Authors Project Team Members

Transcription

1 1 Innovation Potential in Priing and Produt Line Design of SBB Based on the Inreasing Heterogeneity of Customers: An Empirial Analysis of Train Setion and Rush Hour Aess February 2017 Authors Salvatore Maione Lisa Maria Shiestel Prof. Dr. Reto Hofstetter Projet Team Members Prof. Dr. Reto Hofstetter Prof. Dr. Rio Maggi Dr. Stefano Sagnolari Salvatore Maione Lisa Maria Shiestel

2 2 1 Contents 2 List of Figures List of Tables Management Summary Introdution Changes from Original Researh Proposal Idea Generation and Seletion Proess SBB Subsription Struture SBB Priing Struture Overview of Innovative Produt Offers by Other Railway Companies in Europe Segmentation on the Train Loyalty Programs Mobility Partnerships Tiket Options and Priing Servie Innovations Ideation Session Seletion of Produt Line Innovations for Empirial Investigation in Stage Methodology of Choie Models Experimental Design Disrete and Hybrid Choie Models Train Setion Aess Projet Literature Review Soial Influene in Transportation Coneptual Framework Soial Identity Theory, the In-Group, and the Out-Group How Soial Identity Drives Travel Choie Other Relevant Psyhologial Construts in Transportation Study Methodology Desription of Sample Results from Choie Modeling Exploratory Analysis of the Out-Group Derogation Trait Summary and Disussion Study Methodology Desription of Sample Results from Choie Modeling Subgroup Analysis Summary and Disussion Reommendations for SBB Rush Hour Aess Projet Related Literature Our Researh and Pratial Contribution Study Methodology Desription of Sample Results from Choie Modeling Subgroup Analysis Summary and Disussion Reommendations for SBB Conlusion Referenes Appendix... 59

3 3 2 List of Figures Figure 1 Stage-Gate Produt Innovation Management Proess... 6 Figure 2 General Representation of the Hybrid Choie Model Figure 3 Out-Group Derogation Sale (Studies 1 and 2) Figure 4 Full Path Diagram of the Hybrid Choie Model (Study 1) Figure 5 Exemplary Choie Task (Study 1) Figure 6 The Wider the Geographial Aess, the Higher the Utility Figure 7 Aess to the Rush Hour Inreases Utility Figure 8 Aess to the Dediated Setions Inreases Utility Figure 9 The Preferene for a Travel Card Inluding Dediated Setion Aess (vs. Common Setion Aess Only) Inreases With an Inreasing Tendeny Towards Out-Group Derogation Figure 10 Full Path Diagram of the Hybrid Choie Model (Study 2) Figure 11 Hierarhial Struture of the Choie Model (Study 2) Figure 12 Exemplary Choie Task (Study 2) Figure 13 The Wider the Geographial Aess, The Higher the Utility Figure 14 Aess to the Rush Hour Inreases Utility Figure 15 Among the Dediated Setions, the Silene Setion Provides the Highest Utility Figure 16 The Utility of the Dediated Setions Inreases Relative to the Common Setion with Inreasing Out-Group Derogation Figure 17 Full Path Diagram of the Choie Model Figure 18 Exemplary Choie Task Figure 19 Commuters with Formal Constraints Spend More Time at their Commuting Destination than Offiially Required by Formal Constraints Figure 20 The Wider the Geographial Aess, the Higher the Utility Figure 21 Travelers Prefer Seond Class to First Class Figure 22 Unlimited Aess Provides the Highest Utility Figure 23 Commuters Derive a Higher Utility from All Aess Options to the Rush Hour Figure 24 Desription of Train Setion Aess (Train Setion Aess Projet, Survey 1) Figure 25 Desription of Train Setion Aess (Train Setion Aess Projet, Survey 2) Figure 26 Desription of Rush Hour Aess (Rush Hour Aess Projet)... 66

4 4 3 List of Tables Table 1 Top Ideas Suggested in Ideation Session Table 2 Attributes and Attribute Levels for Train Setion Aess Projet (Study 1) Table 3 Cronbah s Alpha of Latent Variables (Study 1) Table 4 Attributes and Attribute Levels for Train Setion Aess Projet (Study 2) Table 5 Cronbah s Alpha of Latent Variables (Study 2) Table 6 Attributes and Attribute Levels for Rush Hour Aess Projet Table 7 Cronbah s Alpha of Latent Variables Table 8 Top Ten Ideas Aording to the Degree of Innovativeness Table 9 Top Ten Ideas Aording to the Expeted Utility for the Target Segment Table 10 Top Ten Ideas Aording to the Impat on the Brand Image Table 11 Top Ten Ideas Aording to the Degree of Innovativeness, Expeted Utility for the Target Segment, Impat on the Brand Image Table 12 Correlation Between Latent Construts (Train Setion Aess Projet, Study 1) Table 13 Relation of Out-Group Derogation to Other Psyhographis (Train Setion Aess Projet, Study 1) Table 14 Relation of Out-Group Derogation to Travel-Related Variables (Train Setion Aess Projet, Study 1) Table 15 Relation of Out-Group Derogation to Seletion of Demographis (Train Setion Aess Projet, Study1) Table 16 Correlation Between Latent Construts (Train Setion Aess Projet, Study 2) Table 17 Correlation Between Latent Construts (Rush Hour Aess Projet, Part 1) Table 18 Correlation Between Latent Construts (Rush Hour Aess Projet, Part 2) Table 19 Train Setion Aess Projet (Study 1) Basi disrete hoie model Table 20 Train Setion Aess Projet (Study 1) Hybrid hoie model with out-group derogation 68 Table 21 Train Setion Aess Projet (Study 2) Basi disrete hoie model Table 22 Train Setion Aess Projet (Study 2) Hybrid hoie model with out-group derogation 70 Table 23 Rush Hour Aess Projet - Basi disrete hoie model Table 24 Rush Hour Aess Projet - Disrete hoie model with interation for ommuters... 72

5 5 4 Management Summary The needs of Swiss travelers are beoming inreasingly diverse and nuaned, providing opportunities for the Swiss Federal Railways (SBB) to inrease revenues by offering more targeted and differentiated offers. In this projet, we generated opportunities for innovative travel ard offerings and empirially investigated two showing high promise. We empirially measured travelers preferenes for variations of travel ard offerings and derived poliy reommendations for SBB based on these values. We used a so-alled hybrid hoie approah for this purpose, whih not only allows measuring onsumers preferenes, but also linking them to their psyhographi and demographi harateristis. In the first projet, we examined travelers preferenes for being able to aess speifi dediated setions on the train that fous on one speifi travel need. Examples of suh dediated setions inlude SBB s urrent silent or family wagons. These setions offer the possibility to travel with other traveler groups who share a similar travel need and to stay away from traveler groups with different travel needs. Soial identity theory has already highlighted that individuals have a general tendeny to separate from different others (i.e., from the out-group). The empirial analysis, however, revealed that the average traveler does not neessarily put a higher value on having aess to suh dediated setions. Only ertain speifi groups of travelers value having this type of aess, suh as individuals who feel highly disturbed by dissimilar others on the train. Although quite speifi, these groups of individuals are quite sizeable. For instane, at least 16 perent of travelers would be willing to pay on average CHF more for having additional aess to one of the four dediated setions. These results imply that separately priing these dediated setions may allow the generation of additional revenues and profits for SBB. Nevertheless, a final deision about who these traveler segments exatly are, whih to target, and what prie points to use requires further researh. In the seond projet, we addressed the highly relevant topi of peak-load priing. Speifially, we investigated whether travelers would be willing to travel at times other than rush hour for a redued prie. We find that travelers, in general, do not like to do so as they have a very strong preferene for unlimited aess, inluding during the rush hour. In fat, they are willing to pay CHF more for a travel ard with unlimited aess during the rush hour than for the least preferred travel ard with limited aess. The formal time onstraints travelers fae, e.g., at their workplae, ould explain this result. They simply annot deviate from the rush hour in a meaningful way and this tendeny is exaerbated for ommuters. Although not representing a large traveler segment, we find that some individuals would be fine with limited rush hour aess. These inlude different types of leisure travelers, suh as pensioners and other non-ommuters. Among the different variations of limited ards that ould be offered, we found that travelers do value the number of times they an aess the rush hour and how long the duration of the rush hour is defined. Our reommendation to SBB is thus to be very autious with rush hour priing as travelling during these times is typially a neessity, sine most travelers simply do not have the required flexibility to do otherwise.

6 6 5 Introdution The needs of Swiss travelers are beoming inreasingly diverse and nuaned (Nguyen & Mariani, 2014). This provides an opportunity for the Swiss Federal Railways (SBB), who ould potentially inrease revenues by offering more targeted and differentiated offers. The urrent produt line of SBB is mainly organized around key servie offerings, suh as the Half-Fare or General Abonnement travel ards. These offerings are provided via one single travel ard, the SwissPass (SBB, 2017). Innovations in the produt line should thus ideally fit into the existing logi of the SwissPass, but ould vary the type of servies offered and the breadth and depth of the produt line, aiming at meeting the diverse needs of travelers. In this projet, we generated opportunities for innovative travel ard offerings and empirially investigated two showing high promise. We followed the typial stage-gate proess of produt innovation management in whih large numbers of opportunities are first identified and then sequentially sreened out until the most promising ones remain, whih are subsequently introdued to the market (Terwiesh & Ulrih, 2009). Our projet featured two main stages and gates (see Figure 1). During the first stage, we identified potential innovations in the produt line through desk researh, disussions with stakeholders of SBB, and by interviewing ustomers using an open innovation approah. Two main ideas passed the gate and remained after this first stage. In the seond stage, we empirially investigated ustomers preferenes for these two ideas using a hypothetial hoie approah. Based on the results of this analysis, we also point to the third stage of the produt innovation management proess in whih the atual revenue potential of the seleted ideas would be estimated; however, this third stage was not part of the sope of this projet. Figure 1 Stage-Gate Produt Innovation Management Proess Methodologially, we applied both qualitative and quantitative approahes. In the first stage, we employed a variety of desriptive and qualitative approahes, suh as desk researh, disussion rounds and brainstorming with different internal and external stakeholders, and an ideation session. In the seond stage, we applied both a disrete hoie approah and a so-alled hybrid hoie approah to investigate ustomers preferenes for the two ideas. This approah allowed for the investigation of the role of psyhographis and demographis as drivers of utility. Traditional eonomists have typially used disrete hoie models to analyze deision makers preferenes assuming a ompletely rational behavior. In ontrast, behavioral sientists have foused on understanding how individuals form deisions and in analyzing the nature of the deision-making proess based on psyhologial theory

7 that does not always satisfy suh an assumption of rationality. Sine the results of these two amps were often quite different, researhers have developed more integrative approahes that aount for both perspetives (Ben-Akiva et al., 2002). This hybrid approah allows us to not only estimate attribute utilities, but also the influene of the proposed soio-psyhologial fators on the utilities. Further, it allows using the olleted data for future market potential and revenue analyses, suh as segmentation, targeting, or market simulations. The first idea that passed the gate to the seond stage (Train Setion Aess Projet) addressed the inreasingly rowded trains faed by SBB (Ungriht, 2010) and the aompanied inreasing heterogeneity of traveler types and behaviors on the trains (Saameli, 2014). Currently, SBB provides speial silene, business, and family setions dediated to partiular individual needs. Although these setions provide additional value to the speifi individuals, they are not pried separately. This projet thus investigates whether it ould be sensible for SBB to extend their servie by expliitly offering aess to suh or other setions on the train, potentially against payment. Greater separation of different traveler segments may offer value to ustomers who urrently feel disturbed by the behaviors of other travelers. Even though SBB is not planning to prie these setions separately in the near future, gaining quantitative insight into the degree to whih travelers value these setions is of high interest to SBB (e.g. for servie hardware deisions in the fleet management). Coneptually, the need to aess these dediated setions an be driven by a traveler s need to travel with other travelers of similar kind and to stay away from travelers with different needs. Take a business traveler, for instane, who would rather travel with other business travelers so as not to be disturbed by the noise potentially reated by other travelers (e.g., rying babies). We apture this general tendeny using the latent onstrut of out-group derogation, i.e., individuals tendeny to asribe more negative harateristis to individuals that are not part of the own soial group (Dasgupta, 2004) and test if it moderates travelers preferenes for dediated setions. We hypothesize that a higher (vs. lower) tendeny towards out-group derogation leads to a higher utility for a travel ard with aess to a dediated setion in addition to the ommon setion (i.e., where diverse types of travelers travel together) and a lower utility for a travel ard with aess only to the ommon setion. Two studies test our oneptual hypothesis and provide insight into travelers preferenes for suh a servie. For the average traveler, we find that having a travel ard with aess to dediated setions does not neessarily provide additional value. Only travelers who have a high tendeny towards out-group derogation would be willing to pay CHF more for the dediated setions (average over the all dediated setions, ompared with having aess to the ommon setion only). Individuals high in out-group derogation have a more negative view of traveling and a more workaholi, frustrated, and status-seeking lifestyle. Additionally, we find that the hoie of eah individual dediated setion over any other setion is driven by a set of individual onsumer harateristis. Therefore, separately priing and offering dediated setions ould make sense for SBB, but would require a highly targeted approah. The seond idea that passed the gate to the seond stage led to the reation of the Rush Hour Aess Projet. It takles the problem of highly rowded trains during the rush hour, whih has been raised in several newspaper artiles over the last years (Bloh, 2011; Riedener, 2012; Valda, 2015; Vontobel & Guanziroli, 2008). Currently, the peak hours are between 7:00 and 8:00 in the morning and between 17:00 and 18:00 in the evening (as determined by SBB). However, also the respetive hours before and after the main rush hour still have a higher apaity utilization than the absolute offpeak hours (Bloh, 2011). SBB has already made a suessful effort to shift the passenger load away from the rush hour by requiring its own employees to avoid it, e.g., work from at home (Weihbrodt et al., 2013; Zemp, 2014). In this projet, we examine the potential of a prie inentive to motivate travelers to hoose travel ard offerings that limit traveling during the rush hour. We measured traveler preferenes for travel ards with either unlimited aess (i.e., valid all day, every day), no aess (i.e., not valid during the defined rush hour interval), or some form of limited aess (i.e., not valid during the defined rush hour interval, exept for a limited number of trips) during the rush hour. We find that travelers derive a substantially higher utility from unlimited aess during the rush hour than from limited aess or no aess. They are willing to pay CHF more to have unlimited aess (ompared with having limited aess for 20 trips per year only during 6.00 to 9:00 7

8 and 16:00 to 19:00). This differene inreases even more (to CHF ) for ommuters. Among the travel ards that limit aess to the rush hour, we find that inreasing the number for trips during the rush hour that are permitted with that travel ard by 10 inreases the utility (CHF 495.-), but this inrease appears marginal relative to a differene of CHF between unlimited and limited aess. A narrower definition of the rush hour (7:00 to 8:00 and 17:00 to 18:00 instead of 6:00 to 9:00 and 16:00 to 19:00) inreases utility by CHF for travel ards with limited aess, but also these inreases are small relative to the overall differene between limited and unlimited aess. These results show, in sum, that unlimited aess provides the most utility to travelers, independent of how the limited aess is designed. It seems that travelers pereive limited aess travel ards as highly inonvenient, even in omparison with travel ards that provide no aess to the rush hour. We further observe that those ommuters who fae formal time onstraints have limited flexibility in their traveling times (i.e., the variane in their travel time is signifiantly lower). The results are thus likely driven by external fators and do not only represent intrinsi traveler preferenes. A prie differentiation strategy based on suh extrinsi fators is not neessarily advisable given that onsumers may pereive it negatively and unfair, potentially harming ustomer loyalty (e.g., similar to the negatively pereived monopolist who perfetly disriminates ustomers 1 (Kahneman et al., 1986)). Limiting the rush hour aess by harging a higher prie is not a reasonable alternative to SBB, unless the formal onstraints of their travelers ould be substantially redued. However, we also explore whih traveler harateristis drive the hoie of one of limited aess and no aess over unlimited aess and find that there are ertain traveler harateristis that lead to the hoie of these options. Therefore, there seems to be potential for an introdution of these travel ard options with a highly targeted approah. The rest of the doument is strutured as follows: We first state and motivate the hanges from the original proposal for this projet. Seond, we outline our idea generation proess (stage-gate stage 1) by desribing the SBB starting environment for the projet and the analyses performed to determine the two ideas we investigate in the seond stage. Third, we explain the hybrid hoie methodology. Fourth, we present both the Train Setion Aess Projet and the Rush Hour Aess Projet. Finally, we summarize the results of both projets and disuss the impliations for SBB. 6 Changes from Original Researh Proposal During the projet, the interests of SBB in terms of projet goals and researh questions have signifiantly hanged. Mainly, they shifted away from the original idea of fousing on produt-line priing with an emphasis on the train setion aess towards investigating preferenes for different rush hour options. These hanges have signifiantly inreased the workload assoiated with the projet at the expense of some of the original goals. Together with SBB, we deided to abandon the produtline optimization and priing part of the projet. Instead, we foused on analyzing travelers preferenes for train setion and rush hour aess. Importantly, we olleted the data in suh a way that the produt-line priing analysis ould still be performed. We simply did not ondut it as it had lost relevane for SBB. The speifi hanges relative to the original proposal are as follows. We foused on the following ore questions as outlined in the original appliation: 1.1, Developing an initial understanding of deision behavior of Swiss travelers; 1.3, Illustrate optimization potential for the urrent produt line (but without market simulation); and 1.4, Linking the produt line explorative to ustomer segments (without segmentation analysis). We did not speifially pursue the following goals: 1.2, Defining prie and ross-prie elastiities of produt line; 1.5, General disussion of aess vs. produt-prie strategies. Please again note that 1.2 ould still be performed based on the data we olleted. Although we did not alulate elastiities, we alulated willingness-to-pay (WTP) values for ertain produt bundles and attribute levels of interest. 8 1 Kahneman, Knetsh, and Thaler (1986) have argued that a monopolist s strategy to maximize their profits by applying a perfet prie disrimination may bakfire, sine individuals regard individual-level prie disrimination as highly unfair.

9 Conerning the detailed questions, we foused on question 2.2 with an emphasis on train setion and rush hour aess. We did not speifially investigate 2.1 and 2.3. We further foused on researh question 3.1, i.e., the ontribution of psyhometri variables, whih we further extended. We deided not to further investigate inentive-aligned approahes, as outlined in 3.2, beause the innovative attributes we investigated were purely hypothetial. 7 Idea Generation and Seletion Proess Following the typial stage-gate proess of produt innovation management, we first generated a number of produt innovation ideas and then sreened them out based on several riteria until the most promising ones remained (Terwiesh & Ulrih, 2009). In order to generate ideas, we first analyzed the starting situation of SBB in terms of their subsription (Chapter 7.1) and priing struture (Chapter 7.2) at the beginning of the projet in Then, we reviewed innovative produt offers of other railway ompanies in Europe and ompared them with the SBB offer to identify possible gaps (Chapter 7.3). In addition, we also involved onsumers in the first stage of the produt innovation proess via an ideation session that will be outlined in Chapter 7.4. Based on these analyses, we seleted two innovative produt offers for whih the onsumer preferenes were empirially analyzed in the seond stage. 7.1 SBB Subsription Struture SBB offers a range of produts for both domesti and international trips. Conerning domesti travel, a main distintion is made between individual tikets and subsriptions. SBB addresses individual tikets mainly to travelers who use publi transportation with a low frequeny. Among these produts, there are ordinary tikets (one-way tikets, round trip, multiple journey ards, and itytiket), daily tikets that allow unlimited mobility throughout all of Switzerland for a whole day, or tikets for groups and disounted tikets. In ontrast, subsriptions (i.e., travel ards) are addressed to travelers who frequently use publi transportation. For instane, the Half-Fare travel ard allows travelers to use most of the publi transport in Switzerland at half-fare. All other travel ards vary by geographial aess, time-wise aess, and validity as defined by the age of the traveler. The General Abonnement (GA) travel ard allows unlimited aess to the entire ountry (train, bus, ferries, et.) and thus provides the most omplete aess. The Trak 7 travel ard allows young people (< 25 years old) with a Half-Fare travel ard to have an unlimited number of trips in seond lass from 19:00 to 5:00. The Point-to-Point travel ard provides aess to a speifi route. The regional travel ards were developed by SBB in ooperation with the loal transport fare networks for those who travel frequently in their own fare network. Thus, they provide the same aess as the GA, but on a smaller geographial sope. Finally, the Inter-Regional travel ards are similar to the Point-to-Point travel ard, allowing travels from one point to another, as well as the use of loal transportation in their home and/or destination area. 7.2 SBB Priing Struture The priing struture of SBB is based on travel harateristis (travel lass, period of validity, number of trips, and sope of geographial aess), the type of payment (monthly, annual, et.), and the harateristis of the traveler (age, presene of disability, and family status). To determine the prie range of subsriptions offered by SBB, we onduted a prie analysis using data from the SBB website and internal douments (SBB, 2014). This analysis allowed the definition of plausible values that served as a starting point for the development of the hoie experiments and the derivation of travelers willingness-to-pay (WTP) based on their utilities (in the seond stage of the projet). During the prie analysis, we first olleted data on all available subsriptions and their respetive pries to obtain a general overview of the prie range. The analysis showed that pries for subsriptions ranged from a. CHF for regional subsriptions, suh as Arobaleno, TVBeo, Onde Verte, Zug, Vagabond with 1-2 zones and seond lass to CHF for the first lass GA whih provides unlimited aess to publi transport in Switzerland. 9

10 7.3 Overview of Innovative Produt Offers by Other Railway Companies in Europe At the beginning of the projet in 2014, we reviewed the offers of other railway ompanies in Europe via desk researh. The goal of this desk researh was to identify offers that were new to SBB and at the same time, interesting from a sientifi point of view. After sreening the European railway market, we were able to divide the innovative offers into five ategories: (1) segmentation on the train, (2) CRM and ustomer loyalty, (3) mobility partnerships, (4) tiket options and priing, and (5) servie innovations. In the following, we summarize the most innovative offers with respet to the SBB offering at the starting point of the projet Segmentation on the Train Several railway ompanies have moved away from the lassial system that divides trains merely into first and seond lass and now offer a more fine-grained segmentation into whih travelers an self-selet. For example, the premium trains of Trenitalia have four lasses: (1) standard, allowing to travel at low ost; (2) premium, allowing to travel at high omfort; (3) business, targeted to business travelers; and (5) exeutive, whih is meant for exlusive trips with the highest omfort (Trenitalia, 2014). The Norwegian railway ompany NSB also removed the two-lass system and offers different oahes, inluding NSB Komfort, whih offers the failities and servies for working on the train, NSB Sove, whih offers beds for night trips, NSB Familie, whih offers servies and failities for families traveling with little kids, and NSB Stille, that requires traveling in omplete silene (NSB, 2014). The Frenh railway ompany SNCF lets its ustomers hoose between first and seond lass and between the two different traveling styles: idzap for travelers who want to soialize and idzen for travelers that want to spend their trip in silene (SNCF, 2014). Nuovo Trasporto Viaggiatori (NTV) is the seond train operator in Italy and a ompetitor of Trenitalia. NTV does not distinguish between lasses, but between the three ambienes: Club, Prima, and Smart with different harateristis and servies provided. The Club ambiene offers meals served at the seats, free Wi-Fi, at-seat TV, and private lounges. The Prima ambiene offers meals served at the seats, free Wi-Fi, and a welome servie. The Smart ambiene offers hanging tables (for newborns), self-servie vending mahines, free Wi-Fi, and one Cinema Coah. In addition, travelers an book a larger seat for an additional fee in the Smart XL Coah (NTV, 2014). Finally, between August 1999 and Deember 2004, the German railway ompany Deutshe Bahn (DB) operated the Metropolitan train exlusively on their Cologne-Hamburg route (Wikipedia, 2009). This train was introdued to ompete with air transport by providing high quality produts and extreme puntuality. The wagons of the Metropolitan were omparatively luxuriously designed and unlike any other produt reated by Deutshe Bahn AG. In the first years of ativity, the Metropolitan did not have a division between first and seond lass, but only one lass with three ategories. This hoie emphasized the high quality servies provided. With the deline in sales, the repartition between first and seond lass was introdued to attrat more ustomers. The ategories used in the first years were the Silene wagons, the Offie wagons, and the Club wagons. The Silene wagons were meant for people that wanted to rest. If they wanted, they ould reeive a pillow, blanket, and earplugs for free. There was also a free audio program and people ould reeive headphones if they wished. Moreover, to prevent travelers in the Silene wagons from being disturbed by regular travelers, these wagons were plaed at the end of the train. The Offie wagons were meant for people that wanted to work on the train. There were no extra servies exept for eletriity outlets at eah seat, mobile phone amplifiers, and a oktail and espresso bar. A selet number of four-person seat groups with a table were available for work alls. Finally, the Club wagons were meant for people who wanted to neither work nor sleep. They ould borrow portable DVD players and DVDs. The lub area as well had a oktail and espresso bar and one of the Club wagons was built without barriers Loyalty Programs Several railway ompanies in Europe offer loyalty programs. SJ (Sweden) offers SJ Prio, where ustomers an ollet points, for instane by purhasing tikets, and then use these points to buy new tikets and a list of other items (SJ, 2014). Trenitalia (Italy) and Renfe (Spain) offer loyalty ards that provide various disounts for offered servies (Renfe, 2014; Trenitalia, 2014). In addition, the 10

11 German railway ompany Deutshe Bahn offers a loyalty ard alled bahn.bonus ard whih allows ustomers to ollet points and then use these points to buy new tikets and gifts (Deutshe Bahn, 2014). The Frenh SNCF offers the SNCF Voyageur Programme, whih allows ustomers to ollet points and then use these points for disounts. In addition, the loyalty ard ontains all information regarding the eletroni tikets purhased, eliminating the need to print the tiket (SNCF, 2014) Mobility Partnerships Like SBB, many other railway ompanies in Europe have entered into partnerships with ar rentals, ar-sharing providers, or bike rentals to offer a omplete mobility servie (Deutshe Bahn, 2014; SJ, 2014; Trenitalia, 2014). Deutshe Bahn and the Austrian ÖBB in addition have partnerships with airlines and Deutshe Bahn with travel agenies for booking (Deutshe Bahn, 2014; ÖBB, 2014) Tiket Options and Priing The Swedish railway ompany SJ and the Frenh SNCF have a so-alled Best Buy Calendar (a monthly alendar of day-to-day offers) that ompares tiket pries (SJ, 2014; SNCF, 2014). The Italian railway ompany Trenitalia offers tikets at lower pries when they are booked far in advane (Trenitalia, 2014). The Belgian railway ompany SNCB offers a part-time travel ard for people that travel on the same route two or three times a week (SNCB, 2014) Servie Innovations The German railway ompany Deutshe Bahn provides several servies in the train stations and on trains, suh as Wi-Fi high-speed aess and an entertainment servie (Deutshe Bahn, 2014). The Frenh railway ompany SNCF provides ativities for kids (SNCF, 2014). On its premium trains, Trenitalia offers internet aess through Wi-Fi onboard and multimedia library with entertainment servies (Trenitalia, 2014). 7.4 Ideation Session In addition to the overview of innovative produt offers by other railway ompanies, we also made an effort to ollet innovative ideas diretly from potential travel ustomers. Therefore, we onduted an ideation session between July 2 and July 4, 2014, whih involved 19 partiipants (10 female, 53 perent; Mage = 26.5) reruited from the population of the Università della Svizzera italiana (USI) students via . During the session, partiipants were asked to list ideas regarding possible innovations for SBB relative to their offer in an online form. We performed a parallel searh approah to ontrol for group dynamis that ould redue the number of ideas produed. The best five ideas were eah rewarded a one-day seond lass tiket valid for all trains in Switzerland. Partiipants suggested 85 ideas omprised of 22 ideas regarding offers in the train, 33 ideas regarding tiket options, 8 regarding offers at the trak, 4 regarding web/mobile options, 2 regarding staff, 4 regarding pre/post trip offers, 6 regarding the timetable, 2 regarding the distribution hannels, and 6 unlassifiable ideas. Subsequently, three oders from SBB and USI independently evaluated eah idea using the riteria innovativeness, expeted utility for target segment, and impat on brand image. The ranking of the ideas by these three riteria as well as average over these riteria an be found in the Appendix Tables 8, 9, 10, and 11. Finally, two projet members from SBB 2 have evaluated the generated ideas on four additional riteria: (1) newness to the market, (2) newness to SBB, (3) sales/market potential, and (4) investment osts. Sorting ideas first by newness to SBB, then by sales/market potential, investment osts, and finally by newness to the market, results of the top ten ideas are listed in Table Christopher Nigg and Roberta Marionni (both SBB Personenverkehr, Preis und Sortiment).

12 12 Table 1 Top Ideas Suggested in Ideation Session No. Idea Desription 1 Seats with multimedia aess to wath movies and read newspapers online Introdue new servie that allows people to wath movies and read newspapers using an online platform 2 Printers in business wagon Introdue printers in business wagon to allow ustomers to print their douments during the trip 3 A digital onboard library Introdue multimedia library for ustomers entertainment during the trip 4 Party trips with dane/drink/datingwagon Speial trips with events where ustomers an dane/drink/date with others 5 ipad/pc onboard Introdue tablet and p onboard for people that want to use the "Online" servies 6 Sommelier servie in business wagon Introdue a sommelier servie (personal wine onsultant at seat) in business wagon 7 Book-rental (station-to-station) Introdue a book-rental servie; ustomers an pik up books in a speifi station and return them in the arrival station 8 Massage seats Introdue new type of seats with "massage" funtion 9 Hostess/steward to aompany kids on Introdue speifi support servies for unaompanied kids their travels 10 Musi available via SBB multimedia database traveling Introdue new funtion that allows ustomers to listen musi at seat using dediated headphones (like a private radio) 7.5 Seletion of Produt Line Innovations for Empirial Investigation in Stage 2 Based on all the information we olleted in the previous analyses, we seleted produt line innovations that warranted further analysis. The final seletion deisions were made during USIinternal meetings and meetings with SBB. We proeeded as follows. Based on the overview of innovative produt offers by other railway ompanies in Europe, we identified five possible diretions for innovation: (1) segmentation on the train, (2) loyalty programs, (3) mobility partnerships, (4) tiket options and priing, and (5) servie innovations. Regarding the segmentation on the train (1), SBB already offers silene and business setions in its first lass wagons and family setions in its seond lass wagons, as well as a family oah on ertain routes. Offering these and other setions independent of first and seond lass ould be a produt line innovation for SBB that would allow atering to a more heterogeneous traveler base. This innovation was previously identified at the initial stage of the idea generation proess as being worthy of investigation also from a researh point of view. 3 Both introduing a loyalty program (2) and extending mobility partnerships (3) to airlines and travel agenies was relatively new to SBB s produt line at the beginning of the projet, but would not diretly address the inreasing heterogeneity of onsumers. Therefore, these options were not onsidered further. In 2014, SBB did not offer speial tiket and priing options, suh as disounts, whih made this a possible produt innovation. Sine tiket options and priing (4) were not the fous of the projet as defined by the original researh proposal, this option was also disarded. Nevertheless, offering part-time travel ards was identified as having potential to address traveler heterogeneity in terms of frequeny of travel. Finally, the servie innovations (5) were also new to SBB at the beginning of the projet and thus onsidered further in the idea generation and seletion proess. Further, the ideas generated from the ideation session were mostly related to servie innovations. The suggested servie innovations reate utility for travelers with different harateristis (for example, some are related to business, others to lifestyle). An agreement between the USI and SBB projet team members led to the deision to fous on offers in the train that ater to the needs of 3 This researh interest was identified in an USI-internal meeting with Prof. Dr. Reto Hofstetter, Prof. Dr. Rio Maggi, Dr. Stefano Sagnolari, Salvatore Maione, and Lisa Maria Shiestel on

13 different onsumer groups (Train Setion Aess Projet). 4 More speifially, it led to an investigation regarding utilities of these speial setions on the train and the potential of priing them separately. This is also in line with the outome of the overview of innovative produt offers of other railway ompanies in Europe. In addition to the existing business, silene, and family setions, the suggested ideas prompted the reation of the lifestyle setion. SBB told us that they have already investigated ertain servie add-ons in the ontext of the GA. For this reason, we abstrated from partiular servies and foused on the train setion aess idea. The train setions an inlude ertain servies. We are agnosti to the partiular servie and look at the setion aess only. Further to the ustomer-driven idea generation proess outlined above, the SBB management had a strong interest in investigating one partiular priing-related produt innovation. In several disussion rounds with members of the SBB priing team and the USI projet team, 5 investigating the priing potential of different aess options to the rush hour was identified as a seond researh diretion, sine it would inform ations related to new priing strategies. Based on these disussion rounds, the Rush Hour Aess Projet was reated, whih is the seond projet that will be outlined in this doument. In summary, after the produt innovation proess, the USI-SBB projet team deided to investigate separately two innovative produt offers. The Train Setion Aess Projet investigates a ustomer need-related produt innovation by estimating the utilities of the different train setion aess options. On the other hand, the Rush Hour Aess Projet investigates a management need-related produt innovation by estimating the utilities of different aess options to the rush hour. Together with SBB, we also made important deisions onerning the sope of the projet. Most importantly, we deided together with SBB to fous our analyses on yearly subsriptions. 6 We also deided to onsider the sope of geographial aess as well as the time-wise validity of the travel ards in our researh. 7 Based on the analysis of the priing struture, initially the prie range was agreed to be set to CHF to in order to show meaningful pries (see Train Setion Aess Projet (Study 1) in Chapter 9.2.3). Throughout the ourse of the projet, however, it was agreed with SBB to set the maximum prie in our studies at CHF in order to be able to analyze possible prie inreases 8 (see Train Setion Aess Projet (Study 2, Chapter 9.4) and Rush Hour Aess Projet (Chapter 10)). In stage two of the projet, we then empirially investigated these two ideas. Below, we present the hybrid hoie methodology used and then we outline the onsumer preferene analysis regarding the two innovative produt offers that we hose to examine. 8 Methodology of Choie Models In the following, we first provide a general overview of the hybrid hoie modeling approah we used. We then disuss how we applied the approah to eah of the projets separately This deision was made during an USI internal meeting with Prof. Dr. Reto Hofstetter, Prof. Dr. Rio Maggi, and Salvatore Maione on USI-SBB and onfirmed during a meeting with Stephan Osterwald (SBB Kommunikation; Verkehrsökonomie, Statistik, Forshungszusammenarbeit), Sean Shwegler (SBB Personenverkehr, Preis und Sortiment), Roberta Marionni (SBB Personenverkehr, Preis und Sortiment), Prof. Dr. Reto Hofstetter (USI), Salvatore Maione, and Lisa Maria Shiestel on in Bern. 5 This deision was made during a meeting with Prof. Dr. Reto Hofstetter, Dr. David Blatter (SBB Personenverkehr, Preis und Sortiment), and Dr. Silvio Stiher (SBB Personenverkehr, Preis und Sortiment) on in Bern. 6 This deision was made during the projet kik-off meeting with Stephan Osterwald (SBB Kommunikation; Verkehrsökonomie, Statistik, Forshungszusammenarbeit), Sean Shwegler (SBB Personenverkehr, Preis und Sortiment), Roberta Marionni (SBB Personenverkehr, Preis und Sortiment), Prof. Dr. Reto Hofstetter, and Lisa Maria Shiestel on in Bern. 7 This deision was suggested during an internal USI meeting with Prof. Dr. Rio Maggi, Prof. Dr. Reto Hofstetter, and Salvatore Maione on and onfirmed during a meeting with Stephan Osterwald (SBB Kommunikation; Verkehrsökonomie, Statistik, Forshungszusammenarbeit), Sean Shwegler (SBB Personenverkehr, Preis und Sortiment), Roberta Marionni (SBB Personenverkehr, Preis und Sortiment), Prof. Dr. Reto Hofstetter (USI), Salvatore Maione, Lisa Maria Shiestel on in Bern. 8 This deision was made via ommuniation with Dr. Silvio Stiher (SBB Personenverkehr, Preis und Sortiment) on

14 8.1 Experimental Design Choie behavior an be influened by two different soures: (produt) attributes and (individual) harateristis (Hensher, Rose, & Greene, 2005). Attributes are related to the desription of a speifi produt (type of aess, subsription duration, et.), whereas harateristis are related to the desription of the individual through soio-eonomi variables and latent variables (e.g., measuring attitudes). Conerning the attributes, two main types of hoie data an be olleted aording the hoie literature: revealed preferene (RP) and stated preferene (SP) data. RP data are data olleted from hoies made in real situations, whereas SP data are olleted from hoies made in hypothetial situations. Both data have advantages and disadvantages. The advantage of RP data is that they ontain atual hoies of individuals and the resulting preferene estimates are highly valid. Sine the alternatives shown in an RP study need to be atually sold, they need to atually exist. This is a problem, e.g., for surveys measuring produt innovations, beause the alternatives are typially at the idea or onept stage and not yet developed or ready to be sold. SP studies alleviate this problem but may suffer from biased responses (Miller & Hofstetter, 2009). Individuals may pay less attention to how they respond to SP surveys, as the hoies they indiate are not onsequential. This is less of a problem, however, if respondents are highly involved into the produt ategory (Hofstetter, Miller, Krohmer, & Zhang, 2013) and if relative and not absolute preferene values are interpreted (Miller, Hofstetter, Krohmer, & Zhang, 2011). In our projet, we intended to measure preferenes for hypothetial offerings and thus hose an SP approah. Customers of SBB are typially highly involved with the brand and its offerings, whih is why we believe this approah ould still generate meaningful values that are indiative of onsumers preferenes. Nevertheless, the resulting absolute values (suh as WTP values for alternatives) should be interpreted with aution, as they may suffer from hypothetial bias. For properly setting up a SP hoie experiment, the stages of the so-alled experimental design proess (as explained in Hensher et al., 2005) need to be followed in line with the statistial model that is planned to be estimated. The proess is omposed of eight stages: (1) problem definition, (2) stimuli refinement as omposed of alternative, attribute, and attribute level identifiation, (3) experimental design onsideration, (4) experimental design generation, and (5) alloating attributes to design olumns, (6) generation of the hoie design, (7) randomization of hoie sets, and (8) onstrution of the survey instrument. We have already partly desribed steps one and two in the previous hapter and we will go into detail again for eah individual projet later in the doument. We will also outline steps six to eight later per individual projet. Therefore, we will only outline step 3 here. In the experimental design onsideration stage (stage 3), a distintion between unlabeled and labeled experiments must first be made. Unlabeled experiments adopt generi titles for the alternatives (i.e., they are alled alternative 1, alternative 2, et.). They are used when the objetive of the researh onerns the evaluation of individual attributes and the number of possible alternatives is large. In labeled experiments, speifi labels are assigned to eah alternative (i.e., brands, produt names, et.). Labelled experiments are used when the alternatives have different attributes (some of them might not be available for some alternatives) and/or different ranges of levels for one or more attributes. Moreover, they make the hoie more realisti. In order to avoid assumptions made by a deision-maker on omitted attributes based on the labels attahed to the alternatives (de Bekker-Grob et al., 2010) but retain the flexibility of a labeled experiment, we adopted an approah in between the labeled and the unlabeled. In the seond study of the Train Setion Aess Projet and the study of the Rush Hour Aess Projet, alternatives have alternative-speifi attributes and generi labels (i.e., alternative 1). In the first study of the Train Setion Aess Projet, we adopted an unlabeled hoie experiment. The experimental design onsideration stage (stage 3) also involves the definition of the number of hoies that a deision-maker has to fae. This number depends on the type of fatorial design hosen. Literature here makes the main distintion between full fatorial designs and frational fatorial designs (Hensher et al., 2005; Rose and Bliemer, 2009). Full fatorial designs use all possible ombinations (hoie tasks) oming from the number of attributes and attribute levels adopted in the hoie experiment. Inreasing the number of alternatives, alternative attributes, or attribute levels 14

15 dramatially inreases the number of ombinations (based on the formula L MA, where L is the number of attributes levels, M is the number of alternatives in the hoie set, and A is the number of attributes per alternative), and onsequently the ognitive burden on respondents. Frational fatorial designs, instead, use only a subset of the total number of ombinations. Hene, eah respondent faes only a (random) seletion of ombinations from the total number of possible ombinations. There exist three ategories of frational fatorial designs as defined by Hensher et al. (2005) and Rose and Bliemer (2009): random designs, orthogonal designs, and effiient designs. In our studies, we adopted an effiient design seleting subsets of ombinations that minimize the standard errors assoiated with the parameter estimates (Huber & Zwerina, 1996). To redue further the number of ombinations shown to respondents and the related ognitive burden, we adopted a bloking tehnique that breaks down the design into different segments (bloks). Bloking a design does not redue its size; rather it only redues the number of hoie sets a respondent has to fae. All the hoie designs have been developed using Ngene software (ChoieMetris, 2014). 8.2 Disrete and Hybrid Choie Models We applied a so-alled hybrid hoie modeling approah. As hybrid hoie models are an enhanement of disrete hoie models, we will first give an overview over the disrete hoie methodology followed by the hybrid hoie methodology. From an eonomi perspetive, hoie models have been widely used within the ontext of random utility theory to predit hoie behavior (Ben-Akiva et al., 2002). These models assume that an individual hooses the alternative that maximizes his/her utility between a set of alternatives based on his/her soio-eonomi harateristis and on the attributes of the alternatives available (Tversky & Kahneman, 1975). Disrete hoie analysis (DCA) analyzes the hoie proess that involves disrete alternatives that are mutually exlusive. It bases on the random utility theory (RUT) proposed by Thurstone (1927). MFadden (1973) expanded it to a multiple hoies version. The underlying idea of RUT is that individual s utility assoiated with eah alternative is a variable not diretly observable by researhers (latent variable) and the observed hoies are indiators (manifestations) of these utilities (Ben-Akiva et al., 1999). These utilities, one for eah alternative available during the hoie proess, an be divided into two parts: systemati and random. The former inludes the explanatory variables, suh as alternative attributes and individual harateristis, whereas the latter takes into aount all the unidentified fators that influene hoies (Boldu & Alvarez-Daziano, 2010). Conerning the random part, Manski (1977) has identified several soures of unertainty: attributes of alternatives not inluded, unobservable hanges in respondent s taste, measurement errors, and instrumental variables. This unertainty arises from the partial and inomplete information that the researher has about the fators that drive the deision-making proess. This implies that hoies ould be explained using a probabilisti framework only. Based on the assumptions assoiated with the distribution of the random part of the utility funtion, it is possible to derive different hoie models (Ben-Akiva & Lerman, 1985). One of the models is the multinomial logit model (MNL) (MFadden, 1973), whih assumes the homogeneity of preferenes aross respondents and that the variane of the random part is independent and identially distributed. In the ontext of disrete hoie models, the study of preferenes fouses on the analysis of available options information and individual onstraints. Due to their fous, several literatures ritiized these models for the inability to desribe the heterogeneity of human behavior. In partiular, soiologists, behavioral eonomists, and ognitive psyhologists have demonstrated that individuals often violate most of the assumptions defined by rational behavior (Kahneman & Tversky, 1979, 1984). Behavioral sientists have foused on understanding how deisions are formed and in analyzing the nature of the deision-making proess based on psyhology theory that not always satisfies that assumption of rationality. MFadden (1986) aknowledged already in the 1980 s that there are several interations of psyhologial fators when an individual hooses an alternative. Forward (2004) and Domarhi, Tudela, and González (2008) have shown that the presene of a relationship between attitudes and behavior an help to understand the deision proess that generates hoies. In this ontext, behavioral hoie analysis fouses on the deomposition of the deision-making proess and 15

16 the identifiation of potential irregularities, whereas preditive hoie models have the goal of prediting the respondent s deisions highlighting regularities in the proess (Ben-Akiva, et al., 2002). Therefore, in order to redue the gap between the two fields mentioned above, researhers in the field of disrete hoie models have worked on the development of these models in order to introdue the proess of preferene formation as well as psyhologial fators that determine human deisions (Bentler, 1980; Jöreskog, 1970; Keesling, 1973; Wiley, 1973). Researhers have developed more integrative approahes that aount for both perspetives (Ben-Akiva, et al., 2002). For instane, substantial researh has been onduted related to inorporating psyhographis into disrete hoie models. This integration is possible using so-alled hybrid hoie models, a methodology in whih latent (unobservable) variables are diretly inorporated in the hoie model via observable variables alled indiators (Ben-Akiva, et al., 1999; Walker, 2001). In partiular, three different methods have been identified to introdue psyhologial latent variables in a disrete hoie model (Walker, 2001): (1) introduing psyhologial indexes diretly in the utility funtions (Green, 1984; Harris & Keane, 1998); (2) performing a fator analysis on psyhologial indies and introduing the fitted variables obtained in the utility funtions (Madanat, Yang, & YING-MING, 1995); and (3) inferring latent attributes of the alternatives as well as onsumer preferenes from hoie data and, then, using indiators to explain latent variables (Elrod, 1991; Elrod & Keane, 1995). In this work, psyhologial indexes have been diretly added in the utility funtions of the disrete hoie framework onsidering these variables soure of heterogeneity. Figure 2 shows a general representation of a hybrid hoie model (Walker & Ben-Akiva, 2002). Figure 2 General Representation of the Hybrid Choie Model 16 Tehnially, latent variables are inorporated through an extension of the standard random utility model. The utility values U are latent variables and preferene indiators y are manifestations of these underlying utilities (Walker & Ben-Akiva, 2002). The utility equation for eah deision-maker n and alternative i an be written as: U in = V(X in ; β) + ε in (1) where Uin represents the benefits of alternative i [I = 1,, Jn] pereived by deision-maker n [n = 1,, N] (Un is a vetor of utilities for deision-maker n); Xin is a vetor of explanatory variables desribing harateristis of alternative i and harateristis of deision-maker n (Xn is a matrix of explanatory variables desribing all alternatives harateristis that deision-maker n has faed); β is a

17 vetor of unknown parameters; V is a funtion of the explanatory variables Xin and the unknown parameters β; εin is a random disturbane for i and n, where εn ~ D(θε) with θε as unknown parameters. RUM assumes that eah individual hooses a speifi alternative if, and only if, this alternative maximizes his/her own utility. In other words, deision maker n hooses alternative i if, and only if, the utility of this alternative (Uin) is greater than or equal to the utility of an alternative j (Ujn) for all the alternatives faed by n. Based on this riterion, the hoie probability equation is: P(i X n ; β, θ ε ) = Prob[U in U jn, j C n ] (2) Eah alternative has a speifi indiator yin that assumes values: (0,1); 1 when deision maker n hooses that alternative (i), 0 for all the other alternatives present in the set of alternatives Cn faed by n. From the previous equations, the following likelihood funtion is derived: 17 P(y n X n ; β, θ ε ) = P(i X n ; β, θ ε ) y in i C n (3) Conerning the aforementioned latent variables, there are tehniques that allow the inorporation of these variables in the model. The hypothesis is that, although these latent onstruts annot be observed and measured diretly, their effets on indiators (measurable variables) an be observed. The objetive is to introdue expliitly these relationships providing information on the unobservable variables, suh as attitudes and pereptions, in the model. By introduing the latent onstruts, the generalized deision model beomes a two-omponent model: the hoie model, previously defined, and additionally the latent variables model. The benefits an be represented, inluding the matrix X*n of unobservable explanatory variables in Equation 4, as follows: U n = V(X n, X n ; β) + ε n (4) Assuming that these latent variables were added, the onditional probability of yn given X*n would be defined as: P(y n X n, X n ; β, θ ε ) (5) To obtain the unonditional probability of yn, the latent variable strutural model must be defined and, afterwards, the onditional probability must be integrated over the distribution of these latent variables. Therefore, the latent variable strutural model is: X n = X n (X n ; λ) + ω n (6) where: the latent variable X*n is a funtion of a set of explanatory variables diretly observed X*n, a set of parameters λ and a disturbane ωn ~ D(θω). From this model, the density funtion of the latent variables, f(x* Xn; λ, θω), ould be derived and used to alulate the unonditional probability equation from Equation 5: P(y n X n ; β, λ, θ ε, θ ω ) = P(y n X n, X n ; β, θ ε )f(x X n ; λ, θ ω ) dx (7) Unfortunately, the estimation of this model is diffiult due to the latent psyhologial fators. Therefore, psyhometri data are used as indiators of these unobservable onstruts through the latent variable measurement model: I n = I(X n ; α) + ν n (8) where: In represents a vetor of indiators related with the unobservable onstruts; X*n represents a matrix of latent variables; α is a set of parameters; νn is a disturbane, νn ~ D(θν). From this model, as done before with Equation 6, the density funtion of the indiators is f(in X*; α, θν). Integrating this

18 density funtion in the previous model (Equation 7), the final form of the integrated hoie and latent variable model is: P(y n, I n X n ; β, α, λ, θ ε, θ ν, θ ω ) = P(y n X n, X n ; β, θ ε )f(i n X ; α, θ ν )f(x X n ; λ, θ ω ) dx (9) This methodology has already been used for many appliations, e.g., for modelling interity mode hoies, inluding omfort and onveniene as latent variables (Morikawa, Ben-Akiva, & MFadden, 2002) or for modelling the rereational hoies of young people involving attitude towards alohol onsumption as unobservable onstrut (Sagnolari & Maggi, 2010). This survey and the relative data analysis above allow insight into the dynami relationships between existing and innovative produt offers in the SBB produt line. Unlike previous studies, where the fous is to analyze the interdependenies between different Mobility Tools (Ciari, Marmolejo, Stahel, & Axhausen, 2014; Sott & Axhausen, 2006; Vrti, Shüssler, Axhausen, & Erath, 2011), the aim of this projet is purely on the different offers of SBB in the publi transport senario. 9 Train Setion Aess Projet In the Train Setion Aess Projet, we investigate whether it ould make sense for SBB to separately prie aess to dediated setions on the train. By definition, publi transportation leads to soial enounters. Espeially during rush hour, travelers often need to stay physially lose to one another in rowded trains. Travelers are heterogeneous and differ on a multitude of dimensions, suh as their behaviors on the train, their needs on the train, or their trip purposes. For some travelers, this diversity itself may reate value. For others, it may reate disutility sine they feel restrited or at unease due to other people that show different behaviors on the train. These individuals may prefer separate setions on the train, inluding business, silent, family, or other setions that allow them to travel with alike individuals and stay separate from others with different travelling needs. We test this possibility in this projet and ask the following researh question: Do dediated train setions that reate separate spaes for people with different travel needs and habits provide value to travelers? To investigate this researh question, we measured and analyzed travelers preferenes for travel ards with aess to either ommon (i.e., setion for all kind of travelers) or ommon plus dediated (i.e., setion for travelers with ertain needs) setions. We examined how both demographis, speial traveler harateristis, and psyhographis drive the preferene for dediated setions. Most importantly, we investigate the explanatory power of the latent onstrut of out-group derogation, i.e., a tendeny to asribe negative harateristis to individuals that are not part of one s own soial group (Dasgupta, 2004). We expeted that individuals that have a high tendeny towards out-group derogation would show a higher preferene for the dediated over the ommon setion only. If this is the ase, providing targeted offers with dediated setion aess ould be a viable strategy for SBB to inrease revenues. In the following setions, we will first shortly review the literature on soial influene. Seond, we will introdue the oneptual framework of the projet. Third, we will give a short overview over other latent onstruts used in transportation researh. Fourth, we will outline the researh design and report the results of the two studies we onduted. Finally, we will summarize the results and offer reommendations for SBB. 9.1 Literature Review Soial Influene in Transportation Transportation hoies are not only influened by individuals intrinsi preferenes, but also by their soial environment. Soial influenes have manifested in several domains of transportation hoie. A substantial body of researh has already investigated the relative strength of soial influene ompared with other fators driving travel hoie (Ferdous, Pendyala, Bhat, & Konduri, 2011; Gaker, Zheng, & Walker, 2010; Kinateder et al., 2014; Sherwin, Chatterjee, & Jain, 2014; Von Sivers, Templeton, Köster, Drury, & Philippides, 2014). Others have explored the role of soial identity 18

19 (Fielding, MDonald, & Louis, 2008; Lois, Moriano, & Rondinella, 2015) and soial norms (Dieplinger & Fürst, 2014; Forward, 2009; Huth & Gelau, 2013; Paris & Van den Brouke, 2008; Riggs, 2017; Shade & Shlag, 2003; Zhang, Shmöker, Fujii, & Yang, 2016) when it omes to traveling. None of this prior researh, however, has looked into the preferenes related to the soial mixing of individuals on the train, whih is the fous of our investigation. 9.2 Coneptual Framework Coneptually, we utilize soial identity theory (Tajfel & Turner, 1979) and researh related to intergroup biases (Hewstone, Rubin, & Willis, 2002). People generally favor the soial group they belong to (the in-group) and try to distinguish themselves from other soial groups (the out-groups) (Dasgupta, 2004; Hogg, 1996; Turner, Hogg, Oakes, Reiher, & Wetherell, 1987). In a similar vein, travelers also belong to different travel-related groups. We propose that travelers who have a higher tendeny to distinguish themselves from different soial groups on the train are more likely to derive utility from being able to travel in a spae where they are separated from their out-group Soial Identity Theory, the In-Group, and the Out-Group For a speifi individual, the in-group refers to the group that the individual is a member of and the out-group to the group he/she is not a member of (Hewstone et al., 2002). This an be a ultural group, a religious group, or more generally, a so-alled psyhologial group, whih is based on psyhologial traits, that are independent of external features. Groups often share beliefs, behavioral norms, and expetations, whih an be learly distinguished from other groups (Efferson, Lalive, & Fehr, 2008). How distintive two different groups are depends on the pereived dissimilarity between those groups on a speifi omparison dimension (Jetten, Spears, & Manstead, 2001). Soial identity theory explains how individuals identify their membership of a ertain group. Aording to this theory, psyhologial group formation is a proess of self-ategorization. Selfategorization is an individuals attration towards a group based merely on the group label and not on the interation with individuals of the group (Hogg & Turner, 1985). This proess is thus based on group-prototypes and not on an individual basis (Hogg, Hardie, & Reynolds, 1995). A prototype of a speifi group is an abstrat representation of average, stereotypial, or normative harateristis of group members in omparison with the average harateristis of non-group members (Hogg et al., 1995). Group prototypes are formed through depersonalization, a proess in whih ognition, pereption, and behavior are defined by the standards of the group, i.e., the group norms, group stereotypes, and group prototypes instead of by personal standards (Hogg & Hardie, 1992; Hogg et al., 1995; Turner, Hogg, Oakes, Reiher, & Wetherell, 1987). Individuals of a ertain group, usually the in-group, are then liked not beause of their individual harateristis, but rather beause they are the embodiments of the group and the more similar an individual is to the group prototype, the more it is liked (Hogg & Hains, 1996). Davis (1984) shows that individuals attration to a group inreases in relation to the level of similar attitudes of the group and the individual. How similar attitudes are pereived depends on a multitude of fators, suh as external fators that inlude the evaluations of the group s intelligene (Davis, 1984). Hene, how muh a person an identify with a partiular group diretly influenes soial attration. Individuals have a general tendeny to asribe more positive harateristis to their in-groups than to their out-groups (Dasgupta, 2004). This phenomenon is alled in-group favoritism and it is omnipresent in human soieties (Lewis & Bates, 2010). On the other hand, out-group derogation desribes the opposite perspetive of the same phenomenon, i.e., that people have a tendeny to asribe negative harateristis more easily to their out-groups (Dasgupta, 2004). This phenomenon is also often alled intergroup bias, whih is defined as a more favorable evaluation of the in-group ompared with the out-group (Hewstone et al., 2002; Levin & Sidanius, 1999). Intergroup biases influene judgments, deisions, and behaviors (Dasgupta, 2004). This leads to individuals feeling onneted to their in-group and wanting to separate from their out-groups (Hogg, 1996; Turner et al., 1987). Intergroup biases are positively influened by intergroup ompetition sine it renders the distintion between in- and out-group more salient (Brewer, 1979). Literature also evidened that individuals differ in their tendeny towards in-group favoritism and out-group derogation (Jost, Glaser, Kruglanski, & Sulloway, 2003; MGregor, Haji, & Kang, 2008; Stangor & Thompson, 2002). 19

20 9.2.2 How Soial Identity Drives Travel Choie We argue that the in- and out-groups do matter to travelers on the train. Eah traveler is member of a speifi travel-related soio-psyhologial group as defined by psyhographi and demographi harateristis that may differ from the soio-psyhologial groups of other travelers on the train. Travelers who do not belong to the same group (i.e., are onsidered as being part of the outgroup), may be pereived as disturbing. In partiular, they might interfere with a person s individually preferred way of traveling. This leads to a general tendeny to separate oneself from the out-group (Brewer, 1979). In addition, the mere awareness of different groups being present on the train as signaled by different behaviors may lead to some form of intergroup onflit (Tajfel & Turner, 1979). Thus, mixing suh different travelers together may ause dissatisfation. The reation of speial setions on the train that are dediated to a speial group of travelers allows travelers to self-ategorize into their respetive soial groups. The label given to the train setion (i.e., business, lifestyle, family, or silene) funtions as an indiator of the soial group, allowing travelers to draw inferenes about the prototype person in that setion. Based on the selfategorization proess, travelers should thus lassify themselves into the setion to whih they feel most soially attrated to in order to minimize intergroup onflits. Therefore, we propose: H1a: When hoosing between travel ard offerings, travelers have a higher utility for travelards that offer additional aess to dediated setions (vs. aess only to ommon setion). On the train, travelers are relatively unlikely to engage in ative interpersonal interation. They onsider sitting with members of the in-group as being less important than sitting separately from outgroup members. Hene, they hoose a train setion based on who they do not want to sit with (negational ategorization) rather than who they want to sit with (affirmative ategorization). Where affirmative ategorization relates to in-group favoritism, negational ategorization relates to out-group derogation (Zhong, Phillips, Leonardelli, & Galinsky, 2008). Consequently, the stronger the tendeny towards out-group derogation, the stronger the desire to separate from the out-group, and the higher the likelihood to hoose a travel ard with dediated setion aess (Ajzen, 1985). Therefore, we propose: H1b: When hoosing between travel ard offerings, travelers with a high (vs. low) tendeny towards out-group derogation derive a higher utility from travel ards that offer additional aess to dediated setions (vs. aess only to ommon setion) Other Relevant Psyhologial Construts in Transportation The transportation literature has already investigated the role of numerous psyhologial onstruts for transportation hoies; however, none has foused on out-group derogation. These onstruts are typially introdued into statistial models in the form of latent variables. They omprise attitudes, pereptions, lifestyle variables, and personality traits. We briefly summarize these onstruts and the related researh below. An attitude is a feeling or opinion about something or someone, or a way of behaving that is aused by this (Cambridge Ditionary, 2017a). Attitudes have shown impat on various deisions in the transportation field. Redmond and Mokhtarian (2001) showed that having a positive pereption of travel helped to explain an inreased utility of ommuting. On the ontrary, they found that a higher family or ommunity orientation led to a lower ommuting utility. Choo and Mokhtarian (2004) investigated the effet of, among other things, attitudes on the hoie of ar type and found that individuals who dislike traveling more would rather drive luxury ars and that individuals who prefer to live in urban neighborhoods prefer smaller ars. Johansson, Heldt, and Johansson (2006) showed that attitudes toward omfort and flexibility influened hoies of transportation mode. Boldu, Bouher, and Alvarez-Daziano (2008) indiated that introduing appreiation of new ar features as a variable into a model for new ar hoies in Canada inreased explanatory power. Kamargianni and Polydoropoulou (2013) found that students attitudes toward walking and yling had an impat on the hoie of these modes. Kuppam, Pendyala, and Rahman (1999) even disovered that attitude variables have a greater explanatory power than demographi variables in their mode hoie. 20

21 A pereption is a belief or opinion, often held by many people and based on how things seem (Cambridge Ditionary, 2017). For example, the pereptions of walkability of the route to work influene the mode hoie for the shool ommute of teenagers (Kamargianni & Polydoropoulou, 2013). Redmond and Mokhtarian (2001) found that introduing pereived ommute benefits as a variable into a model estimating the utility of ommuting inreased explanatory power. A lifestyle is someone s way of living; the things that a person or partiular group of people usually do (Cambridge Ditionary, 2017b). Travelers with different lifestyles tend to make different deisions. Choo and Mokhtarian (2004) found that less frustrated individuals were more likely to drive luxury ars, workaholis (status seekers) were less (more) likely to drive small ars and sports ars. Finally, those who like traveling by ar were more likely to invest more into ars. However, they stated that the diretion of ausality is ambiguous. Redmond and Mokhtarian (2001) showed that status seeking and being a workaholi helped to explain a higher utility of ommuting. Temme, Paulssen, and Dannewald (2008) disovered that with an inreasing desire for flexibility, the propensity to avoid publi transport also inreased. On the other hand, with the inreasing importane of a onvenient and omfortable ommute, the propensity to use publi transport inreased. A personality trait is a way of behaving or a tendeny that is typial for a speifi person (Myers, 2008). Suh traits an inform transportation behavior in different ways. For instane, Johansson et al. (2006) found that being pro-environmentally inlined influened travel mode hoie. Similarly, Boldu et al. (2008) showed that environmental onern helped to explain new ar hoies in Canada. Further, Bekhor and Albert (2014) investigated the impat of a different personality trait, sensation-seeking, and showed that it influened route hoie. 9.3 Study 1 The goal of Study 1 was to test our two hypotheses (H1a and H1b) related to the basi effet of offering dediated setions (vs. only ommon setions) and the moderating role of out-group derogation Methodology Choie Design In order to maximize the rihness of the data without reating a omplex and elaborate hoie experiment, we developed a bloked hoie design. The design onsists of five bloks with 12 hoie tasks per eah blok to ensure an equal distribution of attribute levels within eah attribute. Respondents were randomly assigned to one of five bloks. In eah blok, the order of the hoie tasks was randomized. In eah task, we inluded three produt alternatives and a non-hoie option. We used an unlabeled approah, meaning labels were not used to identify the alternatives, and eah is the ombination of all attributes listed below. Eah alternative had four attributes: train setion aess, geographial aess, travel during rush hour (7:00 8:30 and 17:00 18:30) 9, and prie (see Table 2 for details). The train setion aess attribute desribes setions on the train to whih eah traveler has aess through a travel ard. This attribute had two levels: ommon setion aess and ommon + dediated setion aess. Common setion aess defines aess to the ommon setions of the train. These setions are aessible by any traveler holding a valid tiket. The level ommon + dediated setion aess defines aess to dediated setions of the train (in addition to aess to the ommon setion). These setions differ from eah other in order to satisfy the needs and habits of the travelers. There are different setions dediated to the traveling needs and habits related to (a) business, (b) silene, reading, and relaxing, () families, and (d) lifestyle, musi, and media. With a ommon + dediated setion aess travel ard, it is possible to hoose a setion and hange between setions during eah trip. The type of people present in the different setions produes the 21 9 The deision about this rush hour interval was made during a meeting with Stephan Osterwald (SBB Kommunikation; Verkehrsökonomie, Statistik, Forshungszusammenarbeit), Sean Shwegler (SBB Personenverkehr, Preis und Sortiment), Roberta Marionni (SBB Personenverkehr, Preis und Sortiment), Prof. Dr. Reto Hofstetter (USI), Salvatore Maione, and Lisa Maria Shiestel on in Bern.

22 substantial differene between the two ategories of aess. Travelers with ommon setion aess annot aess the dediated setions, whereas travelers with ommon + dediated setion aess have omplete aess to all train setions. We will refer to the ommon + dediated setion aess as simply dediated setion from now on for simpliity reasons. We provide the detailed instrutions we provided also to respondents in Appendix Figure 24. The geographial aess attribute desribes the geographial extension in whih the subsription is valid. It had six levels: area small (zone) identifies aess to an area large as two zones; area medium (region, anton) identifies aess to an area as large as an entire region/anton; route (> 10 km) identifies aess to a Point-to-Point subsription longer than 10 km; area small (zone) + route (> 10 km) identifies aess to an area as large as two zones and aess to a Point-to- Point subsription longer than 10 km; area medium (region, anton) + route (> 10 km) identifies aess to an area as large as an entire region anton; and area big (ountry) represents aess to publi transport throughout Switzerland. The travelling during rush hour (7:00 8:30 and 17:00 18:30) attribute desribes the subsription validity during the rush hour and it had two levels: no identifies a subsription valid outside the rush hour only and travelers holding that travel ard are not permitted to take the train during rush hour; yes identifies a subsription with no time restritions and it an be used during rush hour. The prie attribute desribes the eonomi outlay to purhase a speifi travel ard. It had four levels ( CHF , CHF , CHF , and CHF ), whih over the range of pries of existing produts provided by SBB. Table 2 Attributes and Attribute Levels for Train Setion Aess Projet (Study 1) Attribute Number of levels Desription Train setion aess 2 Common setion aess; ommon + dediated setion aess (further referred to as dediated setion for simpliity reasons) Geographial aess 6 Area small (zone); area medium (region, anton); route (> 10 km); area small (zone) + route (> 10 km); area medium (region, anton) + route (> 10 km); area big (ountry) Traveling during rush hour 2 No; yes Prie 4 CHF ; CHF ; CHF ; CHF Latent and Other Variables We began developing the sale to measure the tendeny towards out-group derogation based on an existing sale that measures the tendeny towards in-group favoritism related to different soial groups by Lewis and Bates (2010). This sale uses three items related to the strength of identifiation with the group, the preferene for affiliating with in-group members, and the importane plaed on marrying within the group items 10 of whih the first two are partiularly relevant for us. Based on these two items, we extended the sale to additional soial groups, suh as eonomi groups, ultural groups, and status groups. We did so in order to be able to measure the onept in general and not only based on one single group dimension. Additionally, we generated new items to measure in-group favoritism, partiularly in the train ontext, i.e., regarding traveling needs and behaviors, in order to aount for the partiular ontext of our researh. Based on the disussions in several fous groups, we adjusted and validated all sale items. Finally, we oneptually reversed the items to render them onsistent with the definition of out-group derogation. That is, instead of asking for a preferene to be with similar people or a preferene for an affiliation with similar people, we asked for a preferene to distane oneself from different people or a preferene not to affiliate with different people. The resulting sale has 12 items, as shown in Figure For example, for the religious group, the items were worded as follows: How losely do you identify with being a member of your religious group?, How muh do you prefer to be with other people who are the same religion as you?, How important do you think it is for people of your religion to marry other people who are the same religion?.

23 23 Figure 3 Out-Group Derogation Sale (Studies 1 and 2) We also measured different trait variables that prior researh has identified as relevant. These measures an later serve as ontrols in the analysis, if needed. We measured travelers lifestyle with four different sales. First, we measured the degree to whih a traveler has a frustrated lifestyle on a four-item sale by Redmond (2000). We also used the four-item sale to measure a family/ommunity related lifestyle as well as the five-item sale to measure a workaholi lifestyle by Redmond (2000). Whether a traveler leads a status-seeking lifestyle was measured on a three-item sale derived from Eastman, Goldsmith, and Flynn (1999). Further, we measured to what extent a traveler values hedonism on a three-item sale developed by Temme et al. (2008). A seven-item sale to measure the time-onsiousness of travelers was derived from Kleijnen, De Ruyter, and Wetzels (2007). We also measured travel-liking (four-item sale), ommute benefits (five-item sale), and travel stress (fiveitem sale) (Redmond, 2000), the desire for omfort (four-item sale), and for flexibility (three-item sale) (Johansson et al., 2006). Finally, we measured whether travelers onsider traveling as utilitarian or hedoni on a three-item sale (Wakefield & Inman, 2003). All the above-mentioned variables were measured on seven-point sales. The Cronbah s alphas of all sales are reported in Table 3 and the orrelations between the latent variables are reported in Appendix Table 12.

24 24 Table 3 Cronbah s Alpha of Latent Variables (Study 1) Variable Cronbah s alpha Out-group derogation.95 Family /ommunity oriented lifestyle.67 Workaholi lifestyle.63 Status-seeking lifestyle.90 Frustrated lifestyle.58 Travel-liking.49 Hedonism.84 Time-onsiousness.85 Desire for omfort.67 Desire for flexibility.74 Traveling hedoni (vs. utilitarian).81 Commute benefits.64 Travel stress.79 Additionally, we measured a set of variables that were part of the SBB segmentation studies in order to link to the existing segments possible in follow-up analyses (SBB, 2013). These variables inluded the trip purpose (ommuting to and from plae of work, ommuting to and from an eduational institution, leisure trip without overnight stay (exursion, inema, sports, visit), private vaation trip with at least one overnight stay, everyday tasks (onsultation of a dotor, groery shopping, piking somebody up), business trip, other), the predominant travel lass (first lass, seond lass), the need for ommuting as well as the ommuting habits, and the habits when engaging in leisure trips Model Speifiation and Estimation We applied a hybrid hoie framework (Walker, 2001) to the data we olleted, estimating two different models: a basi disrete hoie model without latent variable (Model 1) and a hybrid hoie model (Model 2). Speifially in Model 2, we ombined a disrete hoie model with the latent variable out-group derogation to one joint model. Figure 4 shows the hybrid hoie model framework we used. This path diagram ompletely desribes the relationships among the explanatory variables and the respetive partial models through a set of strutural and measurement equations. The dashed arrows represent the measurement equations (i.e., the evaluation of the latent variable) and the solid arrows desribe the strutural equations.

25 25 Figure 4 Full Path Diagram of the Hybrid Choie Model (Study 1) In order to have high flexibility in the definition of the hoie model (MFadden & Train, 2000) and to take the panel struture of the dataset into aount (Bierlaire & Fetiarison, 2009), we adopted a mixed logit with a normally distributed error omponent (EC). The basi model, Model 1 (Appendix Table 19), is suh an EC model with the error omponent for the panel struture of the dataset and without the latent variable. In this model, as well as in Model 2, we have three utility funtions (U1, U2, U3) aording to the number of alternatives defined in the hoie design (see Chapter ). The utility funtions are speified as follows: U n1t = β 0 + β 1 Train Setion Aess n1t + β 2 Geographial Aess, area small (zone) n1t + + β 6 Geographial Aess, area medium (region, anton) and route (> 10 km) n1t + β 7 Travelling during rush hour n1t + β 8 Prie n1t + μ Z panel_data,n + ε n1t (10) U n2t = β 0 + β 1 Train Setion Aess n2t + β 2 Geographial Aess, area small (zone) n2t + + β 6 Geographial Aess, area medium (region, anton) and route (> 10 km) n2t + β 7 Travelling during rush hour n2t + β 8 Prie n2t + μ Z panel_data,n + ε n2t (11) U n3t = β 0 + β 1 Train Setion Aess n3t + β 2 Geographial Aess, area small (zone) n3t + + β 6 Geographial Aess, area medium (region, anton) and route (> 10 km) n3t + β 7 Travelling during rush hour n3t + β 8 Prie n3t + μ Z panel_data,n + ε n3t (12) In eah equation, n identifies the respondent, t the hoie task and εnt is a random disturbane over people and hoie tasks (see previous explanation in Chapter 8.2.). The hybrid hoie model, Model 2 (Appendix Table 20), is an extension of Model 1 that uses the latent variable in the utility funtions to understand if it inreases the explanatory power of the model (Walker, 2001). We defined 12 equations for this model, one for eah indiator of out-group derogation. Then, onerning the hoie model, we inluded the latent variable in eah utility funtion as an interation with the train setion aess attribute. We omitted these equations from the doument. We estimated these models using the maximum simulated likelihood (MSL) approah implemented in the free software, PythonBiogeme 2.4 (Bierlaire, 2016). Adopting random draws from the probability distributions of the error omponents and the latent variables, the MSL tehnique an

26 estimate the maximum likelihood by replaing the multidimensional non-losed integral with a smooth simulator (Train, 2003). Following proedures suggested by Ben-Akiva, Walker, et al. (2002) and Bhat (2001), we used Halton draws instead of the uniform random draws in order to speed up the estimation. In fat, at the same level of auray of the simulator, models with Halton sequenes adopt a lower number of draws ompared with models with uniform draws. For our models, we adopted Halton draws and we did not inrease this number for two reasons: the parameter estimates were quite stable and the omputational osts inreased onsiderably Proedure With the help of the market researh ageny Intervista AG, we reruited 207 travelers in the German-speaking part of Switzerland for our survey. Only individuals that were planning to purhase a publi transport subsription in Switzerland or renew their urrent one within the next year and who were paying for their subsription by themselves were eligible to partiipate in the survey. Our sample was stratified by age to represent the Swiss population aording to SBB s marketing researh standards (Swiss Federal Statistial Offie, 2014). Travelers were asked to answer a survey lasting approximately 20 minutes. The survey was strutured as follows. First, they were introdued to the general proedure of the survey. Seond, they were asked a set of questions related to their travel behavior and their travel preferenes as defined by the SBB segmentation study (SBB, 2013). Third, they needed to omplete the hoie experiment. Fourth, they had to omplete questions related to the latent variables. Finally, they had to fill out a setion related to their demographi harateristis. The entire survey was onduted in German. Details onerning the proedure of the hoie experiment are provided below. Before the atual hoie experiment, travelers were familiarized with the onept of a travel ard. In the hoie experiment, eah travel ard provided aess to publi transportation in Switzerland for one year. Different travel ards have different aess options to the publi transport in Switzerland as defined by the previously outlined attributes and attribute levels. Travelers were then asked to imagine that they were about to purhase a new travel ard online and that in the following 12 pages they would fae 12 different purhase situations (i.e., hoie tasks) in whih the website would offer them a different set of travel ards from whih they ould hoose. They were also informed that in eah purhase situation, they would have four purhase options. That is, they ould either hoose one of the three displayed travel ard options or leave the store without purhasing a travel ard by liking None. An exemplary hoie task is shown in Figure 5. In the legend, eah attribute and its respetive levels were again desribed in detail. 26

27 27 Figure 5 Exemplary Choie Task (Study 1) Desription of Sample Of the 207 travelers we reruited, 101 were female (49 perent) and the average age was years. Most of the respondents in the sample owned a Half-Fare travel ard (63 perent), followed by the regional travel ards (23 perent), the general subsription (GA, 22 perent), and the Point-to-Point travel ards (6 perent). Six perent of the sample did not have a urrent subsription. The most frequent purposes of travelling by train of respondents in our sample were leisure trips without overnight stays (80 perent), followed by everyday tasks (46 perent). Forty-five perent of the survey respondents indiated that they were ommuting by train to work and 14 perent of the survey respondents were ommuting by train to an eduational institution. Forty-two perent or the respondents were engaging in vaation trips with at least one overnight stay and 21 perent were taking business trips Results from Choie Modeling Model 1 (Appendix Table 19) shows the oeffiient estimates of the basi model that inluded all attribute levels. Eah oeffiient represents the ontribution that its speifi attribute level has on the overall utility of the alternative. In line with this, the y-axis of all graphs measures the utility ontribution of the attribute level represented. The results show that broader geographial aess provides higher utility to travelers (βarea small = -1.12, p <.01; βarea small + route = -.96, p <.01; βroute = -.46, p <.01; βarea medium=.26, p <.05; βarea medium + route =.31, p <.05). For a better interpretability, the utilities assoiated to the geographial aess are represented in Figure 6. The utilities of the attribute levels orrespond to the values of the relative oeffiients (e.g., Uarea small = βarea small) exept for the attribute referene level. 11 For example, we alulated the utility of the attribute level area big (ountry), referene level of the attribute geographial aess, as follows: U area big = ( 1) β area small + ( 1) β area small + route + ( 1) β route + ( 1) β area medium + ( 1) β area medium + route Plugging in the oeffiient estimates, we obtain the following value: 11 We oded this variable using the so-alled effet oding tehnique (Hensher et al., 2005). With this proedure, we obtained the value of the referene level by summing up the oeffiients (multiplied by a fator of -1) assoiated with other levels.

28 Utility Utility 28 U area big = ( 1) ( 1.12) + ( 1) (.96) + ( 1) (.46) + ( 1) (. 26) + ( 1) (. 31) = 1.96 Being allowed to travel during the rush hour provides higher utility vs. not being allowed to do so (βrush hour = 1.82, p <.01, Figure 7). Higher pries provide a lower utility than lower pries as shown by the prie oeffiient (βprie < -.01, p < ). Having aess to the dediated setion provides a higher utility than having aess only to the ommon setion, onfirming H1a. This differene is only marginally signifiant, however (βtrain setion aess =.15, p =.10, Figure 8). Travelers are willing to pay CHF more for aess to the big area (ountry) than for aess only to the small area (zone). They are willing to pay CHF more for aess to the rush hour. Finally, they are willing to pay CHF more for additional aess to the dediated setions. Figure 6 The Wider the Geographial Aess, the Higher the Utility Area small (zone) -.96 Area small (zone) + route (> 10 km) -.46 Route (> 10 km) Area medium (region, anton) Note: Area big (ountry) represents the referene level for this attribute (alulated as explained above). Figure 7 Aess to the Rush Hour Inreases Utility Area medium (region, anton) + route (> 10 km) Area big (ountry) No rush hour aess 1.82 Rush hour aess Note: Rush hour aess no aess represents the referene level for this attribute (here, fixed to zero). 12 Prie is ontinuous in the model. Therefore, we do not show it graphially.

29 Utility of dediated setion (relative to ommon setion) Utility 29 Figure 8 Aess to the Dediated Setions Inreases Utility Common setion Note: Common setion represents the referene level for this attribute (here, fixed to zero). Model 2 (Appendix Table 20) shows the oeffiient estimates of the hybrid hoie model. Outgroup derogation moderates the effet of the train setion aess on utility (βout-group derogation train setion aess = 1.35, p <.05). This onfirms our hypothesis H1b. The oeffiient for train setion aess is now negative beause the interation with the latent variable ompletely absorbed the positive effet that it had in Model 1 (βtrain setion aess = -.99, p <.05). The marginally signifiant positive utility assoiated to the aess to dediated setion shown in Model 1 was owed to the effet of the interation between out-group derogation and the aess to dediated setion. For a better interpretability, the oeffiients of the train setion aess interated with the out-group derogation are represented in Figure 9. Travelers who have a tendeny to separate themselves from dissimilar other groups (high tendeny towards out-group derogation) are willing to pay CHF more for aess to the dediated setions (vs. only the ommon setion). Figure 9 The Preferene for a Travel Card Inluding Dediated Setion Aess (vs. Common Setion Aess Only) Inreases With an Inreasing Tendeny Towards Out-Group Derogation.15 Dediated setion Low out-group derogation (- 1 SD) Average out-group derogation High out-group derogation (+ 1 SD) Note: ommon setion represents the referene level for this attribute (here, fixed to zero) Exploratory Analysis of the Out-Group Derogation Trait In the hybrid hoie model, we found that out-group derogation moderated the effet of the train setion aess on utility. Therefore, we explored how travelers with a high tendeny towards out-

30 group derogation differ from travelers with a low tendeny towards out-group derogation in terms of psyhographis (latent variables, Appendix Table 13), travel-related variables (Appendix Table 14), and seleted demographis (Appendix Table 15). For this analysis, we performed a median split 13 of the out-group derogation variable. Most importantly, we found that travelers with a high (vs. low) tendeny towards out-group derogation lead a more workaholi lifestyle (t = -1.77, p <.10) and pereive higher travel stress (t = , p <.01). This might explain why they have a stronger preferene for the dediated setions. If they need to work on the train and, at the same time, experiene stress, they feel more bothered by other travelers that behave differently. The result that travelers with a high (vs. low) tendeny towards out-group derogation pereive less ommute benefits supports this notion (t = 3.09, p <.01). They do not see how they ould use ommuting time produtively. Similarly, they are more time-onsious. Thus, they might want to use their time produtively on the train, but the urrent setting does not allow them to do so. The dediated setions, however, would give them the possibility to use their travel time in the way they would like to use it. Finally, the positive relation to a status-seeking lifestyle (t = -3.51, p <.01) onfirms existing literature that stated that high-status groups are likely to have a stronger intergroup bias (Dasgupta, 2004; Levin & Sidanius, 1999). Different from our expetation, we did not find a relation between out-group derogation and travel-related variables. We found, however, that travelers with a high (vs. low) tendeny towards outgroup derogation were younger (t = 3.21, p <.01) and more likely to be students (vs. any other oupation, t = -1.67, p <.10). A possible explanation for these results might be that students want to study on the train and, thus, prefer to separate themselves from others that might disturb them Summary and Disussion We found that the average traveler only has a marginally signifiant preferene for the dediated setion aess, partly onfirming H1a. This effet is moderated by the travelers out-group derogation trait, onfirming H1b. We further found that the explanatory power of the model inreased after inluding out-group derogation as shown by an inrease of 13 points in ρ 2, suggesting that this personality trait an indeed explain travel ard hoies to some extent. Travelers with a low tendeny towards out-group derogation prefer the ommon setion aess, whereas travelers with a high tendeny towards out-group derogation prefer the dediated setion aess. Hene, travelers with a high tendeny towards out-group derogation have a higher preferene for a travel ard that provides the opportunity to travel in a dediated setion of the train. Travelers with a high tendeny towards out-group derogation (one standard deviation higher than the average out-group derogation), are willing to pay CHF more for having aess to both the ommon and the dediated setions (relative to the ommon setion only). In an exploratory follow-up analysis, we found that travelers high in out-group derogation have a more negative image of traveling in general as shown by higher travel stress and the pereption of fewer ommute benefits. Their lifestyle an be desribed as frustrated and workaholi. By offering the dediated setions, they might pereive more value in their travel time, whih inreases utility for them in line with their lifestyle harateristis for whih they are willing to pay more. The finding that they have a more status-seeking lifestyle suggests that they may see having aess to the dediated setions as exlusive. In order to obtain greater insight into the preferenes for speifi types of dediated setions that are implementable by SBB, we onduted a seond study where we investigated the preferenes for dediated setions divided into dediated setions: business, lifestyle, silene, and family. 9.4 Study Methodology Choie Design We used the same blok design as in Study 1 with 5 bloks and 12 hoie tasks per blok. Respondents were again randomly assigned to one of 5 bloks. In eah blok, the order of the hoie The average out-group derogation over the sample is 2.02, the standard deviation is 1.16, and the median lies at 1.75.

31 tasks was randomized. Different from Study 1, the prie attribute range was shifted upward to allow for testing of possible prie inreases. Further, the number of levels of the geographial aess attribute was redued to allow for a more balaned design. The period of the rush hour aess attribute was hanged from the original intervals to the intervals from 7:00 to 8:00 and from 17:00 to 18:00 to be onsistent with the narrow rush hour interval definition of the Rush Hour Aess Projet. We introdued an additional attribute as a speifiation of the ommon setion + dediated setion attribute level. Fifth, eah hoie task had one alternative for the ommon setion only attribute level, two alternatives for the ommon setion + dediated setion attribute level, and the none option. Eah alternative had four attributes: train setion aess, geographial aess, rush hour aess (7:00 8:00 and 17:00 18:00), and prie (see Table 4 for details). The train setion aess aess desribes setions on the train to whih eah traveler has aess through a travel ard. This attribute had two levels: ommon setion only and ommon setion + dediated setion. Common setion only defines aess to the ommon setions of the train only. These setions are aessible from any traveler holding a valid tiket. Common setion + dediated setion defines aess to dediated setions of the train (in addition to aess to the ommon ones). The type of people present in the different setions provides the substantial differene between the two ategories of aess. Travelers with ommon setion only annot aess the dediated setions, whereas travelers with ommon setion + dediated setion have aess both to their dediation setion and to the ommon one. Additionally, alternatives with ommon setion + dediated setion level of the attribute train setion aess have an attribute speifying the type of the dediated setion. This attribute had four levels: business, silene, family, and lifestyle. For simpliity reasons, we will refer to ommon setion + dediated setion as dediated setion from now on. The detailed desription we provided to travelers in the survey is shown in Appendix Figure 25. The geographial aess attribute desribes the geographial extension in whih the subsription is valid. It had three levels: area small (zone) identifies aess to an area large as two zones; area medium (region, anton) identifies aess to an area large as an entire region/anton; area big (ountry) represents aess to publi transport throughout Switzerland. The rush hour aess (7:00 8:00 and 17:00 18:00) attribute desribes the subsription validity during the rush hour and it had two levels: no (outside rush hour only) identifies a subsription valid outside the rush hour only and does not permit the traveler to take the train with that subsription during rush hour; yes (no time restritions) identifies a subsription with no time restritions and it an be used during rush hour. The prie attribute desribes the eonomi outlay to purhase a speifi travel ard. It had four levels ( CHF , CHF , CHF , and CHF ). Table 4 Attributes and Attribute Levels for Train Setion Aess Projet (Study 2) Attribute Number of levels Desription Train setion aess 2 Common setion only; ommon setion + dediated setion (further referred to as dediated setion for simpliity reasons) Dediated setion aess 4 Business; silene; family; lifestyle (speifiation) Geographial Aess 3 Area small (zone); area medium (region, anton); area big (ountry) Traveling during rush hour 2 No; yes Prie 4 CHF ; CHF ; CHF ; CHF

32 Latent and Other Variables As in Study 1, we adopted the same 12-item sale to measure the tendeny towards out-group derogation. In addition, the other latent variables we measured are idential. The Cronbah s alpha statistis are reported in Table 5. The orrelations between the latent variables are reported in Appendix Table 15. Table 5 Cronbah s Alpha of Latent Variables (Study 2) 32 Variable Cronbah s alpha Out-group derogation.94 Family /ommunity oriented lifestyle.56 Workaholi lifestyle.62 Status-seeking lifestyle.60 Frustrated lifestyle.81 Travel-liking.51 Hedonism.90 Time-onsiousness.79 Desire for omfort.55 Desire for flexibility.77 Traveling hedoni (vs. utilitarian).83 Commute benefits.66 Travel stress.78 Additionally, we measured a set of variables that were part of the SBB segmentation studies to link to the existing segments possible in follow-up analyses (SBB, 2013). These variables inluded the trip purpose (ommuting to and from plae of work, ommuting to and from an eduational institution, leisure trip without overnight stay (exursion, inema, sports, visit), private vaation trip with at least one overnight stay, everyday tasks (onsultation of a dotor, groery shopping, piking somebody up), business trip, other), the predominant travel lass (first lass, seond lass), the need for ommuting as well as the ommuting habits, and the habits when engaging in leisure trips Model Speifiation and Estimation Mirroring the analysis of Study 1, we estimated two different models: a basi disrete hoie model without latent variable (Model 1) and a hybrid hoie model (Model 2). In the hybrid hoie model, we ombined a disrete hoie model with the latent variable out-group derogation. Figure 10 shows the hybrid hoie model framework we used with the partial models, measurement, and strutural equations.

33 33 Figure 10 Full Path Diagram of the Hybrid Choie Model (Study 2) We adopted a mixed logit with an error omponent (EC model) for both models as in Study 1. In ontrast with Study 1, in this study, we introdued two error omponents to measure the orrelation patterns between alternatives in the hoie sets obtaining a model similar to a nested logit (Train, 2003). Graphially, the hierarhial struture is represented in Figure 11. Figure 11 Hierarhial Struture of the Choie Model (Study 2) In Model 1 (Appendix Table 21) and Model 2 (Appendix Table 22), we have three utility funtions (U1, U2, U3) aording to the number and the type of alternatives defined in the hoie design (see Chapter ). U1 refers to the utility funtion of the alternative with ommon setion only aess, whereas both U2 and U3 refer to the utility funtion of the alternatives with dediated setion aess. These utility funtions are speified as follows:

34 34 U n1t = β 0om.se.only + β 4 Geographial Aess, area small (zone) n1t + β 5 Geographial Aess, area medium (region, anton) n1t + β 6 Rush Hour Aess n1t + β 7 Prie n1t + μ 1 Z panel_data,n + μ 2 Z om.se.only,n + ε n1t (14) U n2t = β 0om.se.ded.se. + β 1 Train Setion Aess, Dediated Setion (Business) n2t + + β 3 Train Setion Aess, Dediated Setion (Life style) n2t + β 4 Geographial Aess, area small (zone) n2t + β 5 Geographial Aess, area medium (region, anton) n1t + β 6 Rush Hour Aess n1t + β 7 Prie n1t + μ 1 Z panel_data,n + μ 3 Z om.se.ded.se,n + ε n2t (15) U n3t = β 0om.se.ded.se. + β 1 Train Setion Aess, Dediated Setion (Business) n3t + + β 3 Train Setion Aess, Dediated Setion (Life style) n3t + β 4 Geographial Aess, area small (zone) n3t + β 5 Geographial Aess, area medium (region, anton) n3t + β 6 Rush Hour Aess n3t + β 7 Prie n3t + μ 1 Z panel_data,n + μ 3 Z om.se.ded.se,n + ε n3t (16) In eah equation, n identifies individual respondents, t the hoie task, and εnt is a random disturbane over people and hoie tasks (see also previous explanation in Chapter 8.2). As in Study 1, Model 2 (Table 22), is an extension of Model 1 that uses the latent variable in the utility funtions to understand if it inreases the explanatory power of the model (Walker, 2001). We estimated these models using the maximum simulated likelihood (MSL) approah implemented in the free software, PythonBiogeme 2.4 (Bierlaire, 2016), using Halton draws Proedure We reruited 505 travelers in the German-speaking part of Switzerland for our survey with the help of Intervista AG. Only individuals that were planning to purhase a publi transport subsription in Switzerland or renew their urrent one within the next year and who are paying for their subsription by themselves were eligible to partiipate in the survey. Our sample was stratified by age in order to represent the Swiss population aording to SBB s marketing researh standards (SBB, 2013). The rest of the proedure of Study 2 is idential to the proedure of Study 1. An exemplary hoie task is shown in Figure 12. Figure 12 Exemplary Choie Task (Study 2) Desription of Sample Of the 505 travelers we reruited, 192 were female (38 perent) and the average age was years. Most of the travelers in the sample own a Half-Fare travel ard (64 perent), followed by general subsription (GA, 25 perent), regional travel ards (25 perent), Point-to-Point travel ard (7

35 perent), and the Trak 7 travel ard (2 perent). Four perent of the respondents in our sample did not own a travel ard. The most frequent purposes of travelling by train of travelers in our sample were leisure trips without overnight stays (79 perent), followed by everyday tasks (52 perent). Fifty perent of the travelers indiated that they were ommuting by train to work and 13 perent of the survey respondents were ommuting by train to an eduational institution. Forty-seven perent of the travelers indiated that they were travelling by train for vaation trips and 22 perent for business trips Results from Choie Modeling We first estimated the basi disrete hoie model to investigate the utilities of the attributes of the alternatives and then we estimated a hybrid hoie model with the interation of the out-group derogation. Eah oeffiient represents the ontribution that its speifi attribute level has on the overall utility of the alternative. In line with this, the y-axis of all graphs measures the utility ontribution of the attribute level represented. Model 1 (Appendix Table 21) estimates the oeffiients of all attribute levels for the entire sample. Broader geographial aess provides higher utility to travelers (βarea small = -1.06, p <.01; βarea medium = -.15, p <.01). For a better interpretability, the utilities assoiated to the geographial aess are represented in Figure 13. The utilities of the attribute levels orrespond to the values of the relative oeffiients (e.g., Uarea small = βarea small) exept for the attribute referene level 14. We alulated the utility of the attribute level area big (ountry), referene level of the attribute geographial aess, as follows: U area big = ( 1) β area small + ( 1) β area medium Plugging in the oeffiient estimates, we obtain the following value: U area big = ( 1) ( 1.06) + ( 1) (.15) = 1.21 Confirming the results of Study 1, we find that aess to the rush hour provides higher utility than no aess to the rush hour (βrush hour aess = 1.34, p <.01, Figure 14). Higher pries provide a lower utility than lower pries as shown by the prie oeffiient (βprie < -.01, p < ). The ommon setion and the dediated setion (silene) provide higher utility ompared with the remaining train setion options (Uommon setion = βommon setion = -.39, p <.05, Udediated setion (silene) = ). Having aess to the dediated setion (family) provides the lowest utility (Udediated setion (family) = -.71, see Figure 15). 17 The dediated setions (business) and (lifestyle) do not differ in their utility (Udediated setion (business) = -.55, Udediated setion (lifestyle) = -.55). 18 Travelers are willing to pay CHF more for aess to the big area (entire ountry) than only the small area (zone). They are willing to pay CHF for aess to the rush hour. Finally, ompared with the ommon setion, they are willing to pay CHF 4.- less for additional aess to the dediated setion (silene), CHF less for additional aess to the dediated setion (business) as well as the dediated setion (life-style), and CHF less for additional aess to the dediated setion (family) We oded this variable using the so-alled effet oding tehnique (Hensher et al., 2005). With this proedure, we obtained the value of the referene level by summing up the oeffiients (multiplied by a fator of -1) assoiated with other levels. 15 Prie is ontinuous in the model. Therefore, we do not report it graphially. 16 This oeffiient is alulated by adding the oeffiient of the dediated setion (silene) to the alternative speifi onstant of the dediated setion aess. U dediated setion (silene) = β dediated setion + β dediated setion (silene) 17 U dediated setion (family) = β dediated setion + (-1) β dediated setion (silene) + (-1) β dediated setion (business) + (-1) β dediated setion (lifestyle) 18 As the oeffiients for the dediated setion (business) and the dediated setion (lifestyle) are not signifiant, the utilities assoiated do not differ from the value of the alternative speifi onstant of the dediated setions.

36 Utiltiy Utility Utility 36 Figure 13 The Wider the Geographial Aess, The Higher the Utility Area small (zone) Area medium (region, anton) Area big (ountry) Note: Area big (ountry) represents the referene level for this attribute (alulated as explained above). Figure 14 Aess to the Rush Hour Inreases Utility No rush hour aess 1.34 Rush hour aess Note: No rush hour aess represents the referene level for this attribute (here, fixed to zero). Figure 15 Among the Dediated Setions, the Silene Setion Provides the Highest Utility Common setion only Dediated setion (silene) Dediated setion (business) Dediated setion (lifestyle) Dediated setion (family) Model 2 (Appendix Table 22) is a hybrid hoie model sine it inludes the latent variable outgroup derogation 19 in the estimation proess to inrease the explanatory power of the model. Outgroup derogation moderates the utility for dediated setions (βout-group derogation x dediated setion =.35, p < 19 The average out-group derogation over the sample is 2.23, the standard deviation is 1.25, and the median lies at 2.00.

37 Utility of dediated setion (relative to ommon setion).10) as shown in Table 22. However, this moderation does not differ between the different dediated setions (see insignifiant interations with the dediated setions in Appendix Table 22). For lear interpretability, the oeffiients of the train setions aess interated with the out-group derogation are represented in Figure 16. Travelers with a high tendeny towards out-group derogation (plus one standard deviation) are willing to pay CHF more for additional aess to the dediated setion (business), CHF more for additional aess to the dediated setion (silene), CHF more for additional aess to the dediated setion (life-style), and CHF more for additional aess to the dediated setion (family) ompared with ommon setion aess only. Figure 16 The Utility of the Dediated Setions Inreases Relative to the Common Setion with Inreasing Out- Group Derogation Dediated setion (silene) Dediated setion (business) Dediated setion (lifestyle) Dediated setion (family) Low out-group derogation (-1 SD) Average out-group derogation High out-group derogation (+1 SD) Note: The utility for eah of the dediated setions inreases to the same extent with inreasing out-group derogation Subgroup Analysis The previous analyses showed that travelers high (vs. low) in out-group derogation have a stronger preferene for the dediated setions. In addition, we wanted to understand what other traveler harateristis may drive the hoie for eah of the dediated setions. Understanding whih types of travelers prefer a ertain dediated setion ould inform targeted produt offerings in the future. In order to do so, we analyzed their hoies and linked them to their demographis, psyhographis (latent variables), and the variables from the SBB segmentation studies. We observed how travelers expressed their preferenes between the different available alternatives and reated one dummy variable per possible alternative (i.e., ommon, dediated setion (silene), dediated setion (business), dediated setion (lifestyle), and dediated setion (family)). Then, we ran several logisti regressions using eah dummy as the dependent variable. Below, we report a summary of the main statistially signifiant insights. The probability to hoose the ommon setion over any of the dediated setions is higher for women (vs. men, βmale = -.23, p <.10). It is also higher for pensioners (vs. travelers with any another oupation, βpensioner = -.90, p <.05). The probability to hoose the ommon setion is also higher for people that urrently own (vs. not) a regional travel ard (βregio =.30, p <.01). Additionally, it is higher for respondents that urrently own a seond lass travel ard (vs. a first lass travel ard, βommon =.55, p <.01). When looking at ommuters only, the probability for ommuters to hoose the ommon

38 setion inreases with an inreasing perentage of trips made by train (relative to other means of transportation, βtrain.perent =.11, p <.01). The probability to hoose the dediated setion (business) over any other setion is higher for men (vs. women, βmale =.34, p <.10) and inreases with age (βage =.01, p <.10). It dereases with an inreasing the number of people in the household (βnum.household = -.13, p <.05). The probability to hoose the dediated setion (business) is also lower for respondents that urrently own a seond lass travel ard (vs. a first lass travel ard, βseond_lass = -.39, p <.05). Looking into the latent variables, we found that the probability to hoose the dediated setion (business) inreases with inreasing hedonism of travelers (βhedonism =.12, p <.05), desire for omfort (βomfort =.13, p <.05), and time onsiousness (βtime_onsiousness = -.12, p <.05). For ommuters only, the probability to hoose the dediated setion (business) inreases with an inreasing ommuting trip distane (βtrip.distane =.37, p <.01). The probability to hoose the dediated setion (silene) over any other setion only differs by ertain fators when looking at ommuters. For ommuters, the probability to hoose the dediated setion (silene) inreases with inreasing trip duration (βtrip.duration =.16, p <.01) and with an inreasing number of trips made during the rush hour (βomm.rush.hours =.07, p <.05). The probability to hoose the dediated setion (lifestyle) over any other setion is lower for travelers who own a regional travel ard (βregional = -.42, p <.01) and if their travel ard provides aess to the seond lass (vs. a first lass, βseond_lass = -.46, p-value <.05). Moreover, the probability to hoose the dediated setion (lifestyle) is lower for ommuters than for non-ommuters (βommuters = -.23, p <.05). The probability to hoose the dediated setion (family) over any other setion is higher for pensioners (vs. travelers with any other oupation, βpens = 1.71, p-value <.01). Analyzing the effet of the latent variables, we found that the probability of hoosing the dediated setion (family) dereases with an inrease in status-seeking lifestyle (βstatus = -.12, p-value <.05), the more a traveler likes traveling (βtravel.liking = -.23, p <.01), and the higher a traveler s desire for omfort (βomfort = -.20, p <.05). It also inreases the more a traveler pereives traveling as hedoni (vs. utilitarian, βhedoni =.11, p <.05) Summary and Disussion In ontrast to Study 1, we do not find a stronger preferene for travel ards that offer additional aess to the dediated setions overall, rejeting H1a. Nevertheless, the preferene for having aess to one of the dediated setions shows a marginally signifiant tendeny to inrease for travelers with higher out-group derogation, onfirming H1b. Travelers with a high tendeny towards out-group derogation (16 perent of the population) were willing to pay on average CHF more for aess to one of the dediated setions. Speifially, they were willing to pay CHF more for additional aess to the business setion, CHF more for additional aess to the silene setion, CHF more for additional aess to the life-style setion, and CHF more for additional aess to the family setion. This is in line with study 1, where travelers with a high tendeny toward out-group derogation were willing to pay CHF more for having additional aess to all dediated setions. This result ould have ourred due to the speifi way we designed the dediated setions; they an be seen as almost mutually exlusive, whih means travelers who like one of them is likely to dislike the others. Our subgroup analysis supports this argument. The dediated setion (business) is preferred by older travelers that urrently have a first lass travel ard, are more time onsious, have a stronger desire for omfort, and are more hedonisti. Conversely, the dediated setion (family) is preferred by pensioners, travelers with a less status-seeking lifestyle, like traveling less, and have a lower desire for omfort Summing up the WTP for all dediated setions in Study 2 leads to a value of CHF whih is in line with the WTP of CHF for all dediated setions that we found in Study 1.

39 9.5 Reommendations for SBB In summary, our results onfirm that the dediated setions on the train provide value to different segments of travelers. We found that none of the setions provides value to everyone, but rather eah of them appeals to speifi types of travelers. However, we found that the more a traveler wants to separate from dissimilar traveler groups (higher tendeny towards out-group derogation), the higher the preferene for one of the dediated setions. These results are theoretially supported by soial identity theory (Tajfel & Turner, 1979). The different setions on the train represent a hoie of soio-psyhologial groups into whih a traveler an self-selet based on what he defines as his ingroup or his out-groups, respetively. This way, travelers with a higher tendeny toward outgroup derogation may be able to redue their pereived travel stress and use their travel time more produtively. These findings imply that separately priing (on average CHF more for aess to one dediated setion than for ommon setion only in order to target 16 perent of the population) these dediated setions may allow the generation of additional revenues for SBB. However, a highly targeted approah is required. Additionally, priing the setions may be operationally feasible as some of them are already implemented (silent, business, and family setions). More researh is ertainly needed to arrive at a final deision about whether and how to provide targeted offerings related to dediated setion aess. For instane, future researh should better define the target segments and link them to SBB s urrent ustomer segmentation, the atual prie points should be speified in detail, and market simulations should be utilized to gauge the potential market share of suh travel ards. Finally, future researh should additionally measure pereptions of fairness of suh offerings. 10 Rush Hour Aess Projet Trains are typially highly ongested during rush hour, harming the satisfation of SBB ustomers (Riedener, 2012). Thus, an important question is how more travelers ould be motivated to be more flexible in their traveling times. This an potentially be ahieved through produt and priing strategies that limit aess to the rush hour. Instead of always allowing full aess to all travelers, less expensive travel ards that restrit aess ould be offered. The lower prie of suh ards may ause travelers to aept limited traveling times. This strategy bears two main questions that we investigated in this researh projet. First, are travelers really willing to aept travel ards with limited aess to rush hour and how muh less would they be willing to pay? Seond, how should rush hour limits be defined to be effetive? These questions are most relevant to travelers who ommute and who are mostly affeted by the negative onsequenes of ongestion (Bloh, 2011; Ungriht, 2010; Zemp, 2014). Additionally, we explored different types of external influenes on ommuting time hoies. Some ommuters are faed with offiial formal onstraints at their ommuting destination, suh as working times that are defined in their work ontrats. Moreover, there may be soial norms that influene ommuting time hoies. Soial norms are the guidelines for the behavior that is aepted in a group without the existene of any offiial, formal, or written rules. They develop due to mutual interation, are understood by the group members, and deviations thereof are soially santioned within the group (Cialdini & Trost, 1998). We examined whether ommuters faed soial norms at their ommuting destination regarding their flexibility or the duration of their working time and whether they influened travel ard hoies. We also examined the role of formal onstraints. Additionally, we examined whether ommuters faed formal onstraints and soial norms in their private environment. In what follows, we will first shortly review the literature related to how ommuters make timerelated traveling deisions. Then, we will introdue the methodology we used and outline how we applied it. Finally, we will report the results followed by a disussion of the impliations for SBB. 39

40 10.1 Related Literature How onstraints influene departing time deisions has already investigated, espeially for ommuting by ar. Chang and Mahmassani (1988) looked at how ommuters make their departing time deisions and travel time estimations with regard to their latest possible arrival time by estimating a dynami model of departure time adjustment. Another study, while building on prospet theory, found that being late had a different impat on ommuters utility than being early (Senbil & Kitamura, 2004, 2005). Senbil and Kitamura (2004) defined three time referene points in the ommuting time deision, whih are the earliest permissible arrival time (sine one does not wish to arrive at work too long before starting), the ideal arrival time, and work start time whih is usually the latest possible arrival time. Similarly, in a study defining two referene points for ommuter deisions, Jou, Kitamura, Weng, and Chen (2008) showed that ommuters reated asymmetrially to gains and losses with losses being defined as arrivals outside the range of aeptable arrival times. As these onstraints lead many ommuters to travel during the same time interval, literature has also broadly investigated the effetiveness of instruments inentivizing rush hour avoidane. Both rewards and punishments were investigated as poliy instruments to distribute travelers more equally over the time of a day. Based on a disrete hoie experiment, Knokaert, Tseng, Verhoef, and Rouwendal (2012) disovered that rewards for avoiding travel during the rush hour are effetive means to influene departure time hoie. When omparing rush-avoidane rewards to rush-use priing, Tillema, Ben-Elia, Ettema, and van Delden (2013) autiously onluded that rewards were more effetive in moving ommuters out of the rush hour. Ozbay and Yanmaz-Tuzel (2008) adopted a different perspetive on the problem. They were not onerned with studying the effetiveness of different poliy instruments, but instead they hose to estimate individuals value of travel time by estimating a model based on individuals travel time deisions as a reation to time-of-day toll priing. They onluded that people s value of travel time (i.e., their willingness to pay for reduing travel time) on average ranged between $15 and $20 per hour depending on departure time hoie and trip purpose with higher values for individuals doing work related trips. To our knowledge, very few studies have reviewed instruments to inentivize peak hour avoidane for publi transport. One study evidened that giving rewards for rush-hour avoidane has a positive effet on moving trips out of the rush hour in the Beijing subway during morning ommutes (Zhang, Fujii, & Managi, 2014). An applied study on how rush hour trains ould be relieved was the Work Anywhere study undertaken by the SBB in ooperation with Swissom (Weihbrodt et al., 2013). They allowed 264 employees to arrange their working day between their home and their workplae to allow them to travel during off-peak hours. This flexibility moved 66 perent of travel time out of the peak hours Our Researh and Pratial Contribution Different from prior literature, we did not investigate the value of saving one hour of travel time. Instead, we examined the value of being allowed (vs. not) to travel during a ertain time interval. We aomplished this by studying the utility for travel ards that do not provide aess to the rush hour or only some form of limited aess to the rush hour in omparison with travel ards that provide unlimited aess to the rush hour. This an then be used as a basis to design an effetive instrument to inentivize rush hour avoidane. We further ontribute to the body of literature on ommuting in publi transportation. Commuters that are ommuting by train (vs. by ar) are faed with less unertainty with respet to their travel time, sine trains have fixed shedules. However, they have the same onstraints regarding work start time. Prior literature studied different time points of these onstraints. We are the first to explore the influene of different types of onstraints (formal onstraints and soial norms in the professional and private environment) on time-related traveling deisions and on the utility for some form of aess to the rush hour. 40

41 10.3 Study We first introdue the methodology by explaining the hoie experiment, the variables we measured, and the model. Seond, we will explain the proedure of the study. Third, we desribe our sample. Finally, we show the resulting hoie models and finally we disuss our results and offer reommendations for SBB Methodology Choie Experiment As in the Train Setion Aess Projet, a bloked hoie design with five bloks and 12 hoie tasks per blok was developed. Respondents were randomly assigned to one of five bloks. In eah blok, the order of the hoie tasks was randomized. The hoie experiment utilized a labelled design to identify the alternatives. Eah alternative is a speifi ombination of the attributes listed below. Eah alternative had four attributes: geographial aess, omfort level, rush hour (aess), and prie that were agreed upon with SBB 21 (see Table 6 for details). The geographial aess attribute desribes the geographial extension in whih the subsription is valid. It had four levels: area small (zone) represents a subsription that provides omplete aess in a small size home (or destination) area; area medium (region, anton) represents a subsription that provides omplete aess in a medium size home (or destination) area; area medium (region, anton) + route > 10 km represents a subsription with area-route ombination and provides omplete aess in a medium size home (or destination) area and an aess to a speifi route; and area big (ountry) represents a subsription with aess to publi transport throughout Switzerland. The omfort level attribute desribes setions on the train with different qualities of servies. It had two levels: first lass represents a subsription with aess to setions on the train with highquality servies; seond lass represents a subsription with aess to setions on the train with basi servies. The rush hour (aess) attribute desribes the types of aess during rush hour. It had three levels: unlimited aess represents a subsription with aess to publi transport without restritions; Limited aess represents a subsription with no aess to the defined rush hour interval exept for a speifi number of trips; No Aess represents a subsription with no aess to the defined rush hour interval. The prie attribute desribes the eonomi outlay to purhase a speifi travel ard. It had four levels ( CHF , CHF , CHF , CHF ). Additionally, alternatives with levels limited aess and no aess for the rush hour (aess) attribute have an additional attribute: rush hour (time frame). It represents the time intervals in whih the subsription is valid for a limited number of trips and not valid anymore after having ompleted these trips ( limited aess ) or is not valid ( no aess ). It had two levels: 7:00 8:00 and 17:00 18:00 (narrow) ; and 6:00 9:00 and 16:00 19:00 (wide). Outside these intervals, there are no restritions. Finally, alternatives with the limited aess level for the rush hour (aess) attribute have an additional attribute: rush hour (number of trips). This represents the number of trips allowed with a subsription with limited aess during rush hour and it had two levels: 20 trips per year, whih provides aess to the defined rush hour interval for 20 trips per year; and 30 trips per year. After having used the defined number of trips during the rush hour, travelers with a limited aess travel ard are not permitted to use publi transport during the rush hour. The desription of all rush hour aess attributes that was provided to travelers during the survey is shown in Appendix Figure The deision about these attributes and attribute levels was made during a meeting with Prof. Dr. Reto Hofstetter, Dr. David Blatter (SBB Personenverkehr, Preis und Sortiment), and Dr. Silvio Stiher (SBB Personenverkehr, Preis und Sortiment) on the in Bern.

42 42 Table 6 Attributes and Attribute Levels for Rush Hour Aess Projet Attribute Number Desription of levels Geographial aess 4 Area small (zone); Area medium (region, anton); area medium (region, anton) + route > 10 km; area big (ountry) Comfort level 2 First lass; seond lass Rush hour (aess) 3 Unlimited aess; limited aess; no aess Rush hour (time frame) 2 7:00 8:00 and 17:00 18:00 (narrow); 6:00 9:00 and 16:00 19:00 (wide) Rush hour (number of trips) 2 20 trips per year; 30 trips per year Prie 4 CHF ; CHF ; CHF ; CHF Latent and Other Variables In this survey, we also added variables related to the SBB segmentation studies (SBB, 2013) beause they would allow linking to segments at a later time. We extended those with further measures regarding ommuting-time deisions. We also inluded demographis and other typially used latent variables as possible ontrols into the survey. In ontrast to prior literature, we added different types of measures for fators that ould onstrain travel time hoies. These inlude soft fators, suh as soial norms and pereived onstraints due to soial norms and hard fators, suh as formal time onstraints. The variables of the SBB segmentation studies inluded the trip purpose (ommuting to and from plae of work, ommuting to and from an eduational institution, leisure trip without overnight stay (exursion, inema, sports, visit), private vaation trip with at least one overnight stay, everyday tasks (onsultation of a dotor, groery shopping, piking somebody up), business trip, other), the predominant travel lass (first lass, seond lass), the need for ommuting as well as the ommuting habits, and the habits when engaging in leisure trips. We additionally measured how muh they ould deviate from their ommuting time. Sine we measured a list of variables in this study, we did not want to overload respondents and thus redued the number of sales ompared with the Train Setion Aess Projet and, in some sales, the number of items that we measured was redued ompared with the original sales. The Cronbah s alphas of the multiple-item measures are reported in Table 7. Finally, only for travelers that self-identified as ommuters, we measured a set of latent variables regarding the type of onstraint ommuters faed, whih we expeted to add explanatory power for the hoies between the different levels of the rush hour (aess), rush hour (time frame), and rush hour (number of trips) attributes. First, with a 14-item sale that we adapted from Bizer, Magin, and Levine (2014), we measured whether respondents have an individual tendeny to omply with and value soial norms at their ommuting destination (e.g., workplae, university, shool, et.). We adapted the same sale to measure whether respondents have an individual tendeny to omply with and value soial norms in their private environment (e.g., family, sports lub, et.). We will refer to these variables as soial norms ompliane (professional) and soial norms ompliane (private) in the rest of the doument. Seond, we measured with a seven-point, three-item sale to what extend respondents feel onstrained in their ommuting behavior by the soial norms at their ommuting destination. We will refer to this variable as pereived onstraints due to professional soial norms. 22 Similarly, we measured to what extend respondents feel onstrained in their ommuting behavior by the soial norms in their private environment with a seven-point, three-item sale. We will refer to this variable as pereived onstraints due to private soial norms. Third, for the soial norms at the ommuting destination, we measured whether respondents fae soial norms regarding the duration of the time of presene at the ommuting destination with a 22 The three items for a ommuter ommuting to work were worded as follows: What other people at my workplae think about when one should arrive and leave onstrains my ommuting behavior strongly., I feel strongly onstrained in my ommuting behavior by the expetations of other people at my workplae., My ommuting behavior is onstrained by what other people at my hello expet from me.. For ommuters ommuting to a plae of eduation, workplae was replaed with plae of eduation. If ommuters indiated another ommuting destination, that destination was shown in the sale item instead. The items were worded aordingly for the private environment.

43 seven-point, three-item sale 23. That is, we measured whether it is a soial norm at the ommuting destination to arrive earlier and leave later than offiially required. We will further refer to this variable as duration norms. Fourth, we measured whether they fae soial norms regarding the flexibility of the time of presene at the ommuting destination with a seven-point, three-item sale 24. That is, we measured whether it is a soial norm at the ommuting destination to always be present during the same time interval and not to vary times of presene. We will further refer to this variable as flexibility norms. Fifth, we measured the desriptive soial norms regarding duration and flexibility by asking whether people at the respondent s ommuting destination atually spend more time at the ommuting destination than offiially required (further referred to as desriptive duration norm ) and whether people at the respondent s ommuting destination are atually always present during the same time interval (further referred to as desriptive flexibility norm ). The Cronbah s alphas of these measures are also reported in Table 7. Finally, as mentioned previously, we used two binary variables to measure whether respondents fae offiial formal time onstraints at their ommuting destination (e.g., workplae, university, shool, et.) or in their private environment (e.g., fixed training shedules sports lubs, piking up kids at shool, et.) that influene their ommuting behavior ( professional formal onstraints and private formal onstraints ). The orrelations between all latent variables are reported in Appendix Tables 17 and 18. Table 7 Cronbah s Alpha of Latent Variables 43 Variable Cronbah s alpha From Previous Literature Family/ommunity oriented lifestyle.65 Workaholi lifestyle.65 Frustrated lifestyle.63 Commute benefits.68 Time-onsiousness.76 Desire for omfort.51 Desire for flexibility.75 Traveling hedoni (vs. utilitarian).85 Travel stress.78 Soial norms ompliane (professional), adjusted 1.00 Soial norms ompliane (private), adjusted 1.00 New Pereived onstraints due to professional soial norms.85 Pereived onstraints due to private soial norms.92 Duration norms.85 Flexibility norms Model Speifiation and Estimation Applying the disrete hoie model framework to the data we olleted in this projet, we estimated two similar models. In Model 1 (Appendix Table 23), we analyzed travelers preferenes without taking into aount any distintion between them. In Model 2 (Appendix Table 24), we 23 We first explained the onept of soial norms to travelers and then had them indiate their degree of agreement to the following three statements: It is a soial norm at my workplae to leave later than offiially required., It is a soial norm at my workplae to arrive earlier than offiially required., It is a soial norm at my workplae to be present beyond the offiially required working time.. 24 We first explained the onept of soial norms to travelers and then had them indiate their degree of agreement to the following three statements It is a soial norm at my workplae that everybody is always present at the same time., It is a soial norm at my workplae that everybody always starts at the same time., It is a soial norm at my workplae that everybody always leaves at the same time..

44 distinguish between ommuters and non-ommuters to extrapolate the differenes in utility among the groups. Figure 17 shows the disrete hoie model framework used. Figure 17 Full Path Diagram of the Choie Model 44 We adopted a mixed logit with an error omponent (EC model) for both models as in the Train Setion Aess Projet (Chapters and ). In Model 1 (Appendix Table 23) and Model 2 (Appendix Table 24), we have three utility funtions (U1, U2, U3) aording to the number and the type of alternatives defined in the hoie design (see Chapter ). Given that we adopted a labeled hoie design, eah utility funtion is assoiated with a speifi produt based on the levels of the rush hour (aess) attribute. U1 refers to the utility funtion of alternatives with no aess during the rush hour, U2 to alternatives with limited aess during the rush hour, and U3 to alternatives with no time restritions (unlimited aess). These utility funtions are speified as follows: U n1t = β 0no.aess + β 1 Geographial Aess, area small (zone) n1t + + β 3 Geographial Aess, area medium (region, anton) and route(> 10 km) n1t + β 4 Comfort Level n1t + β 5 Rush Hour (time frame) n1t + β 7 Prie n1t + μ Z panel_data,n + ε n1t (18) U n2t = β 0limited.aess + β 1 Geographial Aess, area small (zone) n2t + + β 3 Geographial Aess, area medium (region, anton) and route(> 10 km) n2t + β 4 Comfort Level n2t + β 5 Rush Hour (time frame) n2t + β 6 Rush Hour (number of trips) n2t + β 7 Prie n2t + μ Z panel_data,n + ε n2t (19) U n3t = β 0unlimited.aess + β 1 Geographial Aess, area small (zone) n3t + + β 3 Geographial Aess, area medium (region, anton) and route(> 10 km) n3t + β 4 Comfort Level n3t + β 7 Prie n2t + μ Z panel_data,n + ε n3t In eah equation, n identifies the individual respondents, t the hoie task and, εnt is a random disturbane over people and hoie tasks (see also previous explanation in paragraph 8.2). Introduing the ommuting habits in the Model 2 (Appendix Table 24), eah utility funtion has two betas per eah attribute level, one for ommuters and one for non-ommuters. In pratie, we interated eah omponent of the utility funtions (exept for the error omponents and the error term) mentioned above with two dihotomous variables (one per group). (20)

45 45 We estimated these models using the maximum simulated likelihood (MSL) approah implemented in the free software, PythonBiogeme 2.4 (Bierlaire, 2016), using Halton draws Proedure With the help of the market researh ageny Intervista AG, we reruited 504 travelers in the German-speaking part of Switzerland for our survey. Only individuals who were planning to purhase a publi transport subsription in Switzerland or renew their urrent one within the next year and who are paying for their subsription by themselves were eligible to partiipate in the survey. Our sample was stratified by age in order to represent the Swiss population aording to SBB s marketing researh standards (Swiss Federal Statistial Offie 2014). The rest of the proedure of this study was idential to the proedure of the two studies in the Train Setion Aess Projet. An exemplary hoie task is shown in Figure 18. Figure 18 Exemplary Choie Task Desription of Sample Of the 504 travelers we reruited, 267 were female (53 perent) and the average age was years. Most of the respondents in the sample own a Half-Fare travel ard (66 perent), followed by general subsriptions (GA, 24 perent) and regional travel ards (20 perent). Six perent of respondents owned a Point-to-Point travel ard, three perent owned a Trak 7 travel ard, and three perent had no urrent subsription. The most frequent purposes of travelling by train of respondents in our sample were leisure trips without overnight stays (83 perent), followed by vaation trips with at least one overnight stay (49 perent). Forty-six perent of the survey respondents indiated that they were ommuting by train to work and 17 perent of the survey respondents were ommuting by train to an eduational institution. Forty-eight perent of respondents used the train to omplete everyday tasks and 18 perent for business trips. Most of the ommuters (train or ar) were ommuting to work (78 perent), followed by ommuting to an eduational institution (19 perent). The most frequent trip distane ommuters were overing lies between 15 and 50 kilometers (49 perent). Most ommuters indiated that their ommuting trip last approximately 45 minutes or 1 hour (42 perent). Ninety-one perent of ommuters were doing at least 10 perent of their ommuting trip during the wide rush hour interval (6:00 to 9:00 and 16:00 to 19:00). Fifty-six perent of the ommuters that usually travel during the rush hour indiated that they ould deviate their ommuting time by more than 30 minutes.

46 Frequeny Seventy-four perent of ommuters were faed with professional formal time onstraints at their ommuting destination (e.g., ontrats at the workplae, fixed shedules at shool or in university) that influene their ommuting behavior. Only 33 perent of ommuters were faed with private formal time onstraints (e.g., fixed training shedules sports lubs, piking up kids at shool, et.) that influene their ommuting behavior. Those ommuters that were faed with formal time onstraints (professional) were also more likely to additionally pereive onstraints in their ommuting behavior due to soial norms (β =.68, p <.01). Moreover, younger respondents pereived stronger onstraints in their ommuting behavior due to soial norms at the ommuting destination (β = -.01, p <.01). We further found less variane around the peak time of the rush hour both in the morning (σmorning_no_onstraints= 8.54, σmorning_ onstraints= 6.94, F(60,170) = 1.51, p <.05) and in the evening (σevening_no_onstraints= 13.12, σevening_onstraints= 9.31, F(60,170) = 1.98, p <.01) for ommuters with (vs. without) formal onstraints. Additionally, the ommuters with (vs. without) formal onstraints were leaving approximately one hour later from the ommuting destination to return home (t(230) = -2.61, p <.01). Figure 19 looks only at those ommuters with professional formal time onstraints at their ommuting destination. It ompares the latest possible arrival times with their atual arrival times and the earliest possible departure times from their ommuting destination with their atual departure time. The distributions show that many ommuters arrive earlier and leave later than offiially required. The median of the atual arrival time at the ommuting destination is at 7:45 (dashed line), and the median of the latest possible arrival time is at 8:15 (solid line). The median atual departure time is 17:30 (dashed line), and the median of the earliest possible departure time is at 16:15 25 (solid line). It may be possible that ommuters are antiipating train delays during the rush hour and are self-insuring against these delays by leaving earlier. Figure 19 Commuters with Formal Constraints Spend More Time at their Commuting Destination than Offiially Required by Formal Constraints :45 median atual arrival 8:15 median latest possible arrival 16:15 median earliest possible departure 17:30 median atual departure Atual arrival / departure Latest possible arrival / earliest possible departure Finally, ommuters to an eduational eduation (vs. ommuters to work and to other ommuting destinations) pereived less flexibility in varying their arrival and departure times due to 25 We also tested the mean differenes and found signifiant results both for arrival time (t(170) = 4.77, p <.01) and departure time (t(170) = -5.14, p <.01).

47 the soial norms at their ommuting destination (F(2, 229) = 7.29, p <.01). The older a ommuter, the weaker is the pereption that he or she is expeted to be at the ommuting destination always during the same time interval (β < -.01, p <.01) Results from Choie Modeling Model 1 (Appendix Table 23) estimates the oeffiients of all attribute levels for the entire sample. Eah oeffiient represents the ontribution that its speifi attribute level has on the overall utility of the alternative. In line with this, the y-axis of all graphs measures the utility ontribution of the attribute level represented. It shows that more geographial aess provides higher utility to travelers (βarea_small = -.90, p <.01; βarea_medium = -.20, p <.01; βarea_medium_route =.08, p >.10). For a better interpretability, the oeffiients of the geographial aess are represented in Figure 20. The utilities of the attribute levels orrespond to the values of the relative oeffiients (e.g. Uarea small = βarea small) exept for the attribute referene level 26,27. Results indiate that seond lass provides higher utility than first lass (βomfort level =.37, p <.01, Figure 21). Higher pries provide a lower utility than lower pries as shown by the prie oeffiient (βprie < -.01, p < ). Looking at the rush hour aess, we find that unlimited aess to the rush hour provides the highest utility, followed by no aess to the narrow rush hour, and limited aess to the narrow rush hour for 20 trips per year (βunlimited_aess =.10, p =.50; βno_aess = -.79, p <.01; βno_aess = -1.16, p <.01). If the rush hour interval is defined as wide (vs. narrow), the utility for a travel ard with limited aess dereases even more (βtime frame - limited aess = -.25, p <.05). This derease is even stronger for travel ards with no aess to the rush hour (βtime frame-no aess = -.57, p <.01). Additionally, when the number of trips that is allowed per year with a travel ard that provides limited aess to the rush hour inreases from 20 to 30, the utility of that travel ard inreases (βnumber of trips limited =.27, p <.05). In order to have an overview of the alternatives proposed (unlimited aess during rush hour, limited aess during wide rush hour with 20 trips per year, limited aess during wide rush hour with 30 trips per year, et.), we alulated the utility of eah ombination by summing up the utilities of their elements (Figure 22). For example, we alulated the utility of the alternative limited aess during wide rush hour with 30 trips as follows: 47 U lim.wide.30trips = β 0limited.aess + β 5 Rush Hour (time frame) limited.aess +β 6 Rush Hour (number of trips) limited.aess Plugging in the oeffiient estimates, we obtain the following value: U lim.wide.30trips = = Based on this model, we were also able to alulate the willingness to pay differenes between those attribute levels per attribute that provide the highest utility and the lowest utility. Travelers would be willing to pay CHF more for a travel ard with aess to the entire ountry (vs. only aess to a maximum of two zones). They would also be willing to pay CHF more for a travel ard with unlimited aess to the rush hour (vs. with limited aess to the narrow time frame, 20 trips per year) and they would also be willing to pay CHF more for the unlimited aess ompared with no aess to the narrow time frame. When having limited aess to the rush hour, they would be willing to pay CHF less for a travel ard with a width rush hour definition from 06:00 to 09:00 and from 16:00 to 19:00 (vs. narrow from 07:00 to 08:00 and from 17:00 to 18:00). When having no aess to the rush hour, they would be willing to pay CHF less for a travel ard with a wide rush hour definition from 06:00 to 09:00 and from 16:00 to 19:00 (vs. narrow from 07:00 to 08:00 and from 17:00 to 18:00). Inreasing the number of trips permitted during the rush hour from 20 to We oded this variable using the so-alled effet oding tehnique (Hensher et al., 2005). With this proedure, we obtained the value of the referene level by summing up the oeffiients (multiplied by a fator of -1) assoiated with other levels. 27 U area big = (-1) β area small + (-1) β area medium. The oeffiient assoiated with area medium (region, anton) + route (> 10 km) is not signifiantly different from zero and, for this reason, we do not use it to alulate the utility of area big (ountry). 28 Prie is ontinuous in the model. Therefore, it is not reported graphially.

48 Utility Utility 48 inreases the willingness to pay by CHF Travelers would further be willing to pay CHF more for a seond lass (vs. first lass) travel ard. Figure 20 The Wider the Geographial Aess, the Higher the Utility Area small (zone) Area medium (region, anton) Area medium (region, anton) + route (> 10 km) Note: The utility assoiated with area medium (region, anton) + route (> 10 km) is not represented beause the relative oeffiient is not signifiantly different from zero. Figure 21 Travelers Prefer Seond Class to First Class Area big (ountry) First lass Seond lass Note: First lass represents the referene level for this attribute (here, fixed to zero).

49 Utility Unlimited aess Figure 22 Unlimited Aess Provides the Highest Utility -.79 No aess (narrow) -.89 Limited aess (narrow; 30 trips) Limited aess (wide; 30 trips) Limited aess (narrow; 20 trips) Note: The utility assoiated with unlimited aess is not represented beause the relative oeffiient is not signifiantly different from zero. We alulated the other values ombing the oeffiients assoiated with the attribute levels (as explained in the previous page). Model 2 (Appendix Table 24) estimates the oeffiients for all attributes and attribute levels split between ommuters and non-ommuters. We found that ommuters (vs. non-ommuters) derive a higher utility from a travel ard with unlimited aess to the rush hour (βunlimited_ommuters =.61, p <.01; βunlimited_non-ommuters = -.31, p >.10). Commuters (vs. non-ommuters) also derive a higher utility from a travel ard with limited aess for 20 trips when the rush hour is defined as lasting from 7:00 to 8:00 and 17:00 to 18:00 (narrow) (βlimited_ommuters = -.70, p <.05; βlimited_non-ommuters = -1.50, p <.01) and without aess to the rush hour with the rush hour defined as lasting from 7:00 to 8:00 and from 17:00 to 18:00 (narrow) (βnoaess_ommuters = -.29, p >.10; βnoaess_non-ommuters = -1.17, p <.01). While it does not make a differene for non-ommuters how the rush hour is defined when having limited aess (βnoaess_timeframe_non-ommuters = -.11, p >.10), for ommuters a wider definition of the rush hour from 6:00 to 9:00 and 16:00 to 19:00 (vs. narrow 7:00 to 8:00 and 17:00 to 18:00) leads to a lower utility (βlimited_timeframe_ommuters = -.45, p <.01). When the travel ard provides no aess to the rush hour, the definition of the time frame plays a stronger role for ommuters (βnoaess_timeframe_ommuters = -.74, p <.01; βnoaess_timeframe_non-ommuters = -.44, p <.05). For non-ommuters (vs. ommuters) being allowed 30 (vs. 20) trips with a travel ard that allows limited aess to the rush hour provides a slightly higher utility (βlimited_trips_ommuters =.27, p <.10; βlimited_trips_non-ommuters =.29, p <.05). In order to have an overview of the alternatives proposed divided by ommuters or non-ommuters (unlimited aess during rush hour, limited aess during wide rush hour with 20 trips per year, limited aess during wide rush hour with 30 trips per year, et.), we alulated the utility of eah ombination by summing up the utilities of their elements (Figure 23). For example, we alulated the utility of the alternative limited aess during wide rush hour with 30 trips for ommuters as follows: No aess (wide) Limited aess (wide; 20 trips) U lim.wide.30trips_ommuters = β 0limited.aessommuters + β 5 Rush Hour (time frame) limited.aessommuters +β 6 Rush Hour (number of trips) limited.aess_ommuters Plugging in the oeffiient estimates, we obtain the following value: U lim.wide.30trips_ommuters = =.88 Based on this model, we were also able to alulate the willingness-to-pay differenes between those levels per attribute that provide the highest and lowest utility. Commuters would be willing to pay CHF more for a travel ard with unlimited aess to the rush hour (vs. with limited aess to the narrow time frame, 20 trips per year), whereas non-ommuters would be willing to pay CHF more. Moreover, ommuters would be willing to pay CHF more for unlimited aess to the rush hour ompared with no aess during the narrow time frame, whereas non-ommuters

50 Unlimited aess No aess (narrow) Limited aess (narrow; 30 trips) Limited aess (narrow; 20 trips) No aess (wide) Limited aess (wide; 30 trips) Limited aess (wide; 20 trips) Unlimited aess No aess (narrow) Limited aess (narrow; 30 trips) Limited aess (narrow; 20 trips) No aess (wide) Limited aess (wide; 30 trips) Limited aess (wide; 20 trips) Utility of wide aess (relative to narrow aess) would be willing to pay CHF more. When having limited aess to the rush hour, ommuters would be willing to pay CHF less for a travel ard with a rush hour definition from 06:00 to 09:00 and from 16:00 to 19:00 (vs. from 07:00 to 08:00 and from 17:00 to 18:00), whereas there was not a differene in willingness-to-pay for non-ommuters. When having no aess to the rush hour, ommuters would be willing to pay CHF less for a travel ard with a rush hour definition from 06:00 to 09:00 and from 16:00 to 19:00 (vs. from 07:00 to 08:00 and from 17:00 to 18:00), whereas non-ommuters would be willing to pay CHF less. Inreasing the number of trips that are allowed during the rush hour from 20 to 30 inreases the willingness to pay by CHF for ommuters and by CHF for non-ommuters. Figure 23 Commuters Derive a Higher Utility from All Aess Options to the Rush Hour Commuters Non-ommuters Note: The utility assoiated with no aess (narrow) for ommuters and unlimited aess for non-ommuters are not represented beause the relative oeffiients are not signifiantly different from zero Subgroup Analysis As in the previous projet, we observed how travelers expressed their preferenes between the different available alternatives and reated one dummy variable per possible alternative for this purpose (i.e., no aess, limited aess, and unlimited aess). Then, we ran several logisti regressions using eah dummy as the dependent variable. Below, we report a summary of the main statistially signifiant insights. The probability to hoose a travel ard with unlimited aess during rush hour (vs. any other) is higher for ommuters (vs. non-ommuters, βommuters =.48, p <.01) that ommute to and from work (βomm.work =.29, p <.05) or use the train for their business trips (βbus.trips =.32, p <.05). This onfirms the results of the hoie model that distinguishes between ommuters and non-ommuters. When looking only at ommuters, the probability to hoose this alternative inreases when they have private formal onstraints (βpriv.formal onstr. =.46, p <.05). We also found that the probability to hoose this alternative inreases with a dereasing duration norm (i.e., the pereption that it is a soial norm to arrive earlier and leave later than offiially required, βduration norm = -.26, p <.01). It also inreases the

51 less workaholi the lifestyle of a traveler is (βworkaholi = -.16, p <.05) and the lesser a traveler s desire for omfort (βtrav.omfort =.14, p <.05). It also inreases the less a traveler pereives traveling as hedoni (vs. utilitarian, βhedoni = -.13, p <.01). The probability to hoose a travel ard with limited aess to the rush hour (vs. any other) is higher for pensioners (βpensioner =.86, p <.05). This probability is also higher for non-ommuters (vs. ommuters, βommuters = -.44, p <.01), with dereasing hedonism of travelers (βhedonism = -.17, p <.01) and with an inreasing workaholi lifestyle of a traveler (βworkaholi =.15, p <.05). Moreover, it inreases the higher a traveler s pereption of ommute benefits (βomm.benefits =.17, p <.05) and the lesser a traveler s desire for omfort (βomfort = -.16, p <.01). It also inreases the more a traveler pereives traveling as hedoni (vs. utilitarian, βtravel.hedoni =.16, p <.01). The probability to hoose a travel ard with no aess during rush hour (vs. any other) is higher for travelers that urrently own (vs. not) a Half-Fare travel ard (βhalf-fare =.52, p <.05) and for people that do not own a subsription (βno.subsription =.78, p <.05) Summary and Disussion In the first model, we found that travelers had a strong preferene for travel ards that provided unlimited aess to the rush hour ompared with travel ards that provided limited aess or no aess. On average in the sample, travelers would be willing to pay CHF more for a travel ard with unlimited aess to the rush hour than for the least preferred limited travel ard, a ard that provided aess for 20 trips per year to the wide rush hour. Similarly, they would be willing to pay CHF more for a travel ard with unlimited aess than for a travel ard with no aess to the wide rush hour. A possible interpretation of this finding is that for most travelers, aess to the rush hour is indispensable. The result that the utility for limited aess to the rush hour is lower than the utility for no aess to the rush hour is rather surprising. It may be that travelers do not want limits on their travel ards for onveniene reasons. Having to ount the number of trips one has left during the rush hour may be pereived as umbersome. Not hoosing travel ards with limited aess over travel ards with no aess would be an attitude expression regarding that pereption. Nevertheless, travelers are rational regarding the number of trips; they are willing to pay CHF more for a travel ard with limited aess for 30 than for 20 trips. Thus, having less aess to the rush hour makes an eonomially relevant differene in the willingness-to-pay for travel ards. Counter to our expetation, we found that travelers derive a lower utility from a travel ard with aess to the first lass (vs. seond lass). A possible explanation for this result may be overrowded first lass setions during rush hour. If travelers have to stand, even in the first lass, it does not provide the additional utility it is meant to provide ompared with seond lass. In the seond model, we found that ommuters (vs. non-ommuters) have a higher utility for all rush hour aess options. This shows that they are in general more likely to purhase a travel ard than non-ommuters. It is interesting that this is valid even for travel ards with no aess to the rush hour. However, ommuters (vs. non-ommuters) are more sensitive to the definition of the rush hour time frame. Some ommuters may be able to travel outside the narrow rush hour, but if the travel ard does not allow them to travel during the entire wide rush hour time frame, it signifiantly loses value for them (for a travel ard with no aess to the rush hour, the willingness to pay drops by CHF ). This is supported also by our desriptive analysis showing that the median formal onstraints of ommuters lie in the wide rush hour time frame (and outside the narrow rush hour time frame). Nonommuters, on the other hand, are more sensitive to the number of trips permitted with a limited aess travel ard. This shows that there may be some value for limited aess travel ards for nonommuters. Future researh ould examine different designs of the limited aess travel ards that meet the needs of non-ommuters. Further, in a subgroup analysis, we analyzed in detail whih individual harateristis drive the hoie of eah of the aess options to the rush hour. Travel ards with unlimited aess to the rush hour are more attrative to travelers with a more utilitarian travel purposes, i.e., for ommuters that ommute to work and have exat working times or travelers doing business trips. Travel ards with limited aess to the rush hour are preferred by pensioners and other non-ommuters (onfirming the previous results). They are also preferred by travelers with a more workaholi lifestyle. This may be explained by the notion that workaholis have working hours that go beyond the rush hour. Thus, they 51

52 do not frequently travel during the rush hour. The probability of hoosing a limited aess travel ard also dereases the more a traveler regards traveling as utilitarian. Travel ards with no aess to the rush hour are most attrative to seldom travelers as defined by their urrent travel ard ownership. Also ounter to our expetation, we ould not find onlusive evidene of an influene of formal onstraints or soial norms on the hoie between different rush hour aess options. We ran a set of additional models (whih we do not report here) to analyze possible effets of both types of onstraints, but found no lear diretion in the effets. A possible reason may be that we measured formal onstraints and soial norms in an overly generi way. There may different types of formal onstraints and soial norms at the ommuting destination or in the private environment that differently influene travel ard hoies. Future researh ould explore these in detail. Additionally, it might be possible that the aess differene between unlimited aess and the limited aess options we defined as well as no aess was too large to allow for variation due to onstraints. If the limited aess had been more similar to the unlimited aess in terms of the degree of aess it allows, more travelers may atually have had the opportunity to hoose a travel ard with limited aess. Therefore, different designs of limited aess that allow for greater aess than defined in this survey may be a fruitful avenue for future researh Reommendations for SBB Limiting aess to the rush hour to 20 trips per year during the wide rush hour heavily dereases the utility of a travel ard ompared with unlimited aess (the willingness-to-pay dereases of CHF ). Surprisingly, the willingness to pay dereases less when ompletely removing the aess to the wide rush hour (no aess, willingness to pay dereases by CHF ). A possible explanation for this result is the notion that travelers pereive the limitation in form of the number of trips as inonvenient. Travelers want either unlimited aess or no aess at a signifiantly lower prie. Therefore, given these strong sensitivities related to onstraining travelling, we annot reommend priing strategies that impose strong traveling onstraints or that finanially exploit travelers strong need to travel during rush hours over higher pries. Our results also showed that ommuters are generally more likely to purhase a travel ard than non-ommuters, whih is an intuitive result. However, having no aess to the wide (vs. the narrow) rush hour redued the value of a travel ard more for ommuters than for non-ommuters. A travel ard with no aess to the narrow rush hour may be an option for some ommuters. However, when the rush hour is defined as wide, it beame almost useless for them. Implementing limited aess travel ards with a wide rush hour definition are thus likely to attrat signifiantly less travelers. Further, being allowed a higher number of trips during the rush hour inreased utility more for non-ommuters than for ommuters. Therefore, the design of the limited aess travel ards should be studied further while fousing on non-ommuters. The results of our subgroup analysis also hint that there exists a nihe market for travel ards with limited aess to the rush hour. Before implementing suh a travel ard, however, different design options for the limited aess travel ard should be examined. One having these results, the implementation requires a highly targeted approah. 11 Conlusion For both projets, we followed the typial stage-gate proess of produt innovation management, in whih large numbers of opportunities are first identified and then sequentially sreened out until the most promising ones remain, after whih they are subsequently introdued to the market (Terwiesh & Ulrih, 2009). We identified two potential produt innovations in the first stage, whih we empirially investigated further. The first produt innovation idea investigated the utility for dediated setions on the train where only travelers with similar travel needs and habits travel together. The seond produt innovation idea investigated the utilities of different aess options to the rush hour. In the seond stage, we generated insights about these two produt innovations in two separate projets by measuring onsumer preferenes. In the Train Setion Aess Projet, we found evidene in two studies that there is no general preferene for additionally having aess to a dediated setion relative to a ommon setion only. 52

53 Nevertheless, there exists a speifi group of people that prefer aess to a speifi dediated setion (ompared with all other setions). Additionally, the preferene for the dediated setions (vs. the ommon setion) inreases with an inreasing tendeny towards out-group derogation. People that evaluate other people that are different from them negatively, prefer to travel in a setion of the train where they an be separated from those individuals. This way, they are not disturbed by other travelers that show different behaviors. Thus, a mere separation of different types of people reates utility for a subgroup on the market who would be willing to pay for that separation. This may be a possibility for SBB to generate additional revenue with a low finanial investment. As these dediated setions only provide utility to a speifi subgroup of travelers as defined by their demographis, travel harateristis, and psyhographis, priing the dediated setion requires a highly targeted approah. Future researh ould segment the market based on the preferenes for a speifi dediated setion and run market simulations. Market simulations an help define prie points. In the Rush Hour Aess Projet, we found that travelers have a very strong preferene for an unlimited aess to the rush hour ompared with all other options. They want to travel freely with no time onstraints and there is a high willingness to pay for that onveniene. This shows also that for most travelers, being able to travel during the rush hour is ruial. As travelers prefer no aess over limited aess, an implementation of the limited aess travel ards as defined in our experiment is not reommendable for SBB. Nevertheless, speifi subgroups of travelers exist that prefer travel ards with limited aess to the rush hour respetively over travel ards with unlimited aess. Thus, improving the design of the limited aess travel ards in future researh and then implementing them with a highly targeted approah may still be profitable. We also found that ommuters were more sensitive to the definition of the rush hour time frame, while non-ommuters were more sensitive to the number of trips allowed with a limited aess travel ard. This finding in ombination with the results of our subgroup analysis showed that it might be worthwhile to investigate the design of the limited aess travel ard further while fousing on the relevant traveler segment. 53

54 54 12 Referenes Ajzen, I. (1985). From intentions to ations: A theory of planned behavior. Ation ontrol (pp ). Springer. Bekhor, S., & Albert, G. (2014). Aounting for sensation seeking in route hoie behavior with travel time information. Transportation Researh Part F: Traffi Psyhology and Behaviour, 22, Ben-Akiva, M., MFadden, D., Gärling, T., Gopinath, D., Walker, J., Boldu, D.,... Morikawa, T. (1999). Extended framework for modeling hoie behavior. Marketing Letters, 10(3), Ben-Akiva, M., MFadden, D., Train, K., Walker, J., Bhat, C., Bierlaire, M.,... Bunh, D. S. (2002). Hybrid hoie models: Progress and hallenges. Marketing Letters, 13(3), Ben-Akiva, M., Walker, J., Bernardino, A. T., Gopinath, D. A., Morikawa, T., & Polydoropoulou, A. (2002). Integration of hoie and latent variable models. Perpetual motion: Travel behaviour researh opportunities and appliation hallenges, Ben-Akiva, M. E., & Lerman, S. R. (1985). Disrete hoie analysis: Theory and appliation to travel demand (Vol. 9). MIT press. Bentler, P. M. (1980). Multivariate analysis with latent variables: Causal modeling. Annual Review of Psyhology, 31(1), Bhat, C. R. (2001). Quasi-random maximum simulated likelihood estimation of the mixed multinomial logit model. Transportation Researh Part B: Methodologial, 35(7), Bierlaire, M. (2016). PythonBiogeme: A short introdution: Report TRANSP-OR. Bierlaire, M., & Fetiarison, M. (2009). Estimation of disrete hoie models: Extending BIOGEME. Paper presented at the Swiss Transport Researh Conferene (STRC). Bizer, G. Y., Magin, R. A., & Levine, M. R. (2014). The Soial-Norm Espousal Sale. Personality and Individual Differenes, 58, Bloh, U. (2011). Wo es wirklih eng ist. Neue Züriher Zeitung. Retrieved from Boldu, D., & Alvarez-Daziano, R. (2010). On estimation of hybrid hoie models. Paper presented at the Choie Modelling: The State-of-the-Art and the State-of-Pratie: Proeedings from the Inaugural International Choie Modelling Conferene. Emerald Group Publishing. Boldu, D., Bouher, N., & Alvarez-Daziano, R. (2008). Hybrid hoie modeling of new tehnologies for ar hoie in Canada. Transportation Researh Reord: Journal of the Transportation Researh Board, 2082(1), Brewer, M. B. (1979). In-group bias in the minimal tntergroup situation: A ognitive-motivational analysis. Psyhologial Bulletin, 86(2), Cambridge Ditionary. (Ed.) (2017a) Cambridge Ditionary. Cambridge Ditionary. (Ed.) (2017b) Cambridge Ditionary. Cambridge Ditionary. (Ed.) (2017) Cambridge Ditionary. Chang, G.-L., & Mahmassani, H. S. (1988). Travel time predition and departure time adjustment behavior dynamis in a ongested traffi system. Transportation Researh Part B: Methodologial, 22(3), ChoieMetris. (2014). Ngene User Manual & Referene Guide. Choo, S., & Mokhtarian, P. L. (2004). What type of vehile do people drive? The role of attitude and lifestyle in influening vehile type hoie. Transportation Researh Part A: Poliy and Pratie, 38(3), Cialdini, R. B., & Trost, M. R. (1998). Soial influene: Soial norms, onformity and ompliane. Ciari, F., Marmolejo, A., Stahel, A., & Axhausen, K. (2014). Mobility patterns in Switzerland: Past, present and future. Paper presented at the 13th Swiss Transport Researh Conferene. Dasgupta, N. (2004). Impliit ingroup favoritism, outgroup favoritism, and their behavioral manifestations. Soial Justie Researh, 17(2), Davis, J. M. (1984). Attration to a group as a funtion of attitude similarity and geographi distane. Soial Behavior and Personality, 12(1), 1-6.

55 de Bekker-Grob, E. W., Hol, L., Donkers, B., Van Dam, L., Habbema, J. D. F., Van Leerdam, M. E.,... Steyerberg, E. W. (2010). Labeled versus unlabeled disrete hoie experiments in health eonomis: An appliation to oloretal aner sreening. Value in Health, 13(2), Deutshe Bahn. (2014). Deutshe Bahn. Retrieved from Dieplinger, M., & Fürst, E. (2014). The aeptability of road priing: Evidene from two studies in Vienna and four other European ities. Transport Poliy, 36, Domarhi, C., Tudela, A., & González, A. (2008). Effet of attitudes, habit and affetive appraisal on mode hoie: An appliation to university workers. Transportation, 35(5), Eastman, J. K., Goldsmith, R. E., & Flynn, L. R. (1999). Status onsumption in onsumer behavior: Sale development and validation. Journal of Marketing Theory and Pratie, 7(3), Efferson, C., Lalive, R., & Fehr, E. (2008). The oevolution of ultural groups and ingroup favoritism. Siene, 321(5897), Elrod, T. (1991). Internal analysis of market struture: Reent developments and future prospets. Marketing Letters, 2(3), Elrod, T., & Keane, M. P. (1995). A fator-analyti probit model for representing the market struture in panel data. Journal of Marketing Researh, 21(1), Ferdous, N., Pendyala, R., Bhat, C., & Konduri, K. (2011). Modeling the influene of family, soial ontext, and spatial proximity on use of nonmotorized transport mode. Transportation Researh Reord: Journal of the Transportation Researh Board, (2230), Fielding, K. S., MDonald, R., & Louis, W. R. (2008). Theory of planned behaviour, identity and intentions to engage in environmental ativism. Journal of Environmental Psyhology, 28(4), Forward, S. (2004). The predition of travel behaviour using the theory of planned behaviour. Paper presented at the International Conferene of Traffi and Transport Psyhology. Forward, S. E. (2009). The theory of planned behaviour: The role of desriptive norms and past behaviour in the predition of drivers intentions to violate. Transportation Researh Part F: Traffi Psyhology and Behaviour, 12(3), Gaker, D., Zheng, Y., & Walker, J. (2010). Experimental eonomis in transportation: Fous on soial influenes and provision of information. Transportation Researh Reord: Journal of the Transportation Researh Board, (2156), Green, P. E. (1984). Hybrid models for onjoint analysis: An expository review. Journal of Marketing Researh, 21(2), Harris, K. M., & Keane, M. P. (1998). A model of health plan hoie: Inferring preferenes and pereptions from a ombination of revealed preferene and attitudinal data. Journal of Eonometris, 89(1), Hensher, D. A., Rose, J. M., & Greene, W. H. (2005). Applied hoie analysis: A primer. Cambridge University Press. Hewstone, M., Rubin, M., & Willis, H. (2002). Intergroup bias. Annual Review of Psyhology, 53(1), Hofstetter, R., Miller, K. M., Krohmer, H., & Zhang, Z. J. (2013). How do onsumer harateristis affet the bias in measuring willingness to pay for innovative produts? Journal of Produt Innovation Management, 30(5), Hogg, M. A. (1996). Intragroup proesses, group struture and soial identity. Soial groups and identities: Developing the legay of Henri Tajfel, 65, 93. Hogg, M. A., & Hains, S. C. (1996). Intergroup relations and group solidarity: Effets of group identifiation an soial beliefs on depersonalized attration. Journal of Personality and Soial Psyhology, 70(2), Hogg, M. A., & Hardie, E. A. (1992). Prototypiality, onformity and depersonalized attration: A self ategorization analysis of group ohesiveness. British Journal of Soial Psyhology, 31(1), Hogg, M. A., Hardie, E. A., & Reynolds, K. J. (1995). Prototypial similarity self-ategorization and depersonalized attration: A perspetive on group ohesiveness. European Journal of Soial Psyhology, 25(2),

56 Hogg, M. A., & Turner, J. C. (1985). Interpersonal attration, soial identifiation and psyhologial group formation. European Journal of Soial Psyhology, 15(1), Huber, J., & Zwerina, K. (1996). The importane of utility balane in effiient hoie designs. Journal of Marketing Researh, 33(3), Huth, V., & Gelau, C. (2013). Prediting the aeptane of advaned rider assistane systems. Aident Analysis & Prevention, 50, Intervista. (2017). Intervista. Retrieved from Jetten, J., Spears, R., & Manstead, A. S. R. (2001). Similarity as a soure of differentiation: The Role of group identifiation. European Journal of Soial Psyhology, 31(6), Johansson, M. V., Heldt, T., & Johansson, P. (2006). The effets of attitudes and personality traits on mode hoie. Transportation Researh Part A: Poliy and Pratie, 40(6), Jöreskog, K.G. (1970). A general method for estimating a linear strutural equation system. ETS Researh Bulletin Series, 1970(2), i-41. Jost, J. T., Glaser, J., Kruglanski, A. W., & Sulloway, F. J. (2003). Politial onservatism as motivated soial ognition. Psyhologial Bulletin, 129(3), 339. Jou, R. C., Kitamura, R., Weng, M. C., & Chen, C. C. (2008). Dynami ommuter departure time hoie under unertainty. Transportation Researh Part A: Poliy and Pratie, 42(5), Kahneman, D., Knetsh, J. L., & Thaler, R. (1986). Fairness as a onstraint on profit seeking: Entitlements in the market. The Amerian Eonomi Review, Kahneman, D., & Tversky, A. (1979). Prospet theory: An analysis of deision under risk. Eonometria: Journal of the Eonometri Soiety, 47(2), Kahneman, D., & Tversky, A. (1984). Choies, values, and frames. Amerian Psyhologist, 39(4), 341. Kamargianni, M., & Polydoropoulou, A. (2013). Hybrid hoie model to investigate effets of teenagers' attitudes toward walking and yling on mode hoie behavior. Transportation Researh Reord: Journal of the Transportation Researh Board, (2382), Keesling, J. W. (1973). Maximum likelihood approahes to ausal flow analysis. Kinateder, M., Ronhi, E., Gromer, D., Müller, M., Jost, M., Nehfisher, M.,... Pauli, P. (2014). Soial influene on route hoie in a virtual reality tunnel fire. Transportation Researh Part F: Traffi Psyhology and Behaviour, 26(A), Kleijnen, M., De Ruyter, K., & Wetzels, M. (2007). An assessment of value reation in mobile servie delivery and the moderating role of time onsiousness. Journal of Retailing, 83(1), Knokaert, J., Tseng, Y. Y., Verhoef, E. T., & Rouwendal, J. (2012). The Spitsmijden Experiment: A reward to battle ongestion. Transport Poliy, 24, Kuppam, A., Pendyala, R., & Rahman, S. (1999). Analysis of the role of traveler attitudes and pereptions in explaining mode-hoie behavior. Transportation Researh Reord: Journal of the Transportation Researh Board, (1676), Levin, S., & Sidanius, J. (1999). Soial dominane and soial identity in the United States and Israel: Ingroup favoritism or outgroup derogation? Politial Psyhology, 20(1), Lewis, G. J., & Bates, T. C. (2010). Geneti evidene for multiple biologial mehanisms underlying in-group favoritism. Psyhologial Siene, 21(11), Lois, D., Moriano, J. A., & Rondinella, G. (2015). Cyle ommuting intention: A model based on theory of planned behaviour and soial identity. Transportation Researh Part F: Traffi Psyhology and Behaviour, 32, Madanat, S. M., Yang, C., & Ying-Ming, Y. (1995). Analysis of stated route diversion intentions under advaned traveler information systems using latent variable modeling. Transportation Researh Reord, (1485), Manski, C. F. (1977). The struture of random utility models. Theory and Deision, 8(3), MFadden, D. (1973). Conditional logit analysis of qualitative hoie behavior. MFadden, D. (1986). The hoie theory approah to market researh. Marketing Siene, 5(4), MFadden, D., & Train, K. (2000). Mixed MNL models for disrete response. Journal of Applied Eonometris,

57 MGregor, I., Haji, R., & Kang, S. J. (2008). Can ingroup affirmation relieve outgroup derogation? Journal of Experimental Soial Psyhology, 44(5), Miller, K. M., & Hofstetter, R. (2009). Preision priing: Measuring onsumers willingness to pay aurately. BoD Books on Demand. Miller, K. M., Hofstetter, R., Krohmer, H., & Zhang, Z. J. (2011). How should onsumers willingness to pay be measured? An empirial omparison of state-of-the-art approahes. Journal of Marketing Researh, 48(1), Morikawa, T., Ben-Akiva, M., & MFadden, D. (2002). Disrete hoie models inorporating revealed preferenes and psyhometri data. Advanes in Eonometris, 16, Myers, D. G. (2008). Psyhologie, 2. Aufl., Heidelberg. Nguyen, D. Q., & Mariani, D. (2014). Swiss population getting larger, older, more diverse. Retrieved from more-diverse/ NSB. (2014). NSB. Retrieved from NTV. (2014). Italo. Retrieved from ÖBB. (2014). ÖBB. Retrieved from Ozbay, K., & Yanmaz-Tuzel, O. (2008). Valuation of travel time and departure time hoie in the presene of time-of-day priing. Transportation Researh Part A: Poliy and Pratie, 42(4), Paris, H., & Van den Brouke, S. (2008). Measuring ognitive determinants of speeding: An appliation of the theory of planned behaviour. Transportation Researh Part F: Traffi Psyhology and Behaviour, 11(3), Redmond, L. (2000). Identifying and analyzing travel-related attitudinal, personality, and lifestyle lusters in the San Franiso Bay Area. Institute of Transportation Studies. Redmond, L. S., & Mokhtarian, P. L. (2001). The positive utility of the ommute: Modeling ideal ommute time and relative desired ommute amount. Transportation, 28(2), Renfe. (2014). Renfe. Retrieved from Riedener, J. (2012). Überfüllte Züge während Stosszeiten. 20 Minuten. Retrieved from Riggs, W. (2017). Painting the fene: Soial norms as eonomi inentives to non-automotive travel behavior. Travel Behaviour and Soiety, 7, Rose, J. M., & Bliemer, M. C. (2009). Construting effiient stated hoie experimental designs. Transport Reviews, 29(5), Saameli, N. (2014). Handy-Lärm nervt die Pendler. 20 Minuten. Retrieved from SBB. (2013). Segmentierungsalgorithmus. SBB. (2014). SBB. Retrieved from SBB. (2017). SwissPass - the key to your mobility. Retrieved from Sagnolari, S., & Maggi, R. (2010). Modelling heterogeneity in young persons night-time hoies - aounting for attitude towards alohol onsumption. Paper presented at the 10th Swiss Transport Researh Conferene, Asona, Switzerland. Shade, J., & Shlag, B. (2003). Aeptability of urban transport priing strategies. Transportation Researh Part F: Traffi Psyhology and Behaviour, 6(1), Sott, D. M., & Axhausen, K. W. (2006). Household mobility tool ownership: Modeling interations between ars and season tikets. Transportation, 33(4), Senbil, M., & Kitamura, R. (2004). Referene points in ommuter departure time hoie: A prospet theoreti test of alternative deision frames. Paper presented at the Intelligent Transportation Systems. Senbil, M., & Kitamura, R. (2005). Heterogeneity in ommuter departure time deision: A prospet theoreti approah. Applied Researh in Unertainty Modeling and Analysis (pp ). Springer. 57

58 Sherwin, H., Chatterjee, K., & Jain, J. (2014). An exploration of the importane of soial influene in the deision to start biyling in England. Transportation Researh Part A: Poliy and Pratie, 68, SJ. (2014). SJ. Retrieved from SNCB. (2014). Belgianrail. SNCF. (2014). SNCF. Retrieved from Stangor, C., & Thompson, E. P. (2002). Needs for ognitive eonomy and self enhanement as unique preditors of intergroup attitudes. European Journal of Soial Psyhology, 32(4), Swiss Federal Statistial Offie (2014). Population and Households Statistis. Tajfel, H., & Turner, J. C. (1979). An integrative theory of intergroup onflit. The Soial Psyhology of Intergroup Relations, 33(47), 74. Temme, D., Paulssen, M., & Dannewald, T. (2008). Inorporating latent variables into disrete hoie models a simultaneous estimation approah using SEM software. BuR-Business Researh, 1(2), Terwiesh, C., & Ulrih, K. T. (2009). Innovation tournaments: Creating and seleting exeptional opportunities. Harvard Business Press. Thurstone, L. L. (1927). A law of omparative judgment. Psyhologial Review, 34(4), 273. Tillema, T., Ben-Elia, E., Ettema, D., & van Delden, J. (2013). Charging versus rewarding: A omparison of road-priing and rewarding peak avoidane in the Netherlands. Transport Poliy, 26, Train, K. (2003). Disrete hoie methods with simulation. Cambridge University Press. Trenitalia. (2014). Trenitalia. Retrieved from Turner, J. C., Hogg, M. A., Oakes, P. J., Reiher, S. D., & Wetherell, M. S. (1987). Redisovering the soial group: A self-ategorization theory. Basil Blakwell. Tversky, A., & Kahneman, D. (1975). Judgment under unertainty: Heuristis and biases. Utility, probability, and human deision making (pp ). Springer. Ungriht, T. (2010). Überfüllte SBB-Züge - Die Erfindung des Steh-Pendlers. Blik. Retrieved from Valda, A. (2015). Möglihe Trendwende» bei überlasteten Zügen. Tagesanzeiger. Retrieved from von Sivers, I., Templeton, A., Köster, G., Drury, J., & Philippides, A. (2014). Humans do not always at selfishly: Soial identity and helping in emergeny evauation simulation. Transportation Researh Proedia, 2, Vontobel, W., & Guanziroli, S. (2008). Überfüllte Züge mahen Reisende sauer. Blik. Retrieved from Vrti, M., Shüssler, N., Axhausen, K., & Erath, A. (2011). Mobility priing: Zahlungsbereitshaft und Verhaltensreaktionen. Heureka 11 "Optimierung in Verkehr und Transport 2011 in Stuttgart"(FGSV 002/96). Wakefield, K. L., & Inman, J. J. (2003). Situational prie sensitivity: The role of onsumption oasion, soial ontext and inome. Journal of Retailing, 79(4), Walker, J., & Ben-Akiva, M. (2002). Generalized random utility model. Mathematial Soial Sienes, 43(3), Walker, J. L. (2001). Extended disrete hoie models: Integrated framework, flexible error strutures, and latent variables. (Dotor of Philosophy), Massahusetts Institute of Tehnology. Weihbrodt, J., Sprenger, M., Steffen, M., Tanner, A., Meissner, J. O., & Shulze, H. (2013). WorkAnywhere. Retrieved from Wikipedia. (2009, ). Metropolitan. Retrieved from Wiley, D. (1973). The identifiation problem for strutural equations models with unmeasured variables. Paper presented at the Strutural Equation Models in the Soial Sienes, AS Goldberger and OD Dunan (Eds). Seminar Press: New York. 58

59 Zemp, T. (2014). SBB-Mitarbeiter verstopfen Pendlerzüge. Tagesanzeiger. Retrieved from Pendlerzuege/story/ Zhang, D., Shmöker, J. D., Fujii, S., & Yang, X. (2016). Soial norms and publi transport usage: Empirial study from Shanghai. Transportation, 43(5), Zhang, Z., Fujii, H., & Managi, S. (2014). How does ommuting behavior hange due to inentives? An empirial study of the Beijing Subway System. Transportation Researh Part F: Traffi Psyhology and Behaviour, 24, Zhong, C. B., Phillips, K. W., Leonardelli, G. J., & Galinsky, A. D. (2008). Negational ategorization and intergroup behavior. Personality and Soial Psyhology Bulletin, 34(6), Appendix 13.1 Appendix A: Ideation Session Degree of Innovativeness Table 8 Top Ten Ideas Aording to the Degree of Innovativeness Rank Idea Average Innovativeness 1 Books for rent (from station to station) Spooky train, e.g. as a Halloween event tour Soial wagon to get in touh with others / silent wagon Party trips with dane and drink/bar wagons SBB-assistant for kids travelling alone Tiked the prie of whih is based on the distane (e.g. CHF 1.- per km) More prie differentiation by travel time and destination Mobile app for passengers to ommuniate pakage servie by train (from station to station) Playground for kids at the trak 5.67 Table 9 Top Ten Ideas Aording to the Expeted Utility for the Target Segment Expeted utility for the target segment Rank Idea Average Rating 1 Free Wi-Fi in the trains Loyalty program to ollet points Push messages on the phone for delayed trains Wi-Fi in the train rail & fly tikets, e.g. in ooperation with Swiss Air Cheaper tikets for trips that are booked a long time in advane Tiket that is valid for a ertain route for a ertain number of times Power outlets in the train Digital information system in the train (stops, delays, et.) Tiketing mahine in the train

60 60 Impat on the brand image Table 10 Top Ten Ideas Aording to the Impat on the Brand Image Rank Idea Average Rating 1 Speial tikets for speial days (mother s day, Valentine s day et.) SBB-assistant for kids on their trips Pik up servie for women until late in the night Assistane for seniors Free musi and e-books for SBB passengers Free Wi-Fi in the trains rail & fly tikets, e.g. in ooperation with Swiss Air Digital library in the trains Books for rent (from station to station) Bike sharing programs 5.67 Table 11 Top Ten Ideas Aording to the Degree of Innovativeness, Expeted Utility for the Target Segment, Impat on the Brand Image Overall evaluation Rank Idea Average Rating 1 SBB-assistant for kids on their trips Loyalty program to ollet points Free Wi-Fi in the trains rail & fly tikets, e.g. in ooperation with Swiss Air Digital library in the trains Partner ard at half fare for travel partner Tiked the prie of whih is based on the distane (e.g. CHF 1.- per km) 5.22 Pik up servie for women until late in the night 5.22 Books for rent (from station to station) 5.22

61 Appendix B: Desriptive Results Table 12 Correlation Between Latent Construts (Train Setion Aess Projet, Study 1) Out-group derogation Family oriented life-style Status-seeking life-style Workaholi life-style Frustrated life-style Time-onsiousness Hedonism Travel liking Commute benefits Travel stress Desire for omfort Desire for flexibility Travelling hedoni vs. utilitarian Out-group derogation Family oriented life-style.14 Status-seeking life-style Workaholi life-style.29 < Frustrated life-style Time-onsiousness Hedonism Travel liking Commute benefits Travel stress Desire for omfort < a.01 Desire for flexibility < b Travelling hedoni vs. utilitarian b a < Note: a Signifiant at 10% level; b Signifiant at 5% level; Signifiant at 1% level. Table 13 Relation of Out-Group Derogation to Other Psyhographis (Train Setion Aess Projet, Study 1) Variable Relation Mean at low outgroup Mean at high out- t-value 29 p-value derogation group derogation Latent variables Family /ommunity oriented >.10 life-style Workaholi life-style <.10 a Status-seeking life-style <.01 Frustrated life-style <.01 Travel-liking >.10 Hedonism >.10 Time-onsiousness <.01 Desire for omfort >.10 Desire for flexibility >.10 Traveling hedoni (vs >.10 utilitarian) Commute benefits <.01 Travel stress <.01 Note: a Signifiant at 10% level; b Signifiant at 5% level; Signifiant at 1% level. 29 The degrees of freedom for all analyses are 205 if not indiated differently.

62 Table 14 Relation of Out-Group Derogation to Travel-Related Variables (Train Setion Aess Projet, Study 1) Variable Relation Mean at low outgroup Mean at high out- t-value 30 p-value derogation group derogation Trip purpose Commuting to and from work >.10 Commuting to and from an >.10 eduational institution Leisure trip without overnight >.10 stay Private vaation trip with at >.10 least one overnight stay Everyday tasks >.10 Business trip >.10 Other trip purpose >.10 Travel ard ownership GA travel ard >.10 Half-Fare travel ard >.10 Point-to-Point travel ard >.10 Trak 7 travel ard < >.10 Regional travel ard > Travel lass Seond lass (vs. first lass) predominant travel lass Commuting trip Commuting days per week (only ommuters) 31 Trip distane (only ommuters) 32 Commuting during rush hour (vs. not) 33 Commuting with formal time onstraints at ommuting destination (vs. no ommuting and ommuting without formal time onstraints) > > > > <.10 a Leisure trip Frequeny >.10 Trip distane >.10 Note: a Signifiant at 10% level; b Signifiant at 5% level; Signifiant at 1% level. 30 The degrees of freedom for all analyses are 205 if not indiated differently. 31 The degrees of freedom for this analysis are The degrees of freedom for this analysis are The degrees of freedom for this analysis are Lower values mean higher frequeny, but variable is ordinal. 35 Replaed values don t know and not appliable with the mean.

63 63 Table 15 Relation of Out-Group Derogation to Seletion of Demographis (Train Setion Aess Projet, Study1) Variable Relation Mean at low outgroup Mean at high out- t-value 36 p-value derogation group derogation Demographis Age <.01 Gender male (vs. female) >.10 Last ompleted eduation: >.10 ompulsory shool (vs. other) Last ompleted eduation: >.10 professional shool / apprentieship / voational shool (vs. other) Last ompleted eduation: >.10 middle shool / high shool (vs. other) Last ompleted eduation: >.10 higher shools like universities / institutes of tehnology / tehnial ollege / higher voational shools (vs. other) Number of people in household >.10 Employment perentage >.10 Main oupation: employed >.10 Main oupation: student <.10 a Main oupation: housewife / >.10 homemaker Main oupation: pensioner >.10 Main oupation: unemployed >.10 Main oupation: other >.10 Note: a Signifiant at 10% level; b Signifiant at 5% level; Signifiant at 1% level. Table 16 Correlation Between Latent Construts (Train Setion Aess Projet, Study 2) Out-group derogation Family oriented life-style Status-seeking life-style Workaholi life-style Frustrated life-style Time-onsiousness Travel liking Hedonism Commute benefits Travel stress Desire for omfort Desire for flexibility Out-group derogation Family oriented life-style.02 a Status-seeking life-style Workaholi life-style Frustrated life-style Time-onsiousness Travel liking Hedonism < Commute benefits Travel stress b -.38 Desire for omfort Desire for flexibility < a Travelling hedoni vs. utilitarian < a < a Note: a Signifiant at 10% level; b Signifiant at 5% level; Signifiant at 1% level. 36 The degrees of freedom for all analyses are 205 if not indiated differently.

64 64 Table 17 Correlation Between Latent Construts (Rush Hour Aess Projet, Part 1) Pereived onstraints (professional) Pereived onstraints (private) Soial norms ompliane (professional) Soial norms ompliane (private) Duration norm Flexibility norm Desriptive duration norm Desriptive flexibility norm Family oriented life-style Status-seeking life-style Workaholi life-style Frustrated life-style Pereived onstraints (professional) Pereived onstraints (private) 1.00 Soial norms ompliane (professional) Soial norms ompliane (private) Duration norm Flexibility norm Desriptive duration norm Desriptive flexibility norm Family oriented life-style Status-seeking life-style Workaholi life-style.02 a.02 a.03 b.03 b.02 a.02 a a < Frustrated life-style Time-onsiousness Travel liking b Hedonism a a Commute benefits < Travel stress Desire for omfort b Desire for flexibility <.01 <.01 <.01 <.01 <.01 <.01 <.01 < a.15 Travelling hedoni vs. utilitarian Note: a Signifiant at 10% level; b Signifiant at 5% level; Signifiant at 1% level. Table 18 Correlation Between Latent Construts (Rush Hour Aess Projet, Part 2) Time-onsiousness Travel liking Hedonism Commute benefits Travel stress Desire for omfort Desire for flexibility Travel liking.09 b Hedonism a Commute benefits -.09 a -.49 <.-01 Travel stress Desire for omfort a -.09 b Desire for flexibility < Travelling hedoni vs. utilitarian Note: a Signifiant at 10% level; b Signifiant at 5% level; Signifiant at 1% level.

65 Appendix C: Survey Instrutions Figure 24 Desription of Train Setion Aess (Train Setion Aess Projet, Survey 1) Figure 25 Desription of Train Setion Aess (Train Setion Aess Projet, Survey 2)

66 Figure 26 Desription of Rush Hour Aess (Rush Hour Aess Projet) 66