Characteristics of Premium Transit Services that Affect Choice of Mode (2014)

Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

8C H A P T E R 2 Current practice in regional travel forecasting models typically considers the effects of travel times, wait times, frequencies, travel costs, and transfers when evaluating the benefits of transit services and estimating ridership. In many cases, however, models in metropolitan areas with existing rail services require large adjustments to replicated observed ridership patterns. These adjustments usually are designed to increase modeled rail ridership to match observed (counted) values. These adjustments can take several forms, including: • Defining rail as a separate mode in the mode choice model and assigning a mode-specific constant that reflects less perceived times and costs for a rail journey than for a similar bus trip; and • Adjusting the perceived in-vehicle travel time for rail modes so that a minute of time on the train is less onerous than a minute of travel time on the bus. These adjustments vary from metropolitan area to metropolitan area, suggesting that these parameters are not easily transferred without a better understanding of what causes travelers to prefer fixed guideway services to similar bus options. Furthermore, defining rail as a separate mode introduces a series of potential problems when this type of model is used to analyze transit alternatives. Potential issues include: • Mode Definition and Hierarchy. Individual modeled modes are usually organized into a hierarchy of modes with rail being the highest and bus being the lowest. This structure can create counterintuitive results. A typical example occurs when a new rail line is added to an existing system. Existing bus-to-rail trips might be converted to rail-only trips. The model, however, sees only that the trips are defined as rail in both cases and therefore would not assign any value to this conversion beyond whatever time and cost improvements are associated with this project. • Arbitrary Labels and Impedances. These are defined based on vehicle technology rather than service attributes. Not all buses and trains are the same. Some buses operate over-the-road coaches with seating for all travelers and on-board Wi-Fi service. Some trains are crowded rapid transit services with high levels of crowding and lower comfort levels. Service attributes can be included in the development of travel impedances and mode shares in lieu of arbitrary labels to better represent the service being offered. Both potential problems suggest that models could be more robust if they focused more on understanding the impact of a broader range of service characteristics and less on the definition of individual transit submodes. Potentially important transit service attributes not typically considered in transit forecasting models include: • Station or stop design features that provide real-time information about the next transit arrival/departure, security, lighting/safety, shelter, cleanliness of the station, benches, and proximity to services; Important Non-Traditional Transit Attributes

Important Non-Traditional Transit Attributes 9 • On-board features that address seating availability, seating comfort, temperature, cleanliness of the transit vehicle, ease of boarding, and productivity features (e.g., Wi-Fi, power outlets, etc.); and • Other features, such as identification of the transit vehicle, schedule reliability, schedule span, and fare machines. This research effort serves to improve the transit industry’s knowledge of the importance of this broader set of important transit service attributes, focusing on those attributes listed above that are not traditionally considered. Defined in this report as non-traditional attributes, these attributes can influence forecasting models in three distinct ways, by: 1. Presenting a complete picture of the attractiveness of a transit option when calculating the likelihood of using transit or a specific transit mode; 2. Accounting for the fact that travelers have different levels of awareness and willingness to consider different transit options; and 3. Incorporating the effect that traveler attitudes have on the likelihood of using transit and selecting specific transit modes. Effects on the Attractiveness of Transit Key Findings The research team found that non-traditional transit service attributes are important factors in decisions about whether to use transit and which transit service to use. Taken together, the importance of non-traditional transit service attributes is equivalent to 13 to 29 minutes of in-vehicle travel time (depending on the city and the purpose of the trip). Recognizing that specific transit routes either do or do not include each of these non-traditional service attributes, accounting for them properly can have a large effect on the relative attractiveness of each route, and therefore on the measurement of the benefits of each transit option. Research Methods The research team designed an advanced travel survey to support a better understanding of transit choice behavior and specifically evaluate the importance of non-traditional transit service attributes. The non-traditional service attributes considered in this research are included in Table 1. The survey consisted of the following four sections: 1. Demographic and travel characteristics; 2. Attitudes about transit; 3. Ranking of different non-traditional attributes; and 4. Selection of transit options with varied attributes for a typical trip a person makes. This survey was specifically designed so that respondents would make trade-offs between different service attributes, and thereby allowed use of mathematical modeling techniques to value the importance of each attribute in the choice of transit options. The research team designed the survey to understand the relative importance of different levels of comfort, convenience, safety, and other non-traditional transit attributes in mode choice decisions, and to further under- standing of how different people in different contexts have different values for these attributes. Figure 2 presents an example of a trade-off experiment used in the survey. Five transit attributes are featured in the specific example shown in Figure 2. In the survey itself, the respondent would see eight experiments in which the attributes were varied, allowing Non-traditional attributes not typically considered in transit forecasting or planning include station amenities, on-board amenities, and other features, such as reliability.

10 Characteristics of Premium Transit Services that Affect Choice of Mode Bundle Attribute Premium Characteristics Standard Characteristics St ati on /s to p de si gn fe at ur es Real time information about next transit arrival/departure Real time information available No real time information available Station/stop security Enhanced (e.g., emergency call buttons, surveillance cameras, security personnel) No added security features Station/stop lighting/safety Well lit with police presence Normal lighting and no police presence Station/stop shelter Effectively protects you from bad weather Limited or no shelter Proximity to services Close to coffee shop, dry cleaners, grocery, etc. Not close to coffee shop, dry cleaners, grocery, etc. Cleanliness of station/stop Wellmaintained and clean Not wellmaintained Station/stop benches Clean and comfortable Some benches O n bo ar d fe at ur es On board seating availability Always available seats Often crowded; youmight not get a seat On board seating comfort Seats are comfortable and a good size Seats are standard On board temperature Effective air conditioning and heating Some air conditioning and heating Cleanliness of transit vehicle Very new and clean Maintained, but not new Productivity features Wi Fi, power outlets, etc., available Productivity features not available O th er fe at ur es Route name/number identification Easy to identify on outside of transit vehicle Difficult to immediately identify on outside of transit vehicle Reliability One in ten trips are 5minutes late or more One in ten trips are 15minutes late or more Schedule span Transit runs from 4:00 a.m. until 11:00 p.m. Transit runs during rush hours only Transit frequency Arrives every 10minutes in rush hour and every 20minutes in off peak Arrives every 20minutes in rush hour and every 60minutes in off peak Transfer distance Convenient (short walking distance or on same platform) Several minutes’ walk Station/stop distance Within 10minutes’ walk of your home/work Not within 10minutes’ walk of your home/work Parking distance Within 10minutes’ walk from station/stop Not within 10minutes’ walk from station/stop Ease of boarding Easy to board; doors are level with platform/curb Must step up to board Faremachines Fast and easy to use Slow and somewhat confusing Table 1. Non-traditional transit service attribute levels in survey.

Important Non-Traditional Transit Attributes 11 consideration of a wide range of attributes without imposing undue burden on the respondent in any one experiment. This example in Figure 2 provides only one glimpse into a complex survey, but serves to provide context for similar experimental survey methods. More informa- tion can be found in Appendix B. Research Results Once the data were collected, specialized mathematical techniques were used to assess the relative importance of different features. This mathematical exercise resulted in an assess- ment of the importance of each non-traditional attribute in relation to attributes that transit planners and modelers often consider. This value was expressed as equivalent minutes of in-vehicle transit travel time. The concept is analogous to the idea that non-monetary factors (e.g., time or personal injury) can have dollar values for use in economic assessments. Taken together, the importance of non-traditional transit service attributes was valued as equivalent to 13 to 29 minutes of travel time (depending on the city and the trip purpose). Table 2 presents the details underlying that finding for each city and service attribute. Although the combined value of the various premium transit service attributes is significant in all cities and for all purposes, it is also clear that travelers in different cities value different features of the transit system in very different ways. The differences suggest that survey research may be required to estimate similar factors in order to apply this approach in new cities that plan to apply these findings in practice. Figure 2. Example trade-off experiment from the Salt Lake City survey.

12 Characteristics of Premium Transit Services that Affect Choice of Mode Effects on Awareness and Consideration of Transit Options The next potential contribution of non-traditional attributes involves traveler awareness of individual transit options and the degree to which travelers are willing to consider using these options. Inclusion of awareness and consideration in travel forecasting models is a relatively new concept. To date, models typically assume that all modes are available and considered by all individuals or apply simple deterministic rules to sort out whether certain modes are available and considered by an individual. Examples of the latter approach include applying a rule that individuals residing in zero-car households are assumed to not have “drive alone” available, or that individuals residing more than one-half mile from a transit stop are assumed not to have “walk to transit” available in the mode choice model. A more comprehensive approach for determining whether transit is considered as a modal alternative may be influenced by numerous factors. These factors may not have much to do with the physical availability of the mode per se. Personal and household constraints (e.g., the need to drop off a child at school on the way to work), individual attitudes, perceptions, preferences, and simple lack of awareness (information) may all contribute to the non-consideration of transit as a viable modal alternative. Awareness of travel options and factors that affect consideration often are related to individual socioeconomic circumstances that may not be evenly distributed across a metropolitan area. Better understanding of these factors and how they work together to forecast transit usage can improve forecasting procedures. Technical Details In technical modeling terms, the survey approach was designed to support maxi- mum difference scaling (MaxDiff) modeling and choice-based conjoint modeling (choice modeling). MaxDiff measures the importance of individual transit service characteristics with respondents choosing the best and worst options from a set of alternatives. In TCRP Project H-37, eight maximum difference experiments were conducted in each of the three surveys. Choice modeling measures the stated preference of a combination of transit service characteristics with respondents choosing the best alternative. In this project eight stated preference experiments were conducted in each of the three surveys. Both survey approaches were analyzed jointly using multinomial logit (MNL) estimation techniques to identify the relative importance of non-traditional service attributes, while also considering the value of traditional service attributes (i.e., time, cost, and frequency). Current practice in transit and mode choice modeling typically results in a model that is sensitive to the effects of travel times, wait times, frequencies, travel costs and transfers, in addition to mode-specific constants. In theory the mode-specific constants capture the differences in the unobserved attributes of modes, but the constants are also adjusted to match observed ridership volumes and therefore help correct other errors in the travel model system. The goal of TCRP Project H-37 was to improve the reasonableness and interpretability of mode choice models, reducing the extent to which the resulting mode choice model constants dominate the modeled utilities. For more information, the details of the transit service attribute models are presented in Appendix D and the multinomial logit mode choice models are presented in Appendix E.

Important Non-Traditional Transit Attributes 13 Key Findings Three key findings relate to travelers’ awareness and consideration of transit options: 1. Many travelers are not aware of, nor do they consider, transit options that travel models represent as available for their trip. Providing options beyond those considered by travelers will bias the mode choice models because awareness and consideration are more a function of demographics, latent variables, and traveler attitudes than of transit service attributes. 2. Travelers are aware of and consider train alternatives more often than bus. This finding is determined directly from the travel surveys, based on questions about travelers’ consideration of bus and rail modes once availability is accounted for. 3. Incorporating awareness and consideration of transit into statistical estimation work improves the statistical fit of the mode choice models. Mode choice models, estimated with and without awareness and consideration models constraining the choice sets, demonstrated statistical improvement with the inclusion of these models. Attribute Commute Trips Non commute Trips Charlotte Salt Lake City Chicago Charlotte Salt Lake City Chicago Station/stop design features 3.71 4.61 4.97 9.06 1.57 4.42 Real-time information 0.40 * 0.62 1.06 * 0.44 Station/stop security 0.60 0.88 0.85 1.56 0.22 0.84 Station/stop lighting/safety 0.66 0.88 0.86 1.62 0.20 0.82 Station/stop shelter 0.64 1.10 0.86 1.57 0.37 0.69 Proximity to services 0.40 0.84 0.40 0.89 0.47 0.50 Cleanliness of station/stop 0.73 0.42 0.90 1.74 0.15 0.86 Station/stop benches 0.28 0.49 0.48 0.62 0.16 0.27 On-board features 4.58 3.53 5.84 9.47 3.8 10.79 On-board seating availability 1.46 1.23 2.15 3.32 1.41 4.09 On-board seating comfort 0.56 0.51 0.77 1.02 0.41 1.39 On-board temperature 1.20 0.81 1.41 2.42 0.85 2.41 Cleanliness of transit vehicle 0.60 0.44 0.64 1.26 0.39 1.56 Productivity features 0.76 0.54** 0.87 1.45 0.74** 1.34 Other features 8.94 4.92 11.17 10.60 6.14 9.77 Route name/number identification 0.57 0.60 0.63 1.23 0.58 0.61 Reliability 4.59 0.44*** 5.64 0.29*** 4.63 Schedule span 0.52 0.42 0.77 1.47 0.33 0.82 Transit frequency 0.60 0.75 0.82 1.49 0.38 0.71 Transfer distance 0.46 0.72 0.56 1.29 0.12 0.48 Station/stop distance 0.80 0.64 0.92 1.76 0.13 0.84 Parking distance 0.72 0.54 0.84 1.44 0.17 0.71 Ease of boarding 0.08 0.16 0.21 0.52 3.02 0.25 Fare machines 0.60 0.65 0.78 1.40 1.12 0.72 All premium service features 17.23 13.06 21.98 29.13 11.51 24.98 *The attribute was not part of the station/stop design features bundle in the survey for Salt Lake City. ** The attribute was referred to simply as “Wi-Fi” in the survey for Salt Lake City. ***The reliability measure was redefined in the survey for Chicago and Charlotte, so this value is not comparable to the value for Salt Lake City. Table 2. Importance of non-traditional transit service attributes (equivalent minutes of in-vehicle travel time).

14 Characteristics of Premium Transit Services that Affect Choice of Mode This research focused primarily on key findings related to the importance of premium service characteristics and their effect on awareness and consideration, as opposed to broader modeling considerations that go beyond service characteristics. Research Methods Questions about awareness and consideration of transit alternatives were included in the surveys for all three cities surveyed. In the initial survey for Salt Lake City, these questions were exploratory. In the second set of surveys, for Charlotte and Chicago, these questions were more systematic and comprehensive to allow for model estimation of awareness and consider- ation. The following list shows some of the issues related to transit awareness and consideration explored in the Chicago and Charlotte surveys: • Do the survey respondents know the routes serviced at the public transit stop within walking distance of their homes? • Do they know how to travel to where they work, go to school, or places where they went on their most recent trips from the public transit stop within walking distance of their home? • What other types of transportation could they have used for their most recent trip? • Why didn’t they use the transit options available on their most recent trip? • What did they need their car for on their most recent trip? • What about the transit service didn’t meet their needs for their most recent trip? • What other types of public transit did they consider using to make this trip? • For the trip they made, did they know they had an alternative option (together with the associated time, required transfers, and costs of that option)? • Why would they not consider the alternative transit mode option? Survey respondents also were asked to say how informed they are about the survey area’s public transit services in terms of types of service available, routes, schedules, fare options, and so forth (see Figure 3). These survey results demonstrated that one-quarter to one-third of survey respondents are uninformed about transit, while travel forecasting models represent all travelers having full information. Figure 3. Survey respondents’ indications that they are informed about transit for Charlotte, Chicago, and Salt Lake City.

Important Non-Traditional Transit Attributes 15 Awareness and consideration models were developed to identify (1) whether travelers are aware of a transit alternative and (2) whether travelers will consider the transit alternative. The results of these models were used to constrain the choices available to travelers in the mode choice models. Awareness and consideration of transit are handled using choice set models as part of the following two-step decision process: Step 1. An individual’s awareness of an option must be determined based on demographic, trip, and attitudinal characteristics. Step 2. The willingness of an individual to consider an alternative must be determined based on awareness and demographic, trip, and attitudinal characteristics. The complete choice set for each individual is formed because of awareness and consideration of the transit options (bus and rail). It is assumed that an individual who has a car available to make the trip is aware of the option to use it and always considers it in the choice set. Consequently, the car option enters the choice set in a deterministic way. Research Results Direct analysis of the surveys provides evidence that typical models overstate the availabil- ity of transit options as compared to the options that are reported by respondents as being available. As shown in Figure 4, respondent awareness is less than the network representation of transit availability for all cities and transit submodes. The differences between respondent awareness and the network representation of bus availability are consistent across all three cities (16% less for Charlotte and Salt Lake City and 13% less for Chicago). The differences between respondent awareness of and network representation for rail were smaller than for bus in two cities Technical Details In technical modeling terms, awareness and consideration were examined using joint bivariate binary probit models to first identify whether travelers were aware of a transit alternative and then to constrain these choices to identify whether travelers would consider the transit alternative. The Joint Bivariate Binary Probit model is a generalization of the probit model that is used to estimate several correlated binary outcomes jointly. The results of these models were used to constrain the choices available to travelers in the mode choice models. This study explicitly accounts for attitudes, perceptions, and values in modeling transit awareness and consideration. The models in this study consider attitudinal factors as possible explanatory variables to account for factors that are tradition- ally unmeasured, unobserved, and relegated to being absorbed in the random error term. A key question that merits consideration is the extent to which modal level-of- service variables should enter the awareness and consideration model specifications. It may be hypothesized that people are more aware of and would give greater consideration to transit modes when transit level of service is greater, more com- petitive with the automobile, and of high quality. In the current study, transit awareness and consideration is modeled whenever transit is available. More information is presented in Appendix F.

16 Characteristics of Premium Transit Services that Affect Choice of Mode (6% less for Charlotte and 7% less for Chicago). In Salt Lake City, however, travelers were 25% less likely to be aware of rail options than was suggested by the network models. These results may reflect real differences in awareness or different assumptions in the network representation across cities. Table 3 reports the survey results for consideration of transit alternatives in Chicago and Charlotte for bus and rail modes. In Charlotte, 71% of travelers who report having rail as an available mode would consider taking the train, whereas only 55% of travelers who report having an available bus option would consider taking bus. In Chicago, those percentages are 83% and 56%, respectively. Even among travelers willing to consider a given mode of transit, a higher proportion selects rail than selects bus. Sequential models were estimated for awareness and consideration, with consideration models limited to choices that travelers were aware of. Bus and train were represented as individual Note: The awareness questions in the Salt Lake City survey were changed when conducting the Charlotte and Chicago surveys, so these results may not be directly comparable. 0% 10% 20% 30% 40% 50% 60% 70% 80% Respondents aware of bus availability Network representa
on of bus availability for respondents’ trip Respondents aware of train availability Representa
on of train availability for respondents’ trip Charloe Chicago Salt Lake City Pe rc en t o f S ur ve y Re sp on de nt s Figure 4. Respondents’ awareness of bus and rail modes available for a trip. Note: Total available in this context represents availability reported by the respondents. Charlotte Chosen 191 156 50% 62% Not Chosen 189 96 50% 38% Chosen 207 429 62% 69% Not Chosen 126 190 38% 31% 592 745 100% 100%Total Available Not Considered 259 126 44% 17% Considered 333 619 56% 83% Chicago 354 100% 100%Total Available 690 Not Considered 310 102 45% 29% Considered 380 252 55% 71% Bus Train Bus Survey Respondents Percent of Total Train Table 3. Consideration of bus and rail modes.

Important Non-Traditional Transit Attributes 17 choice alternatives in both the awareness and consideration models. One primary question for these models is whether representing a traveler’s awareness and consideration of transit will improve the ability of the mode choice model to explain travel behavior. Mode choice models were estimated with and without awareness and consideration constraints to evaluate the statistical improvement in the models by accounting for these choice set constraints: • In Chicago, final log-likelihood was 5790 and 4720 for commute trips and non-commute trips, respectively; with awareness and consideration models to constrain, the choice set was 5908 and 4870 without these constraints. • In Charlotte, the final log-likelihood was 7134 and 3373 for commute trips and non-commute trips, respectively; with awareness and consideration models to constrain, the choice set was 7250 and 3278 without these constraints. Log-likelihoods represent the likelihood that a given function describes the probabilities that underlie the data in these surveys. The difference in log-likelihood here is significant, based on a statistical goodness-of-fit test (chi-squared) of approximately 100 points difference in log-likelihood resulting in significance beyond the 0.01 level. These results demonstrate that the models that include awareness and consideration are significantly better than the models without awareness and consideration, based on the estimation of the models; however, further research is necessary to evaluate the difference in the model predictions of transit ridership. The Role of Traveler Attitudes The third role for non-traditional attributes is in determining how traveler attitudes affect transit usage. Attitudes were obtained from travelers on driving, walking, and taking transit. These traveler attitudes and their impact on transit ridership were evaluated in three different cities using sequential estimation of traveler attitudes and modes and simultaneous estimation of traveler attitudes and modes. Both for sequential and simultaneous estimation, the traveler attitudes enhanced the estimation of the mode choice models by complementing the other socioeconomic factors represented in the models. In all three cities, the attitudes affected the choice of transit versus automobile much more than the choice of bus versus rail. Key Findings There is evidence that different attitudes about transportation affect the choice between transit and automobile. Although this is interesting and supported by other research, it was not the focus of this research and so it was not explored further. Based on model estimation results, and in Chicago and Charlotte specifically, there is no evidence that attitudes about transportation affect the choice between bus and train. There is, however, some evidence that traveler attitudes affect the awareness and consideration of transit, which will influence the choice set available for mode choice. Research Methods Traveler attitudes were obtained for 18 attitudinal questions from the survey in Charlotte and Chicago and for 15 attitudinal questions in Salt Lake City. Each attitudinal question had five potential responses (strongly disagree, somewhat disagree, neutral, somewhat agree, or strongly agree). In Charlotte and Chicago, traveler attitudes were obtained from all respondents, while the earlier survey in Salt Lake City targeted these questions to specific respondents (six questions were for transit users and nine questions were for non-transit users). As a result of these differences in the surveys, some statistics can be obtained and analyzed from The log-likelihood is a function of the parameters of the mode choice model. The objective of mode choice models is to maximize the log-likelihood; therefore, higher values of log- likelihood are preferred.

18 Characteristics of Premium Transit Services that Affect Choice of Mode all three cities while other analyses can only be performed on survey records from Charlotte and Chicago. For example, more respondents from Salt Lake City indicated that they are willing to increase the frequency of transit usage than did respondents from Charlotte and Chicago. As shown in Figure 5, some 61% of Salt Lake City respondents indicated that they could use transit more frequently. By comparison, respondents from Charlotte and Chicago share similar attitudes toward the possibility of increasing transit usage: In both these cities, 37% of respondents indicated that they could use transit more frequently, which suggests that the potential market share for transit is limited to travelers who feel that public transit is a viable option. Another important element of the surveys was questions about willingness to walk, which is a strong indicator of travelers who may choose to walk to transit services. Respondents were asked about a recent trip that they took. Willingness to walk is not consistent across bus and rail modes or in different cities, but some trends can be observed. Figure 6 shows Chicago and Charlotte respondents’ willingness to walk by mode of travel (auto, bus, and rail) for their current Statement: “If I wanted to, I could use public transit more frequently.” Pe rc en t o f S ur ve y Re sp on de nt s Figure 5. Willingness to increase transit usage for Charlotte, Chicago, and Salt Lake City. Pe rc en t o f S ur ve y Re sp on de nt s Figure 6. Willingness to walk to transit by reference trip mode for Charlotte and Chicago.

Important Non-Traditional Transit Attributes 19 trip. For each level of walking time (up to 5 minutes, up to 10 minutes, and up to 20 minutes), rail travelers are somewhat more likely to report that they are willing to walk to transit than are bus travelers. This outcome suggests that travelers might be more willing to walk farther to rail transit. It is also possible, however, that this outcome indicates that rail users must, on average, walk farther because there is a greater distance between rail stations than most bus stations. Factor analysis of the Chicago and Charlotte survey data was used to determine the most significant attitudinal factors affecting change of mode. Five attitudinal factors were found to be significant in the awareness, consideration, and mode choice models and therefore contributed to explaining travel behavior in these models. There are two challenges to including attitudinal factors in travel forecasting models: 1. The optimal number of factors from a statistical standpoint is too complex for interpretation and therefore less helpful to planners. For example, in this research three factors tended to favor auto modes (pro-car attitude, transit averse, and low transit comfort level) and two factors tended to favor transit modes (pro-transit attitude, and environment, productivity, and time savings). The interpretation of the factors would be much more straightforward if it were limited to the pro-car and pro-transit attitudes. Further analysis of the attitudinal factors demonstrated that these two factors could be supported by the surveys and it may not be necessary to include as many attitudinal statements in the surveys to estimate these factors. 2. Forecasting attitudinal factors requires either a separate model to estimate the attitudinal factors that are input to the various models or a model that can simultaneously estimate traveler attitudes and mode choice or awareness and consideration. In TCRP Project H-37, a simultaneous model to estimate traveler attitudes as a function of socioeconomic variables within mode choice was developed to demonstrate how this can be done. The results of this model indicate which socioeconomic variables are important for each attitudinal factor. In addition, a utility is associated with the bus and rail modes that indicates some differences between these attitudinal factors and mode choice. These research tests can help to guide future inclusion of traveler attitudes in mode choice models. Technical Details Traveler attitudes were developed using factor analysis to correlate traveler sur- vey responses into groups with similar attitudes. The Chicago and Charlotte factor analysis produced five attitudinal factors that were significant in the mode choice models: pro-transit; consciousness (e.g., of environment, productivity, and time savings impacts); pro-car; transit averse; and low transit comfort level. The Salt Lake City factor analysis produced two significant attitudinal factors for transit users (convenience/inclination and service availability) and two attitudinal factors for non-transit users (inclination and discomfort/inaccessibility). The non-transit user factors were not significant in the mode choice model estimation process. The integrated choice and latent variable (ICLV) models provide an opportunity to estimate traveler attitudes as a function of socioeconomic variables within mode choice where the multinomial logit (MNL) models require that traveler attitudes be developed outside the mode choice models. This allows us to forecast these attitudes within the mode choice model rather than having to develop a separate model. For more information, see Appendix G for details on the factor analysis for traveler attitudes and Appendix H for details on the ICLV models for mode choice.

20 Characteristics of Premium Transit Services that Affect Choice of Mode Research Results Table 4 presents the equivalent minutes of in-vehicle travel time for latent variables in the mode choice models. Most of the latent variables reflect large impacts on the choice of transit versus auto, but only few differences between the choice of bus and rail. The few differences are important to understand premium services: • Bus travelers are more informed about transit for commute travel than are train travelers, which may reflect the need to understand a more complex system of bus routes given that outbound and return bus trips may be on different routes due to timing and frequency. • Train travelers in Chicago are more willing than train travelers in Charlotte to walk more than 10 minutes for a train for all trips. These response data are consistent with the prior summary of the survey data shown in Figure 6. • Travelers in Chicago are more likely than travelers in Charlotte to be willing to walk more than 2 minutes for commute trips on a train than on a bus. • In Charlotte, travelers with pro-transit and pro-environment attitudes are slightly more likely than travelers in Chicago to choose train over bus, and travelers with pro-car attitudes (including travelers who are transit averse and/or have a low transit comfort level) are slightly less likely to choose train over bus. Summary of Key Findings There are a number of benefits to accounting for non-traditional factors and recognizing traveler attitudes or awareness and consideration in mode choice. Non-traditional service attributes, such as on-board and station amenities, are important differentiators for premium transit. Premium service attributes account for a range of 13 to 29 minutes of in-vehicle travel time based on MaxDiff scaling models. Commute Non-Commute Explanatory Variables Bus Train Bus Train Chicago Very Informed About transit 8.84 Pro-Transit Attitude 38.2 38.2 33.32 33.32 Environment, Productivity, and Time Savings 15.16 15.16 11.89 11.89 Pro-Car Attitude -24.76 -24.76 -24.53 -24.53 Transit Averse -5.44 -5.44 -9.42 -9.42 Low Transit Comfort Level 5.32 5.32 Not Willing to Walk More than 2 minutes -27.52 -27.52 -41.11 -41.11 Willing to Walk 10 or more minutes 7.08 8.68 Charlotte Very Informed About Transit 21.91 12.91 29.16 29.16 Pro-Transit Attitude 14.5 14.5 22.37 23.11 Environment, Productivity, and Time Savings 15.55 15.55 32.68 34.11 Pro-Car Attitude -21.82 -21.82 -22.47 -23.32 Transit Averse -2 -2 -7.58 -7.95 Low Transit Comfort Level -14.86 -14.86 -25 -26.11 Not Willing to Walk More than 2 minutes -4.59 -11.55 Willing to Walk 10 or More Minutes 7.68 7.68 24.63 24.63 Note: Auto modes are not included here because their equivalent minutes of travel time for these variables are zero. The cases where bus and train coefficients did not reflect significant differences were estimated together. Table 4. Equivalent in-vehicle travel time (in minutes) for traveler latent variables in mode choice models. Latent variables are those that cannot be directly observed. In this study, examples of latent variables include traveler attitudes, willing- ness to walk, and how informed travelers are.

Important Non-Traditional Transit Attributes 21 When comparing modal availability predicted by network path-building models, travelers are more likely to report rail service being available than bus service. This may be because bus systems are more complex than train systems and bus stops are less visible than train stations. Consideration of transit options does affect sub-modal choices, with 12% to 14% of travelers with rail available reporting that rail was not considered for the trip and 27% to 38% of travelers with bus available not considering bus for the trip. Awareness and consideration models were esti- mated and used to constrain mode choice sets, which does statistically improve goodness-of-fit for mode choice model estimation, but the impact on forecasted ridership by mode is unknown. Traveler attitudes do influence the choice of transit or auto, but do not consistently affect the choice of bus or train for different types of trips or in different cities.