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.

A-1 A p p e n d i x A Literature and Practice Reviews Contents A-1 Overview A-2 Awareness of Transit Services A-7 Transit Service Attributes A-27 Case Studies of Transit Attribute Evaluations A-38 Applied Models A-52 Transit Agency Interviews A-62 Summary Overview The review of the literature and current practice covered three aspects of transit planning: awareness of transit services, transit service attributes, and how mode choice models incorporate premium transit services. The lack of awareness and familiarity with transit seems to be significant and there is not yet abundant research on this topic. Awareness about premium transit services can increase due to visibility of stations or right-of-way over more conventional services. Branding and marketing campaigns can also increase awareness, but this was not the focus of our research. Current mode choice models assume perfect knowledge with regard to awareness and consideration of transit modal alternatives and the research clearly shows this not to be true. This review led to a clear focus on modeling awareness and consideration for transit services and using these models to constrain choices available within the mode choice model. To support better behavioral models, it is necessary to extend the conventional set of explanatory variables to include new variables and methods that relate specifically to the decision-making process. Current practice in mode choice modeling typically results in models that are sensitive to the effects of travel times, wait times, frequencies, travel costs, and transfers, and include large mode-specific constants. In theory, the mode 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 majority of the literature and practice review focused on evaluating non-traditional transit service attributes that could inform mode choice models and transit networks for planning analysis. The long list of attributes was organized into nine categories: monetary cost, journey time, convenience, comfort, accessibility, productivity, information services, fare payment, and safety. These attributes were refined for use in market research conducted in three cities for this study in three groups: on-board amenities (comfort, productivity, information services, and

A-2 Characteristics of premium Transit Services that Affect Choice of Mode Practitioners have struggled to quantify these additional service attributes and to measure traveler’s reactions to these service attributes. This review highlighted the need for an in-depth study to quantify these additional service attributes and to incorporate them in travel demand forecasting models. Contents of the Appendix This appendix presents the findings from a review of the literature and the practical experience in these areas, focusing primarily on identification of distinguishing transit service features and their relative importance in mode choice and transit customer satisfaction. A few successful transit industry anecdotes related to upgrading “non-traditional” transit service amenities are discussed to provide context for the research. The discussion is based on detailed responses obtained from staff at a few transit agencies and MPOs which have been reported in Appendix C. Further, the appendix outlines current attempts in research and practice to understand mode choice and improve the reasonableness and interpretability of mode choice models, reducing the extent to which mode constants dominate the utility equations. The review considers the extent to which the general public is aware of transit services, and whether the presumption of complete knowledge in travel models is reasonable. Finally, the ways that non- traditional transit attributes have been included in mode choice models are discussed. These reviews together informed and helped focus the data collection effort for TCRP Project H-37 and begin to suggest opportunities for advancement of the practice. This appendix presents a more detailed review of research around identification and quantification of non-traditional transit service attributes as well as case studies pertaining to attribute evaluation and incorporation of these attributes in model applications. Awareness of Transit Services Mode choice models typically assume that if a transit option is available nearby, it is part of the traveler’s mode choice set and has a probability of being used (i.e., models assume “perfect knowledge” with regard to the mode choice set). In reality, travelers are often simply unaware of the transit options available to them. Therefore, when calculating the market share for transit modes, accurately defining the mode choice set and eliminating any zero-probability options will produce more accurate ridership forecasts. Moreover, to the extent that travelers are more aware of different modes, it is possible that this difference in awareness explains some of the observed difference in ridership not explained by level-of-service. This has potential implications for regions considering and analyzing new transit modes for which the regional model is not calibrated. Transit Awareness and Familiarity The lack of awareness and familiarity with transit seems to be significant, though there is not yet abundant research on this topic. For TCRP Report 63: Enhancing the Visibility and Image safety), station amenities (comfort, accessibility, information services, fare payment, and safety), and other attributes (journey time, convenience, and information services). Traditionally, travel demand models have underestimated ridership on premium services. People have speculated that this is related to public perception of safety awareness, brand visibility, and various service attributes that are not typically included in mode choice models.

Literature and practice Reviews A-3 this study lived in an area with readily available transit alternatives, 21% did not know that transit was available (FIGURE A-1). More than twice that number, 44%, reported being either “not very familiar” or “not at all familiar” with public transportation services in their area. Unfamiliarity with public transportation is also prominent in major transit markets. A study for the Regional Transportation Authority for Chicago (Northwest Research Group, Inc., 1999) found that 38% of randomly selected residents in the transit service area had not ridden transit in the past year, with 19% reporting they were “somewhat unfamiliar” with transit services and an additional 36% “very unfamiliar” with transit. Source: (TCRP Report 63, 2000) FIGURE A-1. Awareness of transit availability. of Transit in the United States and Canada (2000), individuals in a variety of transit markets were asked their perception of transit availability; and while all respondents contacted in An interesting social experiment was conducted at UCLA in the summer of 2008 to get employees to try public transit (Gould, 2010). As gas prices were increasing dramatically during the summer of 2008, UCLA tried to motivate SOV commuters to switch to transit by providing a free transit pass for 12 weeks in return for turning in their employee parking pass. Researchers found that there were several factors that influenced employees’ successful conversion to public transit and several other factors that caused some employees (28%) to return to driving upon completion of the program. It was felt that gaining familiarity with using transit over the 12-week period contributed to the program’s success because committed transit riders were able to become comfortable with routes and schedules, became more relaxed with the experience, and ultimately found the bus less stressful than driving. Conversely, participants who ultimately went back to driving and reclaimed their parking passes, indicated that they did, in fact, continue to drive a few times per week throughout the program duration.

A-4 Characteristics of premium Transit Services that Affect Choice of Mode While many individuals are unaware of transit in general, determining the differences in awareness between premium and conventional services is of particular importance to this research. Typically, those supporting a positive premium service bias cite the improved quality of non-traditional, more qualitative attributes like comfort and convenience. However, another possible reason for premium transit’s perceived appeal is that premium transit services are more visible and therefore travelers are more aware of their existence. For example, newer Bus Rapid Transit (BRT) systems often stand out due to branding, and their right-of-way may be painted or visibly marked in some way. Rail stations tend to be highly visible—as are the tracks and rail cars—and major infrastructure investments receive more publicity owing to cost and the need for infrastructure improvements. The characteristics and marketing of premium services that heighten visibility may resolve some of the currently unexplainable preference for premium transit. The most obvious way to become aware of a transit service is to physically see it. Conventional bus service may seem visible because it is typically well established and geographically widespread; however, bus stops are often poorly marked and the route and schedule of the service can be difficult to determine (FIGURE A-2). Premium bus services, on the other hand, typically include many improvements that increase the visibility of the service. Improvements in bus stops such as clear signage, seats and shelters, or off-board ticket vending, bring attention to the service, while branding on the bus exterior captures attention and distinguishes the bus from conventional services. Also, premium bus services occasionally operate in bus lanes or high occupancy vehicle (HOV) lanes, and marked or painted lanes can bring attention to the bus service and its potentially improved reliability and travel time. In New York City, a BRT service introduced in 2007 incorporates many of these visible service improvements and has shown a significant increase in ridership (J. Barr, pers. comm., September 17, 2008). The new BRT alternative, branded Select Bus Service (SBS) by the New York City Metropo- litan Transportation Authority (MTA), runs along Fordham Road in the Bronx. Bus lanes are painted a separate color from the regular street with large signs declaring the lanes as bus lanes (FIGURE A-3). New bus shelters have been constructed to offer better visibility and improve security, and ticket- vending machines have been placed at bus stops and eliminate on-board payment (FIGURE A-4). The SBS buses, which are FIGURE A-2. Conventional bus stop. FIGURE A-4. Select bus service stop in Bronx, NY. FIGURE A-3. Fordham bus lane, Bronx, NY.

Literature and practice Reviews A-5 the same type as conventional buses, are thoroughly rehabilitated and cleaned for the new service, and are equipped with signal priority and on-board cameras. The buses are also “wrapped” with a brand logo. Awareness of the Fordham Road service and its high priority has increased, likely due to the “hard-to-miss” painted lanes and that some on-street parking was taken away (which makes non-riders aware that something has changed). Ridership is up in the corridor, with a 20% increase in ridership on the SBS over the former limited stop route (a much higher increase than what MTA buses have experienced), though it is not yet clear how much of this increase comes from local bus riders who switched to SBS. Like the SBS, the Xpress bus service operated by the Georgia Regional Transit Authority has expanded its number of routes from 2 to 27 in the last five years with little to no advertising (R. Alexander, pers. comm., September 26, 2008). The buses serve as “billboards,” with branding and website/phone information prominently displayed on the sides. Further details on these projects that have enhanced the visibility of transit services and as a result influenced the ridership can be found in Appendix C. Light rail and commuter rail, while not as geographically dispersed as conventional bus, are highly visible and the routes can be simpler to understand. Suburban rail stops often have parking lots with clear signage and larger stations, and the tracks and rail cars are typically easily noticed. The fixed track and sequence of stations also indicate the service route. New premium transit services may also be more visible upon opening because the introduction of new capital improvements is likely to be discussed on television, in newspapers, and among travelers, increasing awareness. For example, a study of the FasTracks rail and BRT service improvements in the Denver area (The Kenney Group, 2007) showed that, “three years removed from the FasTracks campaign and the campaign communications, half of survey respondents feel informed about FasTracks plans.” Respondents reporting the primary source of this information cited newspaper (55%) and television news (31%) as opposed to the Regional Transportation District’s website (3%) or its newsletters or emails (1% each). The introduction of premium transit is also often accompanied by targeted marketing campaigns. Marketing has been shown to significantly improve ridership, due to increased awareness of premium transit services. Brog and colleagues (Barta et al. 2007) in particular have detailed the impact of targeted marketing—specifically the IndiMark® program—on ridership increases. Individualized Marketing (IndiMark) is a dialogue-based technique for promoting the use of public transportation based on a targeted, personalized, customized marketing approach. The success of this technique has shown the importance of distributing information to heighten awareness and increase ridership. Results from two projects show that soft policies such as IndiMark can in fact double ridership (Barta et al. 2007). The introduction of the “Saarbahn,” a light rail system in the Saarland region in Germany, led to an increase of 28 public transportation trips per person per year, but in combination with IndiMark, the increase was 56 trips per person per year. In Portland, Oregon, the introduction of the MAX light rail line increased transit use by 16 trips per person per year. Those targeted by IndiMark increased transit use by 32 trips per year.

A-6 Characteristics of premium Transit Services that Affect Choice of Mode One additional example, the River Line light rail operated by New Jersey TRANSIT, illustrates how the many visible aspects of a new premium rail service can work together to increase ridership (T. Marchwinski, pers. comm., September 30, 2008; see later in this appendix for more details). This service, introduced in March 2004, connects southern New Jersey to Camden and Trenton and to the Northeast Corridor line to Philadelphia via Port Authority Transit Corporation (PATCO) service. Many specific infrastructure improvements contributed to the increased awareness of the River Line. Train stations were built with ticket-vending machines, phones, a public announcement system, digital signs to show delays and alerts, platforms, and full signage. The service was accompanied by many new park-and-ride facilities as well as an upgraded transfer point with PATCO, offering a pleasant pedestrian-friendly plaza for transferring between services. Public debate over cost and right-of-way improvements heightened awareness, and once the right-of-way was constructed, the 45 grade crossings were easily noticed by auto travelers and even led to safety trainings conducted in schools, introducing children to the River Line. Strategic marketing campaigns were also launched to increase awareness of the River Line (FIGURE A-5). Initially, the service was priced lower than the existing bus despite the many benefits provided by the premium service. Also, upon introduction and for the first year, the service was advertised through newspapers, brochures, and connecting services like the Northeast Corridor. Websites, which still continue (http://www.riverline.com), were developed and offer promotions to destinations like the aquarium or entertainment centers for concerts, increasing weekend trips in both the short and long term. While the exact impact of the project’s visibility on ridership has not been measured, quantitative evidence suggests that visibility played a key role. Surveys showed that after service opened, 15% of people were riding just to “check it out.” Meanwhile, overall ridership increased significantly, with 25% of River Line riders having switched from the existing bus, 50% switching from auto modes, 5% using the service to transfer from the Northeast Corridor to PATCO, and the remaining ridership coming from induced new trips. The existence and extent of marketing likely contribute to heightened awareness and increased ridership and can be studied alongside other attributes of premium services that heighten visibility to determine the level of awareness and, more specifically, who is aware and under what circumstances. Model adjustments accounting for differences in awareness can be developed and applied so that the market potential for a particular mode can be more accurately estimated. FIGURE A-5. Large crowds during the opening ceremony for the River Line.

Literature and practice Reviews A-7 Transit Service Attributes The attributes explaining mode choice must be identified and appropriately described in order to estimate each mode’s market share. Mode choice models typically specify level-of- service attributes such as travel time, cost, access time, wait time or headway, and transfers. These attributes are considered strong predictors of mode choice, and they are also readily quantifiable, making it easier to measure their importance to travelers and to incorporate them in travel models. The downside, however, is that these level-of-service attributes only account for a portion of the variation in mode choice behavior and therefore can be poor mode choice predictors. These level-of-service attributes alone do not adequately differentiate among transit options, and models are left to capture the remaining attributes in mode-specific constants. In reality, there are several attributes known to be important in mode choice decisions and transit customer satisfaction that are not traditionally included in mode choice models. These non-traditional attributes tend to either be qualitative (e.g., comfort and safety) or quantitative but difficult to measure (e.g., reliability). Practitioners assume that all the unspecified characteristics of a particular mode are captured in the model’s error term and that adding mode-specific constants can then reflect the magnitude of the difference in preference among modes beyond the specified attributes. Use of mode constants without representing other factors in the models is, however, not generally a sufficient way to represent differences among modes. One reason for the insufficiency of mode constants is that any error introduced in prior stages of the forecasting process will be incorporated in the constant. A problem lies with the transferability of mode constants when new premium transit services are introduced with varied levels of services and amenities. A second problem relates to the Federal Transit Administration (FTA) New Starts project evaluation criteria related to ridership, known as user benefits. User benefits is a measure of the difference in the aggregate utility of different alternatives, and heavy reliance on mode- specific constants has been shown to bias this measure. There are recent changes to the project evaluation criteria regarding user benefits, so a review of these criteria should be completed before use. One way to improve mode choice models is to identify the most important non-traditional attributes that contribute to the preference for premium transit, quantify the importance of these attributes, and realistically incorporate them into travel models. This is a difficult task because so many of the service attributes that distinguish premium transit from traditional transit services are qualitative factors. However, while quantifying these non-traditional attributes is challenging, there is much literature identifying these factors, and some attempts have been made to quantify their relative importance and their effects on the traveling experience. Finally, while few attempts have been made in practice to incorporate a variety of non-traditional attributes in models, real-world efforts to adjust mode constants and to specify a modest number of non- traditional attributes have been made, and the results of these efforts can offer key insights into predicting mode choice.

A-8 Characteristics of premium Transit Services that Affect Choice of Mode Traditional Transit Service Attributes Most mode choice models characterize service quality in terms of travel time (in vehicles, walking, waiting, and transferring) and cost (fares, fees at park/ride lots), and some models capture the effects of a few other measured attributes (number of transfers, transit/pedestrian friendliness at the beginning and/or end of the trip). Walk time and wait time are usually specified separately from in-vehicle travel time (IVTT) because time spent out of the vehicle has typically been found to be two to five times more onerous than IVTT (Litman 2007). Time spent walking and waiting during a transfer is also accounted for separately, but transfers are generally thought to impose additional costs through increased unreliability, additional mental effort, and by splitting IVTT into a greater number of stages, which breaks up time that could be more productive with fewer but longer journey stages (Li 2003). These costs can be captured by adding a coefficient specifying the number of transfers and assessing a transfer penalty, estimated as an extra 5 to 15 minutes of IVTT (Horowitz and Zlosel 1981). The monetary cost of a transfer is captured in the cost coefficient along with the fare, parking cost, and any additional fees. Service frequency can be included in models as a proxy for wait time; however, research has shown that improvements in headway provide greater benefits for high-frequency services than low-frequency ones, and can therefore be specified nonlinearly. In one study, a 1-minute decrease in headway for a service departing every 5 minutes was equivalent to 1 minute of IVTT savings, while the same improvement for an hourly service provided roughly half that benefit (Litman 2007). Finally, pedestrian friendliness, while not necessarily a service attribute over which the transit agency has control, is occasionally included in models to account for variation in the quality of the accessibility between the station and activity locations. Non-Traditional Transit Service Attributes Mode choice models account for the different costs associated with the different stages of a transit journey, but the costs of each stage can still vary considerably based on the conditions in which they take place. Wait time or IVTT spent in dirty, crowded, or unsafe conditions can make a trip seem more onerous, whereas the ability to be productive and enjoy a smooth ride in a comfortable seat can make traveling significantly more enjoyable. Factors beyond the journey time, such as the ease of planning or executing a trip, also impact the overall trip cost and therefore the attractiveness of a particular mode. TABLE A-1 presents a list of attributes that can impact the cost of travel beyond those traditionally applied in models. To provide structure, the attributes have been organized into nine categories: monetary cost, journey time, convenience, comfort, accessibility, productivity, information services, fare payment, and safety. The attributes listed were identified in quantitative and qualitative research studies as well as customer satisfaction surveys conducted for the Chicago Transit Authority (CTA), Washington Metropolitan Area Transit Authority (WMATA), and the Sacramento Regional Transit District. The full list of attributes measured in these customer satisfaction surveys along with some anecdotal accounts and quantitative valuations of a subset of these attributes can be found later in Appendix A.

Literature and practice Reviews A-9 TABLE A-1. Transit attributes. Monetary Cost Producvity Cost of one-way ride/pass Ability for ac vity Parking cost Ac vity services - WiFi Journey Time Entertainment Access/Egress me Journey enjoyment Wait me Informaon Services In-vehicle me General Understandability of schedules/routes Reliability Accuracy of informaon Right of way Ease of geng informa on by phone/online Bus goes to front of line at red light Effec veness of customer service Bus gets priority at traffic light Availability of service change informa on Convenience Noficaon of service changes Transfers Number of transfers Availability of customized local informaon Transfer walk me Staon/stop Schedule/map availability Transfer wait me Availability of real-me informaon Transfer monetary cost Usefulness of digital displays Time to transfer before assessed second fare Clear/ mely announcements Quality of transfer (same vs different plaorm) Visibility of signage Transfer informa on Staff availability Schedule/route coordinaon w/in b/w agencies Staon egress informaon Span/Frequency Service frequency ] On-board Visibility of route names/numbers on outside Service hours Schedule/map availability Geographic coverage Clear/mely announcements on board (if any) Express service Visibility of staon name from inside train Comfort Driver knowledgeable of schedules/routes Staon/Stop Shelter Driver explains reasons for delays Seats/benches Fare Payment Cleanliness Pass/fare card purchase loca on availability Vandaliza on Ticket vending machine availability Maintenance/repair Ease of purchasing pass/fare card Staon design/layout Ease of recharging fare card Staon building materials Ease of obtaining refund/replacement fare card Staon art Fare integraon with other agencies On-board Layout/design Mandatory off-board payment Seat configuraon Proof of purchase by fare inspectors Seat comfort Ease of paying fare on-board Load factor Change availability Seat availability Safety Heang/cooling/venlaon Staon/stop crime daylight Smoothness Staon/stop crime nighme Quietness On-board crime daylight Cleanliness/appearance interior/exterior On-board crime nighme Smell Parking lot crime daylight Space for luggage/belongings Parking lot crime nighme Restrooms Presence of surveillance cameras Accessibility Presence of emergency call buons Pedestrian friendliness Presence of security personnel and/or police Parking General visibility/open sightlines Bicycle accommodaon Lighng Distance from entrance to pla orm Accidents Elevators/escalators Availability of on-board emergency exits Wider passages and stairways Pla orm surface Low-floor/no steps Wide entry Availability of handrails Stopping posion of bus/train

A-10 Characteristics of premium Transit Services that Affect Choice of Mode Qualitative and Quantitative Research on Premium Service Attributes The following sections of this appendix present anecdotal accounts and quantitative valuations for a subset of the attributes listed in TABLE A-2. The sections are organized by category, and the attributes described are those most frequently cited in research, those most valued by travelers, and those that transit agencies have some degree of control over. Note, however, that after reviewing the existing literature, it is apparent that no standard method exists for quantifying, or valuing, non-traditional attributes. Because of the many different techniques and methods used, it is difficult to compare attribute valuations across studies. There are differences in the presentation of attributes (pictorial, text) that can impact the respondent’s valuation during survey work, particularly regarding qualitative attributes (e.g., a picture of a crowded bus may have more emotional resonance than the text “crowded bus”). There are also differences in the specification of attributes that confound any specific comparisons across studies (e.g., the following might be used in different studies to describe the ride quality (1) smooth, (2) quiet, (3) smooth and quiet, (4) smooth, quiet, and clean). There are differences in the analytical techniques employed (e.g., stated preference, maximum differences, regression) that can lead to different results due to fundamentally different approaches. And there are differences in the benchmarking of attribute importance (e.g., indexed, equivalence in percent change in IVTT, equivalence in flat IVTT), where the final valuation can be expressed somewhere on a scale of 0–100, or as a benefit equal to a 10% decrease in IVTT, or as a benefit equal to 10 minutes of IVTT. One study in particular (Douglas and Karpouzis 2006) is frequently cited in this appendix because it is a comprehensive study of both on-board and station attributes. This study, however, presents results in terms of the increased on-board time respondents would be willing to accept for a 10% improvement in customer satisfaction for a specific attribute. While this study is unique in that customer satisfaction ratings are used to derive attribute importance ratings, the study ultimately examines and presents data in a way that is difficult to compare with other studies. Other differences exist between studies that further complicate comparisons. There are differences in the geography (Australia, Britain, United States) that can impact the value of a bus shelter or heating and cooling. There are differences in demography (middle class, lower class, older, younger) that can affect the value of attributes such as real-time information or a no-step bus entry. There are differences in the existing physical conditions (safe, high-crime, modernized, aged) that can impact the value of security cameras or the appearance of the building. These differences can significantly affect the relative value placed on attributes. As attributes are presented in the following sections, the absolute values of the attributes are often provided (e.g., a dedicated right-of-way for BRT offers the same benefit as a 10-minute reduction in travel time). However, as discussed, it is difficult to compare these values across studies. Most often, the best comparisons that can be drawn relate to the relative importance of the attribute among other attributes evaluated in the same study. For example, if a 10-minute time savings is the most important attribute evaluated, a dedicated right-of-way is also very important.

Literature and practice Reviews A-11 TABLE A-2. Relative attribute importance. Reference Documents D ou gl as e t a l 20 06 Li tm an 20 07 H en sh er e t a l 20 03 R SG JF K 20 06 Sw an so n e t a l 19 97 Pe pp er e t a l 20 03 Sp itz e t a l 20 07 R SG N JT 20 07 Journey Time Reliability Right of way Moving to front at signal Signal priority Convenience Headway Staon Comfort Cleanliness Staon building Shelter Seang High quality materials Art On board Comfort Heang & cooling Layout & design Seat comfort Quietness Cleanliness Smoothness Seat configuraon Luggage rack Mullevel Seat material Crowding Seat capacity Seat availability Crowded seat Stand 20+ minutes Crush stand 20+ minutes Load factor 160% Accessibility Entry steps Bus pulls to curb Stop w/in walking Good sidewalks Elevators/escalators Wider passages/stairs Informaon Services Real me informaon General maps/metables Local maps/metables Announcements Fare Payment Pre boarding POP Tickeng Safety Cameras/emergency call Security day Security night Lighng Visibility Indicates highest relave importance within the study Indicates lowest relave importance within the study

A-12 Characteristics of premium Transit Services that Affect Choice of Mode Journey Time Travel time is an attribute traditionally accounted for in mode choice models and by path- builders. However, there are aspects that are not well represented but seem to be important, including reliability and specific design elements that accommodate and give priority to transit vehicles. Reliability. An unreliable transit service imposes obvious costs to travelers. Whether the service has an unreliable arrival time at the boarding station or an unreliable IVTT, the traveler faces an uncertain arrival time at the final destination. If the unreliability is anticipated, a traveler can depart earlier and allow for uncertainty in the total trip time, but this also imposes a cost. The costs of arriving late due to unreliability are evident and can range from significant to incidental depending on the journey purpose. One study conducted in Australia (Hensher et al. 2003) used a stated preference exercise to quantify the impact of various bus stop and on-board service attributes, including reliability. While choosing among service options, respondents indicated the cost of a bus being 5 or 10 minutes late, compared with an on-time service. Results showed that the cost from an additional minute of delay was equivalent to 2.1 additional minutes of IVTT. However, in many areas, the variation in delay is much larger than 10 minutes, and it is likely that delays exceeding 10 minutes would impose relatively higher costs (i.e., the cost of unreliability is nonlinearly related to the amount of delay). In fact, other research has estimated an additional minute of unexpected delay at 3.7 times the cost of an additional minute of IVTT (Litman 2007). Perhaps the most realistic way to present unreliability is the same way one would respond to the question, “How long does it take it to get there?” In an area with unreliable travel times, the answer would likely be that, “Most times it takes as little as x minutes, but it could take as much as y minutes.” However, the challenge is presenting unreliability realistically—as the combination of the probability of delay and the amount of delay—while still making it tangible for respondents. Unreliability, however, was successfully measured in this format and incorporated in an air passenger transit access model developed and implemented for travelers accessing John F. Kennedy International Airport. In this study, the measure of unreliability was presented in a stated preference survey as, for example, “1 in 10 trips, the service is 15 minutes late,” and produced statistically significant coefficients (RSG 2006a). The value of the estimated coefficients for this measure of unreliability ranged from 0.5 to 1 minute of equivalent IVTT per minute of delay incurred 10% of the time, depending on the market segment. To implement this Because the attributes are organized and presented using nine categories and the attributes tested within a study often span many categories, the value of an attribute relative to the others from the same study can be obscured. Therefore, this section provides brief summaries of each valuation study and the quantitative results from that study. TABLE A-2 also presents the relative value of the attributes within each study, and can be used to compare the relative value of an attribute across all research. Presumably, the attributes most frequently receiving high-importance scores are the most important. Finally, where possible, any differences in the relative value of an attribute are explained contextually as the attributes are introduced. For instance, if safety attri- butes were highly valued in one study and not another, perhaps it is simply a result of the fact that one area has a greater threat of crime.

Literature and practice Reviews A-13 variable in the regional model, Global Positioning System speed data were used to relate average peak-period delay to the standard deviation in travel time. Higher average delay resulted in a higher likelihood of extreme delay, and this deviation in delay was used to estimate the 90th percentile travel time and compute the resulting delay. Route Accommodations. Transit level-of-service can be improved by operating transit vehicles in a dedicated right-of-way, or in HOV lanes, as well as by cheaper methods such as accommodating queue-jumping and allowing signal pre-emption (FIGURE A-6). In 2007, RSG used a maximum differences scaling conjoint (MaxDiff, or best-worst conjoint) analysis to value BRT attributes in a heavily congested corridor in New Jersey (RSG, 2007). Here, a dedicated right–of-way was found to be particularly highly valued, which is not surprising since the primary benefits of a right-of-way—improved reliability and travel time—are greatest where traffic is heaviest. Also, while respondents did not specifically answer how much time they thought a right-of-way might save them, the MaxDiff results showed that a right-of-way provided the same benefit as a 10-minute reduction in travel time. If it is assumed that the majority of benefits from the interpreted that respondents, on average, estimated that this attribute would cut approximately 10 minutes from their trip. Other “route accommodation” attributes evaluated in this study included “moving to the front of the line at red lights” and “getting priority when coming to traffic lights.” Both these attributes provided benefits of less than a 5-minute reduction in travel time, indicating that respondents, given their awareness of travel conditions, believed that these options would provide lower time savings. Currie (2006) discusses various features of bus vs. rail modes in the context of transit- oriented development (TOD). He assesses the relative merits of bus and rail and notes the importance of examining the bus vs. rail issues in isolation from market climate and development viewpoint to understand modal-only influences. While the article primarily examines the strengths and challenges of bus vs. rail factors in relation to successful TOD, the research examines several features that are relevant to mode choice: newness, permanence, and bus stigmatization. One advantage that light rail holds over local bus and somewhat surprisingly bus rapid transit is the perception of newness. Light rail routes often replace old bus routes and are introduced as an important new mode while bus rapid transit routes replace buses with buses. Taken as a factor in the success of TOD, and by inference mode choice, rail wins out over bus to some degree. With respect to bus stigmatization, and relatively speaking of course, “buses have a bad image”. Transit operators are trying to change this image but buses are often still perceived as being second-class forms of transportation. In his research, Currie challenges this perception and compared how transit riders felt about on-street bus, dedicated bus rapid transit, light rail, and heavy rail. He found a preferential bias for rail over on-street bus with the benefit valued at FIGURE A-6. Atlanta XBus using HOV ramp. right-of-way are improvements in travel time, it can be

A-14 Characteristics of premium Transit Services that Affect Choice of Mode between 4 to 10 minutes of travel time. He also found similar results in preference and benefits for fixed-guideway bus rapid transit over on-street bus, suggesting that the technology may be less important than the reliability and service quality provided by a dedicated guideway. In practice, several successful BRT services have implemented one or more of these route accommodation improvements, including the Kansas City MAX, which, “uses dedicated lanes during rush hour and has the ability to prolong green lights at intersections to remain on schedule” (Kansas City Area Transportation Authority 2008). Convenience. The relative convenience of transportation options is often cited as a critical element in mode choice decisions. Mode choice models traditionally represent convenience using transfer and headway variables. However, the quality and ease of the transfer is often not represented, and the span of transit service is a detail that often challenges travel forecasters. Transferring. As described earlier, transfers are specified in many mode choice models because of their impact on the convenience of a transit trip. Transfer walk time, wait time, and monetary cost are captured in traditional attributes and any additional burden of transferring can be accounted for in a transfer penalty; however, it is likely that the magnitude of the transfer penalty can vary. For example, after accounting for the walk, wait, and monetary cost, cross- platform rail transfers may be still inherently more convenient and simpler than transfers involving a bus. Likewise, a second or third transfer is presumably more onerous than the first. Also, the amount of time allowed before a new transfer or ticket must be purchased as well as the coordination of schedules, and routes within and between transit agencies may impact the magnitude of the transfer penalty. The Utah Transit Authority (UTA) recently began operating the FrontRunner commuter parallel to the I-15 corridor at 30-minute headways during the day with a distance-based fare that costs as much as $6.50 one-way without a discount pass. Initial ridership in 2008 is approximately 5,000 riders on the average weekday. One particularly unique aspect of the FrontRunner service is that an extremely high share of riders using the service transfer to reach their final destination (M. Crandall, personal communication, September 2008). Anecdotal evidence suggests that as many as 90% of the FrontRunner riders heading to downtown Salt Lake City transfer to other transit lines. The exact transfer rate has not yet been estimated, but the simple fact is that the FrontRunner was designed to terminate at a new intermodal center on Salt Lake City’s industrial and redeveloping west side, approximately ¾ mile from the central business district (CBD) and the majority of downtown. The UTA designed the intermodal center and service routing to ensure convenient and easy transferring between the FrontRunner and light rail and bus services. This example illustrates that well-designed transfers can lead to a successful integration of transit services. FIGURE A-7. UTA's FrontRunner. rail service between Ogden and Salt Lake City (FIGURE A-7). This new service operates

Literature and practice Reviews A-15 Span of Service. The span of transit service for a particular transit line obviously limits or defines the availability of that service. This service attribute tends to be tracked in transit customer satisfaction surveys and is clearly important. Representing the span of service in a model is sometimes complicated because the span of service in reality is not uniform over the day or within a given time period. Travel models tend to be aggregated along several dimensions, including time of day. This time-of-day aggregation has implications for transit network modeling related to service availability and headway. For instance, if a traveler makes a trip at midnight but the only available transit service ends at 11 p.m., the transit service is technically not an option. Models, however, tend to be aggregated with respect to time period, and trips at 11 p.m. and midnight may both fall within the “night” time period, during which some transit service is available but the specific availability can only be approximated. Adding a span-of-service attribute better captures service availability that is lost in time period aggregation. Variation in headway, in addition to service availability, is also obscured by time period aggregation. During 1 hour of a 3-hour morning peak period, express bus headways may be 10 minutes, while outside of that 1 hour, the headways could be much higher. In this case, a modeler is faced with the difficulty of averaging the headway during the entire morning peak while the span-of-service attribute can better account for headway costs and benefits facing each traveler in each time period. Station/Stop Comfort A transit trip made in comfortable conditions is more appealing and imposes lower costs on the traveler. At transit stations and stops, cleanliness, seating, shelter, and the layout and appearance of the station building all impact the perceived cost of waiting. Cleanliness. A clean and well-maintained station or stop has been found to be particularly highly valued in studies of both rail and bus services. In a British research study, “the difference between the dirty, vandalized stop and the clean, well-maintained one was the largest magnitude valuation,” among 30 bus stops and on-board attributes (Swanson et al. 1997). Improved cleaning and maintenance were the second most important attribute among New York City (NYC) train station improvements (next to the provision of surveillance cameras and emergency call buttons) (Spitz et al. 2007), and cleanliness was also near the top of the list of station improvements in the Douglas research (fourth out of 24 station attributes). TABLE A-3 shows the increased amount of time travelers would willingly accept for a 10% improvement (in customer satisfaction ratings) of various station-comfort attributes measured in the Douglas and Karpouzis (2006) research. TABLE A-3. Value of station improvements. Type of Station Improvement Additional Time* (Increased On-board Time) Cleanliness 19% Station Building 17% Weather Protection 07% Platform Seating 05% Platform Surface 05% *Value of improving station attribute rating from 50% to 60% Source: Douglas and Karpouzis, 2006

A-16 Characteristics of premium Transit Services that Affect Choice of Mode Shelter and Seating. While not as highly valued as station and stop cleanliness, shelter and seating are routinely identified and quantified in research. In the Douglas and Karpouzis (2006) work, these attributes were relatively less valued, although that research examined train stations where shelter and seating are more likely to be provided than at bus stops. Two bus studies showed higher value for these attributes, with a seat and shelter providing benefits equivalent to a 43% reduction in IVTT in the Hensher et al. (2003) study. Station Building. Research into the layout and appearance of the station building showed different results depending on the specificity with which the attributes were described. For instance, in the Douglas and Karpouzis (2006) study, the general attribute “design and layout of Main station building” was the fifth most important station attribute (of 24), with a 10% improvement in customer satisfaction ratings providing benefits equal to a 17% reduction in travel time. However, when station appearance attributes were specifically described, as in the NYC station amenities study, “using high-quality/attractive materials such as granite” and “station art such as mosaics and stained glass” were the least valued of 12 station improvements (Spitz et al. 2007). This would suggest that functional improvements to station building design are preferred to aesthetic improvements. On-board Comfort While on-board a transit vehicle, cleanliness, crowding, and the vehicle layout also impact the perceived cost of traveling. Layout and Seating. Once on-board, the perceived cost of in-vehicle time is greatly affected by the level of comfort. The interior layout and seating attributes are often identified as important contributors to transit attractiveness, though the relative values of these attributes have been found to differ across research studies. In the Douglas and Karpouzis (2006) study, the general attribute “layout and design” was particularly highly valued. Along with air conditioning, this attribute offered the most benefit of any on-board improvement (29% increase in IVTT for a 10% improvement). However, any specific improvements resulting from specific train design elements cannot be discerned from this study. A variety of specific on-board attributes were valued using an adaptive conjoint analysis in a study conducted for New Jersey TRANSIT (Pepper et al. 2003). This study was conducted after New Jersey TRANSIT considered the introduction of multilevel coaches to increase seating capacity on trains traveling in the New York Metropolitan Area. One specific aspect of train layout, seat configuration, was tested, with respondents indicating only a slight preference for walkover seating (where seats can be flipped to face the direction of travel) over fixed seating. A more substantial preference was found for both these configurations over transverse, or subway style, seating, likely because transverse seats offer less capacity and comfort. The value from walkover seating versus transverse seating was equal to an 8-minute savings in IVTT. FIGURE A-8 shows the relative importance of the seating-related attributes measured in this study on an indexed importance scale.

Literature and Practice Reviews A-17 Source: Pepper et al., 2003 FIGURE A-8. Importance of scores for on-train attributes. In the same study, respondents evaluated six different seat types and were willing to pay roughly 20% more in fare ($0.90 per trip) to have the most preferred seat instead of the least preferred seat, or about half that ($0.45) to have the most preferred seat compared with the seat type currently used on New Jersey TRANSIT trains (FIGURE A-9). As was shown in FIGURE A-8, seat comfort was of particular importance to respondents. Seat comfort was also the third most important on-board attribute in the Douglas and Karpouzis (2006) study behind air conditioning and layout. Source: Pepper at al., 2003 FIGURE A-9. Importance scores for seat types.

A-18 Characteristics of premium Transit Services that Affect Choice of Mode As shown in FIGURE A-8, additional layout and seating attributes such as seat material or the increased seating options from a multilevel train were less strongly valued. In September 2008, the CTA conducted two seating configuration studies, the max- capacity rail car experiment (Chicago Transit Authority 2008a) and a seat-less bus experiment (Chicago Transit Authority 2008b), to assess configuration preferences within high-capacity vehicles and to observe how customers used the additional space within the vehicles afforded by the altered configurations. In the high-capacity rail car experiment, 45% of customers preferred the new rail car configuration with less seating and more open space over the standard configuration (36%), but it was observed that this was likely due to the fact that most of these customers were interviewed while standing. The additional open space made it easier for riders to get into the car and move quickly to seats. Nearly three-quarters of respondents had carried on one or more items and the new rail car configuration afforded more space to accommodate these items, mostly stored between their feet. Based on survey and observational data and customer comments, it was recommended that the new rail cars work well but that they should be used during optimal peak periods. Results of the seat-less bus experiment showed that 54% of riders preferred the standard seating arrangement on the bus vs. 34% who preferred the seat-less arrangement. While the open arrangement again provided more space for carry-on articles, only 36% of bus riders had carry- on bags. Customers preferred getting a seat and suggested that the seat-less design be used only on crowded routes and during peak periods. Crowding. The level of crowding on-board has an impact on both comfort and the ability to engage in productive activity. Crowding can be expressed as the availability of seating— whether a traveler gets a seat—and also the load factor—the ratio of occupancy to the total number of seats. Including both attributes can improve model fit since the benefit of getting a seat (or the cost of standing) depends on the level of crowding. Returning to the multilevel coach study (Pepper et al. 2003), a seating capacity attribute was measured in addition to the multilevel attribute to identify the value placed on multilevel coaches specific to the increased capacity the coaches provide. Not surprisingly, given the crowded conditions existing at the time, it was the increased capacity of the coaches, and not the increased seating options, that offered the greater benefit—an equivalent of over 8 minutes of IVTT. A stated preference study by Hensher et al. (2003) found that the benefit from having a seat for an entire trip nearly offset the cost of IVTT, with respondents equating the benefit to an 88% decrease in IVTT. Respondents were willing to accept an approximately 30% increase in IVTT to only have to stand part of the way. That even sitting part of the way provided positive utility would indicate that travelers are accustomed to high load factors and therefore standing for the majority of bus trips. This could explain the high value placed on having a seat for the entire trip.

Literature and practice Reviews A-19 TABLE A-4 and TABLE A-5 illustrate the added explanatory power gained from specifying both load factor and seat availability. A train at full capacity (200% load factor) increases the cost of IVTT by 74%. Standing for 20 minutes or longer increases the IVTT cost by 81%. However, the combination of these two conditions—crushed standing—increases perceived IVTT costs by 152% (Litman 2007). TABLE A-4. Value of on-train load factor. Level of Crowding (Load Factor) Crowding Factor (Additional Time) 80% 0% 100% 10% 160% 60% 200% 74% Source: Douglas and Karpouzis, 2006 TABLE A-5. Value of on-train crowding. Level of Crowding Crowding Factor (Additional Time) Crowded seat 17% Stand 10 minutes or less 34% Stand 20 minutes or longer 81% Crush stand 10 minutes or less 104% Crush stand 20 minutes or longer 152% Source: Litman, 2007 An additional study modeling commuter trips between Long Island and Manhattan (RSG 2006b) presented and modeled seat availability as a probability—the number of times the rider gets a seat out of ten trips. This specification produced statistically significant coefficients, and when interacted with IVTT, found that the benefit from seating increased roughly linearly with trip duration (TABLE A-6). TABLE A-6. Value of seat availability. Description Marg. Rate of Substitution Seating availability very short trip (IVTT < 15 min.) $1.44 Seating availability short trip (IVTT 15 & < 30 min.) $3.83 Seating availability long trip (IVTT 30 min.) $6.47 Source: RSG, 2006b

A-20 Characteristics of premium Transit Services that Affect Choice of Mode Other Comfort. As mentioned previously, air conditioning (along with layout and design) provided the largest on-board benefit to respondents in the Douglas and Karpouzis (2006) study (TABLE A-7). The importance of air conditioning and heating on-board, as well as in the station, would presumably vary considerably based on climate. Quietness, smoothness, and on-board cleanliness were also found to be important to travelers in numerous studies. While the relative preference for these attributes varied in studies, they were consistently among the most valued comfort characteristics. Smoothness was found to be particularly important to respondents in the Swanson et al. (1997) bus study, where the benefits from a smooth ride compared with a rough ride were the second most highly valued. TABLE A-7. Value of train improvements. Type of Train Improvement Additional Time* (Increased On-board Time) Air conditioning 29% Layout 29% Seat comfort 24% Quietness 24% Cleanliness 22% Smoothness of ride 21% *Value of improving train attribute rating from 50% to 60% Source: Douglas and Karpouzis, 2006 Accessibility As identified in research customer satisfaction studies, transit accessibility has several aspects to it, including the trip to the station, movement through the station to the vehicle, and boarding the vehicle. Mode choice models traditionally represent walk and drive accessibility to the station at least in terms of travel time, but other accessibility details are often ignored, though they can impact travel time or transit utility measurably. Auto and Walk Accessibility. Transit travel begins with the access trip to the station or stop. For those living in less urban areas who need to drive to access transit, parking availability can make public transportation significantly more attractive. The existence of parking facilities and their capacity can also differentiate services like commuter rail from those operating in more pedestrian-friendly urban areas. For those accessing transit on foot, the level of walk accessibility, or pedestrian friendliness, can also influence the willingness to use public transportation. Areas with consistent sidewalks and clear crosswalks make transit access safer and easier. As described well in TCRP Report 95, Chapter 17, the proximity of the transit station to surrounding real estate is one of the most important attributes of a successful transit system (Evans et al. 2007). One example cited in this appendix presents data from the San Francisco Bay Area from a 2006 travel survey done

Literature and Practice Reviews A-21 by the Metropolitan Transportation Commission. This study found that the transit share for commute trips was: 42% for trips where both the residence and workplace were within 0.5 mile of a transit stop/station; 28% for trips where the workplace was within 0.5 mile of a transit stop/station but the residence was not; 16% for trips where the residence was within 0.5 mile of a transit stop/station but the workplace was not; and 4% for trips where neither the residence nor workplace were within 0.5 mile of a transit • • • • stop/station. Research by Cervero (1993), also discussed in TCRP Report 95, Chapter 17, shows comparable patterns (FIGURE A-10 and FIGURE A-11). Each of these studies considers contexts that are unique and have confounding factors, but the relationship between proximity and use is striking. Source: Cervero, 1993 FIGURE A-10. Work trip rail share by distance from residence to station.

A-22 Characteristics of Premium Transit Services that Affect Choice of Mode Source: Cervero, 1993 FIGURE A-11. Work trip rail share by distance from workplace to station. Goals related to drive accessibility and walk accessibility conflict. The space required to park vehicles takes up and fragments the land adjacent to the station and diminishes walk accessibility. Additionally, conflicts between pedestrians and vehicles may more likely occur to the extent that both try to access a particular station. Travel models tend to represent both walk and drive accessibility, but clearly issues related to capacity, safety, and comfort mentioned above can apply to the access trip. As discussed in TCRP Report 95, Chapter 17, transit agencies occasionally have the opportunity to work with developers to design TODs and address aspects related to accessibility to the transit station (Evans et al. 2007). Two interesting examples are from King County Washington, as described by Shelton and Lo (2003) and presented in FIGURE A-12. The design of these TODs considered how to integrate transit parking within the development, as well as policies to promote transit ridership, such as providing residents of the developments with free transit passes. The TODs were located adjacent to transit centers.

Literature and Practice Reviews A-23 Source: Shelton and Lo, 2003 FIGURE A-12. Two King County bus transit-oriented developments. The above examples certainly illustrate the complexity of transit planning, but from a forecasting standpoint the representation of the parking and fare policies in mode choice is relatively straightforward. However, the policies outlined above suggest that people who choose to live in these developments might be more likely to use transit on account of socioeconomic factors, or a desire to live near transit, or because of the free transit pass. If this is the case, understanding how these factors influence or relate to auto-ownership or trip-distribution patterns is important to consider in order to forecast transit usage correctly. Mode choice models could capture some of these impacts with pedestrian environment or TOD variables, but the underlying behavior is quite complex. In-station Accessibility. Elevators and escalators improve movement and access throughout stations and have therefore been studied in various customer satisfaction surveys and in the study of NYC station amenities (Spitz et al. 2007). In this study, travelers valued the benefit of escalators equal to roughly 4 fewer minutes of IVTT, while the most valued attribute—the addition of surveillance cameras and emergency buttons—provided benefits equal to about 7 minutes of IVTT. The same study also found that wider passages and stairways throughout the station provided half the benefit of escalators, or 2 fewer minutes of IVTT. The value of elevators and escalators was also measured in the Douglas and Karpouzis (2006) work, with a 10% customer satisfaction improvement yielding the benefit of a 16% decrease in IVTT.

A-24 Characteristics of premium Transit Services that Affect Choice of Mode Ease of Boarding. Primarily quantified in bus studies, the ease of boarding can significantly impact a large portion of the traveling population, with wider doorways and ground- level entry and aisles serving as attractive features on many bus rapid transit services. Estimates of the effect of this attribute in the Hensher et al. (2003) study showed that the design of the entry-way, in terms of width and number of steps is the most important attribute for certain segments of the population. Ease of train boarding was also a highly important attribute in the Douglas and Karpouzis (2006) study, with a 10% improvement in customer satisfaction equating to a 24% decrease in IVTT—third highest value of on-board attributes after air conditioning and layout. Two other ease-of-boarding attributes that were identified in the literature include the ability of the bus to be able to pull up to the curb and the material used to make station platforms. For the latter, a slippery surface potentially makes boarding more dangerous. The UTA recently implemented BRT service on 3500 South, and one important new aspect of the service is the design of the buses (FIGURE A-13). The buses are designed with low floors and three doors for easier and quicker boarding and alighting. The buses are new vehicles that have very large windows and comfortable seating, all of which has increased customer satisfaction. FIGURE A-13. UTA's MAX BRT vehicle. Productivity/Enjoyment Research studies by Lyons et al. (2007) and Ory and Mokhtarian (2005) posit that in- vehicle time, rather than being wasted, can be enjoyable and/or productive and therefore can possess a positive utility. Ability for Activity. In the Lyons et al. (2007) study of rail passengers in Great Britain, 23% of respondents felt they made “very worthwhile use” of their travel time and an additional 55% made “some use” of their time while traveling. Worthwhile use does not mean that this time was necessarily economically productive (for some people, sleeping on the train was very worthwhile), but this does indicate some utility gained from travel time. Furthermore, the increasing dissemination of electronic devices, whether for work or leisure, is shown to decrease the cost of in-vehicle time. Over one-fifth of rail passengers carrying electronic devices reported that such devices made the time on the train significantly less onerous and 46% agreed that they made the time seem to pass more quickly. Activity Services (WiFi). To the extent that on-board attributes can facilitate the productive and/or worthwhile use of travel time, there is potential to increase the utility of transit travel. Recently (2008), the UTA introduced on-board wireless Internet service on its

Literature and practice Reviews A-25 FrontRunner commuter rail line to increase the appeal of the service (FIGURE A-14). Wireless Internet is also available on certain of UTA’s express buses, and as this amenity becomes more pervasive, it could serve as a large draw for passengers. According to the UTA, the popularity of WiFi exceeded expectations, with 1,000 passengers, or 1 in 9, using the service. FIGURE A-14. On-board wireless Internet service. Journey Enjoyment. Ory and Mokhtarian (2005) identify a number of reasons why individuals travel simply for the sake of enjoyment: adventure seeking, variety seeking, independence, control, status, buffer, exposure to the environment, scenery and other amenities, synergy (excess travel if productive), escape, curiosity, conquest (taking a new route through an unfamiliar area), physical exercise, and the therapeutic value of movement/travel. As transit travel becomes more comfortable, individuals may be more inclined to use transit modes for many of the reasons above. The same study also asked individuals to rate how they “feel” about travel by various modes whether or not they actually use the modes on a regular basis. Notable to the differentiation between premium and conventional services, 31% of respondents liked, or strongly liked, travel by rail compared with only 8% for travel by bus. Information Services One subject of much discussion in research is the impact of travel information on travel choices. Many studies have examined the ability of real-time service information to mitigate the costs associated with wait time, unreliability, and transfers since it is the uncertainty of arrival that increases the perceived time and therefore the cost of waiting (Li 2003). Studies have also frequently explored the effect of providing more basic service information such as timetables and maps at stations and stops. Route Information. The Swanson et al. (1997) study estimated the value of both general and locally customized paper-based information at stops, as well as the importance of Countdown, a real-time travel information service. The results showed that service information is highly valued, yet the high-technology Countdown system is actually less preferred than the provision of locally customized information at stops. In NYC train stations, “information on platforms and walls” was found to be worth nearly 3 minutes of IVTT, while real-time information at New Jersey BRT stations was found to be equivalent to a 5-minute reduction in IVTT (Spitz et al. 2007).

A-26 Characteristics of premium Transit Services that Affect Choice of Mode FIGURE A-15. MTA's off-board fare vending machine. The somewhat inconclusive value of maps, timetables, or real-time information suggests the need for additional study of the effects of these attributes. In areas with high service frequency, timetables are often ignored. For those who only use transit for frequent or routine trips, such as to and from work, maps and timetables may be unnecessary. However, work trips often demand punctuality, and particularly in areas with unreliable service, real-time information may be highly valued, allowing travelers to change plans mid-trip or to minimize anxiety from arrival uncertainty. As technology develops and laptop and handheld devices become more prevalent, real-time information may be increasingly available with less cost for infrastructure improvement. Announcements. As with real-time information, clear announcements both in the station and on-board are hypothesized to reduce the uncertainty of transit travel as well as the mental effort required for travel (Li 2003). The NYC station amenities study (Spitz et al. 2007) found clear announcements to be the third most valued attribute behind “improved cleaning and maintenance” but ahead of the benchmark attribute, “five-minute travel time savings on your train trip.” Further, the understandability, or intuitiveness, of routes and schedules, the accuracy of information, the availability of obtaining information over the phone, and the notification of service changes are also identified as attributes affecting the perceived cost of transit travel. Fare Payment Proof of Payment. There have been many new fare payment methods introduced in recent years. One method, proof of payment, requires passengers to purchase a ticket before boarding and tickets may be checked only after being seated—a method significantly reducing the perceived cost of a trip by reducing station dwell times. In a study quantifying BRT attributes (RSG 2007), the attribute, “fare payment is available pre-boarding and is fast,” was worth nearly 5 minutes of IVTT, while “proof of payment by inspectors increasing speed and ease of boarding” was approximately two-thirds as valuable. These attributes, while improving travel time, were still valued less than attributes improving headways, reducing access and egress time, shelter, and real-time information. The respondents in the study lived in a less urban environment, perhaps with fewer stops or fewer passengers boarding at each stop. Also, boarding may have occurred while in traffic, mitigating the savings from reduced dwell time. Payment Ease and Availability. Customer satisfaction surveys examine many different aspects of in-station fare payment, including the availability of places to purchase a ticket, the number of fare card vending machines, the ease of purchasing a ticket or pass, and the integration and automation of fares among modes. The introduction of innovative fare payment methods (FIGURE A-15) reflects advances in technology but also the value of improvements to passengers. In fact, “ticketing” was the second most important station attribute tested in the Douglas and Karpouzis (2006) study.

Literature and practice Reviews A-27 Safety Safety is a significant concern for many transit travelers, particularly personal safety from crime. This can be crime in a parking lot, at a station or stop, or on-board, and the perceived threat of crime increases at night. Safety from accidents is also frequently cited as a concern for travelers, with fixed-guideway services and the presence of emergency exits perceived as easing safety concerns. While safety is difficult to quantify, among 11 NYC train station attributes tested in one study (Spitz et al. 2007), “adding surveillance cameras and emergency call buttons” was the most highly valued attribute—equal to a 7-minute improvement in IVTT. The same study found less preference for other safety improvements, including “enhanced lighting on station platforms” and “improved visibility and open sightlines.” These attributes were valued at roughly 3 minutes of IVTT. Interestingly, lighting at bus stops, despite being the only safety-related attribute measured, was of relatively little importance to respondents in the Swanson et al. (1997) study. On an indexed willingness-to-pay scale from 0 to 100, the benefit of a clean, well-maintained bus shelter compared with a dirty, vandalized shelter scored a 100, while the benefit of lighting at bus stops scored a 26. This study was conducted in a less urban area, however; with presumably less crime, aesthetic concerns can replace those for personal safety. Case Studies of Transit Attribute Evaluations Valuing Rail Service Attributes through Rating Surveys Douglas and Karpouzis (2006) The authors conducted research on behalf of RailCorp – a transit rail agency in New South Wales, Australia – using customer satisfaction ratings of individual service attributes to explain overall ratings of service. A regression model estimated the share that each attribute contributed to the overall satisfaction rating. Survey respondents were further asked the amount of travel time savings required to rate the “on-train time” attribute as “excellent” – the top score. Using the average rating score for each attribute, the value of the attribute could be expressed in terms of the in-vehicle time respondents were willing to accept for a 10% improvement in the individual attribute rating.

A-28 Characteristics of premium Transit Services that Affect Choice of Mode The tables below (TABLE A-8, TABLE A-9, and TABLE A-10) display the percentage of in-vehicle time respondents would accept for a 10% improvement in each attribute rating. TABLE A-8. Value of service improvements. Type of Service Improvement Additional Time* (Increased On-board Time) Reliability 222% On-train time 100% Service frequency 76% Seat availability 36% Train security day 21% Station security day 21% Train security night 12% Station security night 12% *Value of improving train attribute rating from 50% to 60% Source: Douglas and Karpouzis, 2006 TABLE A-9. Value of train improvements. Type of Train Improvement Additional Time* (Increased On-board Time) Air conditioning 29% Layout 29% Ease of boarding 24% Seat comfort 24% Quietness 24% Train outside 22% Cleanliness 22% Smoothness of ride 21% Announcements 12% Lighting 10% Graffiti 09% *Value of improving train attribute rating from 50% to 60% Source: Douglas and Karpouzis, 2006 TABLE A-10. Value of station improvements. Type of Station Improvement Additional Time* (Increased On-board Time) Staff 33% Ticketing 28% Bus 21% Cleanliness 19% Graffiti 19% Station building 17% Subway-overbridge 16% Lifts & escalators 16% Lighting 14%

Literature and practice Reviews A-29 TABLE A-10. (Continued). Type of Station Improvement Additional Time* (Increased On-board Time) Announcements 12% Information 12% Signing 10% Station on-off 07% Weather protection 07% Toilets 07% Platform seating 05% Platform surface 05% Taxi 03% Telephone 03% Retail 03% Car park 02% Car drop-off 02% Bicycle 02% *Value of improving station attribute rating from 50% to 60% Source: Douglas and Karpouzis, 2006 Service Quality–Developing a Service Quality Index in the Provision of Commercial Bus Contracts Hensher, Stopher, and Bullock (2003) A stated preference study was conducted evaluating various attributes of bus services in New South Wales Australia. A literature review, interviews with bus operators, and a pilot study identified 13 major dimensions of service quality from the user’s perspective, and three levels were developed for each attribute (TABLE A-11). TABLE A-11. Attributes and attribute levels in the SP experiment. Attribute Level 1 Level 2 Level 3 Bus travel time 25% less Same 25% more Bus fare 20% less Same 20% more Ticket type Cash fare Pre-purchased bus- only 10-trip ticket or weekly Integrated (bus and other mode) Buses per hour at this bus stop 50% more service Same as now 50% less service Time of arrival at bus stop On time 5 minutes late 10 minutes late Time walking to bus stop Same An extra 5 minutes An extra 10 minutes Seat availability on bus Seated all the way Stand part of the way Stand all of the way Information at bus stop Timetable and map Timetable, no map No timetable, no map Access to bus Wide entry, no steps Wide entry, 2 steps Narrow entry, 4 steps Bus stop facilities Seats only Seats under cover No seat or shelter Temperature on bus Too hot Just right Too cold Driver attitude Very friendly Friendly enough Generally unfriendly General cleanliness on-board Very clean Clean enough Not clean enough Source: Hensher, Stopher, and Bullock, 2003

A-30 Characteristics of premium Transit Services that Affect Choice of Mode A paper survey was administered to nine service segments—three bus depots and three route types per depot. These nine segments were treated as individual nests in order to scale coefficients across segments, which allowed direct comparison of coefficient values across segments while also accounting for preference variation between the segments. The same coefficients were evaluated in each segment. TABLE A-12 presents results from the nested logit model estimation. Attributes with insignificant coefficients were excluded from the results. TABLE A-12. Final model used to identify the importance weights and scale differences between segments for scheduled routes. Segment importance and scale weights (t-value in brackets)* Attribute S1 S2 S3 S4 S5 S6 S7 S8 S9 Travel time (minutes) -0.0333 -0.0346 -0.0249 -0.044 -0.0396 -0.0356 -0.028 -0.0272 -0.0362 (-3.8) (-3.2) (-1.5) (-4.9) (-3.9) (-3.2) (-3.3) (-2.7) (-2.1) One-way bus fare ($) -0.6519 -0.7136 -0.7508 -0.5592 -0.6394 -0.5948 -0.6256 -0.5543 -0.5543 (-4.5) (-4.4) (-4) (-4.3) (-4.6) (-4.4) (-4.2) (-2.9) (-2.9) Unreliability (minutes) -0.0317 -0.0322 -0.0626 -0.0399 -0.0649 -0.0119 -0.0116 -0.01127 -0.1029 (-1.8) (-1.4) (-1.7) (-2.6) (-3.3) (-0.5) (-0.8) (-3.9) (-1.9) Access time to bus stop (minutes) -0.0248 -0.0725 -0.0859 -0.0081 -0.0449 -0.0696 -0.0128 -0.0567 -0.0768 (-2.0) (-3.9) (-3.4) (-0.8) (-3.4) (-3.4) (-1.1) (-3.6) (-2.7) Bus frequency (/hour) 0.0923 0.084 0.2729 0.049 0.0858 0.1187 0.0869 0.144 0.0523 (-3.0) (-2.0) (-2.8) (-2.0) (-2.6) (-2.2) (-2.8) (-2.9) (-0.6) Seat all way (1,0) 0.6529 0.6661 0.5159 0.438 0.4622 0.531 0.7734 0.356 0.9531(-3.8) (-3.0) (-2.5) (-3.1) (-2.8) (-2.1) (-4.7) (-1.9) (-2.0) Stand part way (1,0) 0.2367 0.2367 0.2367 0.2367 0.2367 0.2367 (-2.5) (-2.5) (-2.5) (-2.5) (-2.5) (-2.5) No timetable, no map (1,0) -0.185 -0.4216 -0.1372 -0.2464 -0.2913 -0.2033 -0.121 (-1.4) (-2.3) (-1.1) (-1.5) (-1.9) (-1.2) (-0.5) Narrow 4 steps (1,0) -0.4455 -0.1535 -0.5709 (-2.7) (-0.8) (-3.1) Wide entry 2 steps (1,0) -0.5124 -0.4899 -0.5748 (-3.2) (-2.7) (-3.3) Seat only at stop (1,0) 0.6102 0.6102 0.6102 0.1851 0.1851 0.1851 0.1851 0.1851 0.1851 (-4.2) (-4.2) (-4.2) (-2.5) (-2.5) (-2.5) (-2.5) (-2.5) (-2.5) Seat under cover at bus stop (1,0) 0.6102 0.6102 0.6102 0.1851 0.1851 0.1851 0.1851 0.1851 0.1851 (-4.2) (-4.2) (-4.2) (-2.5) (-2.5) (-2.5) (-2.5) (-2.5) (-2.5) Very clean bus (1,0) 0.3228 0.3228 0.2262 0.3228 (-2.9) (-2.9) (-1.7) (-2.9) Very friendly driver (1,0) 0.1704 0.1704 0.2089 0.2263 0.2263 (-1.4) (-1.4) (-1.7) (-1.9) (-1.9) VTTS ($/h) 3.06 2.92 1.99 4.72 3.72 3.59 2.68 2.94 3.92 No. of observations** 580 511 472 454 646 336 463 304 122 Scale value 0.9835 0.5019 0.6326 1 0.727 0.4212 1.065 1.0727 0.837 -4.6 -3.8 -4.4 (fixed) -4.7 -3 -5.6 -4.4 -3.2 Log-likelihood -3848.9 Pseudo-R2 0.69 * Missing attribute weights mean that the attribute was too insignificant to report for the segment where it was highly non-significant. ** The minimum number of observations per respondents was one and the maximum was 3 (i.e., 3 or less SP experiments completed). Note: School children on passes have been excluded. Source: Hensher, Stopher, and Bullock, 2003

Literature and practice Reviews A-31 Measuring Bus Passenger Preferences Swanson, Ampt, and Jones (1997) A study of various bus transit attributes performed in Britain aimed to carry out a, “...study of willingness to pay for bus service and infrastructure improvements in order to inform the project appraisal process.” “The objectives were to identify those service attributes of concern to respondents, meaning those of which they express awareness, how large changes in those attributes need to be for passengers to recognize a difference, and the best way of summarizing and presenting this experience to respondents in a pictorial format.” Artist sketches were deemed the best method for displaying service attributes, offering enough context to make the situation representative (e.g., a bus shelter on rural road as opposed to a city street) without the distracting level of detail from a photograph. A protocol analysis was conducted to understand passengers’ decision-making processes, from which attributes were selected and grouped into journey stages (TABLE A-13). TABLE A-13. Journey stages and attributes tested. Stage of Journey Attributes Tested 1. Pre-trip information Maps Timetables Customized local information Telephone information services 2. The bus stop infrastructure Type of shelter Lighting Cleanliness and state of repair 3. Waiting at the bus stop Fixed information display Real-time information (i.e., countdown and/or telephone services) Service reliability 4. The bus at the curbside Compulsory or request stop Ease of identifying correct bus Stopping position of bus Design of vehicle entry steps 5. Encountering the driver Driver appearance Driver helpfulness Driver identification Availability of change 6. Moving to your seat Level of crowding Design of luggage storage area Seating configuration Quality of vehicle motion 7. Traveling in a seat Types of seats Spaciousness of seats Type of ventilation Cleanliness Travel time 8. Leaving the bus Provision of information on bus Number and location of doors Source: Swanson, Ampt, and Jones, 1997

A-32 Characteristics of premium Transit Services that Affect Choice of Mode Respondents first evaluated attributes from three randomly selected journey stages. Two alternatives were presented—one sketch matching the respondent’s current situation and one sketch of a hypothetical situation—and respondents provided the direction and magnitude of their preference on a 100 point scale. An additional exercise determined monetary valuations for service improvements, asking respondents their likelihood of paying an increased fare given a certain bundle of service improvements. A final “capping exercise” asked respondents to rank areas for improvement, and then showed respondents a text-based stated preference exercise where they evaluated improvements in the highest ranked areas against a fare increase. This determined the maximum willingness to pay for any improvement package. Individual models were estimated, with relevant results indexed and presented (TABLE A-14). TABLE A-14. Selected valuations, expressed as indices. Improvement Mean WTP, Indexed* t-ratio Dirty, vandalized shelter vs. clean, well-maintained shelter -100 -9.1 Rough vehicle motion vs. smooth motion -89 -7.0 Guaranteed provision of customized information at stops vs. none 85 9.1 Highly crowded vs. low crowding -81 -6.8 Countdown 76 9.0 Guaranteed current style information at bus stops vs. none 75 8.8 Dirty bus interior vs. clean -72 -8.5 Best improvement to reliability vs. current (long headway) 66 6.0 Best improvement to reliability (short headway) 60 5.1 Interaction between countdown and best reliability improvement (long headway) -58 -4.1 Interaction between countdown and best reliability improvement (short headway) -57 -4.5 Medium-smooth vehicle motion vs. smooth motion -54 -4.9 Bus able to pull in close to curb 49 4.8 Bus shelter with roof and end-panel vs. no shelter 47 5.6 Medium crowding vs. low crowding -40 -3.9 Driver gives change when needed 34 3.6 Electronic display of next bus stop name 33 6.5 Lighting at bus stops 26 4.4 Roomy seats vs. cramped seats 25 5.0 Low-floor buses vs. high steps 20 2.4 * The largest willingness-to-pay valuation has been indexed at 100. Source: Swanson, Ampt, and Jones, 1997

Literature and Practice Reviews A-33 Source: Resource Systems Group, Inc., 2007 FIGURE A-16. NJ Transit MaxDiff conjoint screenshot. FIGURE A-17 model results (normalized to a 0 to 100 scale) indicated that increased headways, expanded coverage (e.g., bus stop within walking distance of work), and decreased travel times were the most important factors to the respondents, while factors such as a level boarding platform or bus branding were of the least importance. NJ Transit MaxDiff Conjoint Training: A Guide to Designing and Preparing MaxDiff Surveys and Analyzing and Interpreting MaxDiff Data Resource Systems Group, Inc. (2007) A study of Bus Rapid Transit (BRT) attributes measured the preferences of individuals traveling in a heavily congested corridor in New Jersey. A maximum differences scaling conjoint analysis was performed, displaying four attributes at a time, and asking respondents to select the attribute considered most important and the attribute considered least important (FIGURE A-16).

Source: Resource Systems Group, Inc., 2007 FIGURE A-17. NJ Transit MaxDiff results. 0 10 20 30 40 50 60 70 80 90 100 Level BRT Platforms at stations: easier boarding and deboarding BRT Bus is branded “ROUTE 1 EXPRESS” BRT costs the same as a regular bus Good sidewalks between BRT stop and your home POP by inspectors, increasing speed and ease of boarding Good sidewalks between the BRT stop workplace BRT costs $1 less than an auto trip for same trip BRT gets priority signal when coming to a traffic light BRT jumps to the head of the line at red light at intersection BRT bus vehicle is a hybrid BRT is 20 minutes in peak and 60 minutes the rest Fare payment for bus is available pre-boarding and is fast BRT has convenient Transfers to NEC BRT takes 5 minute less travel time than current trip Real time info of arrivals at bus station and online BRT Bus stations are enclosed and out of weather BRT costs $2 less than an auto trip BRT bus vehicle more clean, quiet, comfortable vs regular bus BRT stop w/in walking distance of home BRT Bus has dedicated right-of-way BRT is 15 minutes in the peak 30 minutes the rest BRT is 10 minute less travel time than your current trip BRT stop w/in walking distance of work BRT is 10 minutes in the peak period 20 minutes the rest Normalized Score

Literature and practice Reviews A-35 Qualitative and Quantitative Approaches for Studying Transit Stations Spitz, Greene, Adler, and Dallison (2007) The value of individual train station amenities was quantified for ten recently improved NYC train stations. Focus groups helped identify station characteristics of particular importance to travelers, after which a maximum differences scaling conjoint analysis was performed to attribute. Security concerns featured prominently with surveillance cameras and emergency call buttons found to be most important and enhanced lighting and visibility of moderate importance (FIGURE A-18). Source: Spitz, Greene, Adler, and Dallison, 2007 FIGURE A-18. NYC train stations MaxDiff results. Customers’ Perspectives on Using Multilevel Coaches to Increase Rail System Capacity Pepper, Spitz, and Adler (2007) New Jersey Transit planned to purchase multilevel coaches to address a critical passenger capacity issue on trains accessing New York’s Penn Station; but first, a study was conducted to determine how the multilevel coaches should be designed to both provide the needed additional system capacity and to reflect customers’ preferences. The study used adaptive conjoint analysis 0 10 20 30 40 50 60 70 80 90 100 Station art such as mosaics or stained glass Using high quality/attractive materials throughout the station, such as granite Wider passages and stairways Availability of subway maps on station platforms and walls Improved visibility and open sightlines in the station Enhanced lighting on station platforms Escalators in the station 5 minute travel time savings on your train trip Clarity of announcements from public address speakers Improved cleaning and maintenance Adding surveillance cameras and emergency call buttons Normalized Score Error bars represent 95% confidence interval measure the relative importance of the amenities as well as the strength of preference for each

A-36 Characteristics of premium Transit Services that Affect Choice of Mode to value various interior attributes including seating configuration and seat design, which were directly related to the amount of seated (and standee) capacity that the coaches would provide. The indexed importance scores (utilities) shown in TABLE A-14 (in prior section) indicate respondents’ strength of preference for the various on-board attributes. Level-of-service attributes including travel time, fare, and service frequency were also measured to serve as benchmarks and to allow for estimation of respondents’ willingness to pay for attribute improvements. Customer Satisfaction Surveys Chicago Transit Authority 2003 and Washington Metropolitan Area Transit Authority (2006) Customer Satisfaction Measurement. Findings of Customer Surveys. TABLE A-16 shows the pretest criteria for selection of customer satisfaction variables from a survey conducted for Sacramento Regional Transit. TABLE A-15 shows rail and bus service attributes measured in customer satisfaction surveys. • • • TABLE A-15. Rail and bus service attributes measured in customer satisfaction surveys. Dimension Attributes Convenience Using Metrorail/bus for shopping trips Using Metrorail/bus for work trips Using Metrorail/bus for entertainment trips Making transfers Parking at rail stations (METRORAIL survey ONLY) Riding experience Cleanliness of rail cars/buses Cleanliness of rail stations/bus stops Comfort of the overall ride Smell of rail cars/buses Temperature inside rail cars/buses Availability of seating when riding on train/bus Comfort of seats on the train/bus Number of people on the train/bus Number of bus stops that have shelters (METROBUS survey Only) Safety From accidents while riding From crime during daylight hours while riding From crime during nighttime hours while riding At bus stops/rail stations during daylight hours At bus stops/rail stations during nighttime hours In Metro parking lots during daylight hours In Metro parking lots during nighttime hours Access Distance of the nearest bus stop from home (METROBUS survey ONLY) Distance of the nearest bus stop from destination (METROBUS survey ONLY)

Literature and practice Reviews A-37 TABLE A-15. (Continued). Dimension Attributes Access (cont’d) Frequency of buses from home to closest Metrorail station Number of transfers needed to get to final destination Wait time at start of trip Availability of parking at rail station (METRORAIL survey ONLY) Vertical transportation One or more elevators were not working at a rail station One or more escalators were not working at a rail station Reliability Trains/buses getting to the destination on time Stops were announced by train/bus operators Metrobus arriving more than 5 minutes early or late (METROBUS survey ONLY) Having to wait more than 15 minutes for the next train (METRORAIL survey ONLY) Customer service Satisfaction with helpfulness of bus operators (METROBUS survey ONLY) Satisfaction with the level of service of Metro personnel in rail stations (METRORAIL survey ONLY) Satisfaction with clarity of operator announcements at stops Fares Value of ride fare Satisfaction with cost of riding Process of purchasing farecards and passes Process of obtaining refunds or replacement farecards or passes Cost of parking at Metrorail stations (METRORAIL survey ONLY) Communications Utility of digital displays – PIDS Understandability of route/schedule information Responsiveness of WMATA Timeliness of schedule information Information availability Source: Chicago Transit Authority 2003 and Washington Metropolitan Area Transit Authority (2006) TABLE A-16. Pretest criteria for selection of customer satisfaction variables. Attributes Safe and competent drivers Buses - Trains running when schedule says Safety from crime on-board vehicles Frequency of service on weekdays Total travel time for your trip Security at light rail stations and bus transfer points Friendly, courteous operators Notification of service disruptions Freedom from nuisance behavior of other passengers Availability of schedule information Frequency of delays for repairs - emergencies Cleanliness of vehicles Trains and buses which are not overcrowded Visibility of security staff on the light rail system Helpfulness of telephone information center Fare you pay to ride RT

A-38 Characteristics of premium Transit Services that Affect Choice of Mode TABLE A-16. (Continued). Attributes Availability of shelters and benches at stops Cleanliness of light rail stations and bus transfer point Time buses start running in morning Usefulness of RT website Time buses stop running in evening Safety of vehicle parked at light rail station Connecting bus service at train stations - main bus stop Process of filing a complaint Stop announcements made by operators Accessibility of vehicles to persons with disabilities Availability of bike racks Frequency of service on weekends Reliability of wheelchair lifts Source: Transit Marketing, LLC and CJI Research Corporation (2006) Applied Models Practitioners have been attempting for many years to improve their models’ ability to estimate transit ridership and, in particular, to represent the distribution of ridership among conventional bus and premium modes such as rail or BRT services. Their attempts have been focused in three areas: (1) a more realistic representation of transportation supply, (2) verification of travel patterns, and (3) incorporating non-traditional attributes of premium modes into the modal choice decision. The first two areas are critically important to producing credible forecasts, but the third is the focus of this appendix. A summary of various methods of incorporating non-traditional attributes of premium modes in model choice models and techniques employed in their application based on eight case studies is provided here. Further technical details of the modeling and forecasting procedures along with relevant coefficients and parameters have been provided in this appendix. Incorporating Attributes of Premium Modes Practitioners implementing models in areas with successful rail systems have often struggled with the fact that no matter how carefully transit services are represented inside their models, ridership is not estimated accurately on these premium services. Practitioners are also aware of experience in other areas that suggests that premium mode ridership is higher than what would be projected through traditional elasticities and attributes. TCRP Report 118: Bus Rapid Transit Practitioner’s Guide, for example, notes that new BRT systems in six cities experienced higher ridership increases than their traditional attributes would normally indicate (Kittleson, et al. 2007). The reasons behind underestimates of premium transit services are not usually known but have been speculated to be related to public perception of safety, heightened awareness, brand visibility, and various service attributes that are measurable but not typically included in most forecasting models. An example of the latter is peak travelers considering off-peak

Literature and practice Reviews A-39 service frequency in their choice process since they could elect to travel home in the middle of the day in case of a family emergency. Even if these factors were known with greater certainty, practitioners would struggle to inform their models about these new transit attributes. The quantification of some attributes is a problem. Concepts such as safety, system awareness, ease of use, and branding are compelling components of the choice process but difficult to quantify by an objective standard. Without quantification, however, it would be difficult to incorporate these attributes properly within travel forecasting models. Other attributes, such as reliability and seat availability, could possibly be correctly quantified and represented in travel models. However, travelers’ reaction to these attributes and those currently unquantifiable are unknown beyond general ridership patterns and anecdotal evidence. Given the uncertainty about modal preference and the difficulties of quantifying the underlying factors, practitioners trying to match observed transit usage typically use simplified approaches that try to represent a general preference toward certain modes without explicitly representing the reasons that these preferences might exist. Often this is done by introducing transit mode-specific constants favoring premium modes within the mode choice utility function. The purpose of these constants is to represent the sum of all non-traditional attributes that accrue to travelers who elect to use the premium service during the course of their trip. The value of the incremental mode-specific constant generally varies between 10 and 15 minutes of equivalent transit IVTT. This reflects the perceived difference between conventional bus and premium modes. Another less commonly employed technique is to discount the perceived travel time of the premium mode. This approach suggests that premium travel time benefits are proportionate to the time spent on a premium service and has the additional advantage that similar judgments of the merits of different transit options are made in both path choice and mode choice. This helps to present different alternatives that are selected within mode choice, accounting for demographic, attitudinal and the build environment, in addition to level-of-service vectors. Two recent efforts have attempted to refine both of these general approaches by apportioning benefits across several specific modal attributes. The FTA’s latest modeling guidance for New/Small Starts projects relates the size of the mode-specific constant to specific attributes such as schedule-free service, passenger amenities, and branding (FTA 2011). Attributes are assigned certain values. The presence of an attribute is required to increase the size of the mode-specific constant by the assigned amount. TCRP Report 118 offers a similar process for new BRT systems. The values are based on professional judgment in both efforts. This, in part, has led to further research to better address this issue. TABLE A-17 summarizes known case studies on this topic. The eight case studies described here represent a cross section of practitioners’ efforts to date. Non-traditional attributes were incorporated into the travel model in seven of the cases. Two case studies are guidance on accounting for non-traditional attributes in models. Two other case studies are applied research efforts whose results were not yet applied to model sets. In most cases, practitioners have attempted to account for the aggregate impact of all unmeasured attributes rather than focus on particular attributes. This occurred in seven of the

A-40 Characteristics of premium Transit Services that Affect Choice of Mode eight case studies and likely is a reflection of the difficulty of quantifying non-traditional attributes and assessing the traveler’s reaction to changes in those attributes. Two case studies attempt to relate incremental benefits to specific components of premium modes. The FTA New/Small Starts modeling guidance accrues benefits depending on the extent of changes to the following attributes: service reliability, branding, visibility, learn- ability, schedule-free service, hours of frequent service, and passenger amenities. TCRP Report 118 suggests accruing benefits depending on the presence of these attributes: running ways, station amenities, vehicle attributes, service patterns, intelligent transportation systems (ITS) applications, and branding. TABLE A-17. Case study summary. Model Applicaon# Case Study Aributes Phase Technique 1 FTA New/Small StartsModeling Guidance Reliability, branding, visibility, learn ability, schedule free service, hours of frequent service, passenger amenities Mode Choice Incremental bias constant 2 TCRP 118 – BRTPraconer’s Guide Running ways, staon amenies, vehicle aributes, service paerns, ITS applicaons and branding Post model Percentage adjustment toridership 3 Chicago Transit Authority & Metra New Starts Alternaves Analysis Walk ability, unmeasured rail preferences Auto ownership, path building, mode choice Ulity variable, travel me discount (15%) 4 Discounted travel me coefficient (models for Denver Regional Transportaon District and New York Metropolitan Transit Authority) Sum of all unmeasured LRT aributes Mode Choice Discounted travel me coefficient (30% for Denver, 25% for New York) 5 Charloe New StartsTravel Demand Model Aributes of formal park ride lots Mode Choice Shadow price penalties of 3 9 minutes on informal park ride lots 6 Southeast Florida Regional Planning Model (version 6.5) Sum of all unmeasured premium mode attributes Mode Choice Incremental mode specific bias constant 7 Lower Manhaan Jamaica/JFK Transportaon Project Seang availability Mode Choice(suggested) Ulity variable 8 Chicago Transit Authority Smart Card Acvity Analysis Revealed bus vs. rail preference Mode Choice (suggested) Discounted travel me (42%) and wait me coefficients (34%)

Literature and practice Reviews A-41 Only one case study focused on a particular attribute. The Lower Manhattan-Jamaica/JFK Transportation Project studied a number of different attributes, including seat type (bench or forward/reverse), transfer type (whether transfer occurred at a single platform), and seating availability. Only seating availability was found to be statistically significant. The model developed for the CTA and Metra New Starts Corridor Alternatives Analyses applies a general walk-ability attribute in several different phases of the model (AECOM 2006a). The walk-ability attribute is intended to capture the impact that different area types have on the likelihood of using transit. Although not specifically geared toward premium transit, walk-ability may do a better job than the standard area type definition. Application Techniques All of the case studies apply some type of modification to the mode choice model. Four case studies apply an incremental bias constant to premium mode utilities. Incremental bias constants calibrated in areas with premium modes in operation have values ranging from 3.9 to 71.5 minutes, estimated using standard mode choice calibration techniques. The FTA recommends that bias constants be limited to 5–15 minutes of equivalent IVTT—considerably less than some of the constants that practitioners have used in operational models. Three case studies included a discount on premium mode IVTT coefficient. Discounts range from 25% to 30% in application. The CTA Smart Card Analysis suggests a discount of 42% based on revealed preferences. A second case study, the CTA and Metra Alternative Analysis model, applies a 15% discount to travel time prior to path-building. The travel time discount is carried forward into mode choice via skims. In these cases, the value of the discount was determined by trial-and-error until the model results matched observed values. In a third case, an analysis of revealed and stated preference survey data in Germany quantified an IVTT coefficient for rail that was 25% lower than the in-vehicle time coefficient for bus. Two case studies involve incorporating a new variable in the mode choice utility to handle unmeasured attributes. For the CTA and Metra Alternative Analysis, the mode choice utility included the walk-ability factor. The same variable was also added to the auto-ownership model. The Lower Manhattan-Jamaica/JFK Transportation Project estimated a seating availability variable to be included in the commuter rail utility. To improve the model’s ability to estimate riders bound for the CBD, the Charlotte New Starts Travel Demand Model applied shadow prices to park-ride lots served by express buses (Woodford 2007). The shadow prices ranged from 3 to 9 minutes. The shadow prices were determined by calibrating the shadow price until the model results matched observed values. Finally, TCRP Report 118 suggests applying a percentage-based ridership bonus for BRT services of up to 25% (after accounting for time, frequency, and cost) to account for the perceived benefits of various BRT amenities. Case Studies of Model Applications The major catalyst for incorporating non-traditional attributes or mode-specific constants/elasticities occurs when model validation efforts using similar weights and traditional

A-42 Characteristics of premium Transit Services that Affect Choice of Mode attributes for bus and premium transit are found to be insufficient. This was the reason in six case studies. Two case studies provided guidance on incremental bias constants/elasticities in recognition of evidence that capturing traditional attributes was not sufficient to estimate ridership of new premium modes. Two other case studies researched non-traditional attributes but did not incorporate them in the travel model. This section describes the case studies whose characteristics were summarized in the previous sections. The first two studies involve applied research into the impact of unmeasured attributes on premium modes ridership. These two studies are guidance documents rather that specific travel models. The last five case studies examine existing practitioner applications of unmeasured attributes. Each involves a series of adjustments that attempt to more fully represent the demand for transit services. Modeling Guidance on New/Small Starts Policies and Procedure—FTA (2007) For many years the FTA has allowed metropolitan areas with existing rail systems to adjust their travel models to replicate ridership patterns on fixed-guideway and conventional bus systems. Even in cases where models were carefully developed with accurate representations of total travel and transit times, models often required adjustment to properly forecast rail and bus ridership. Frequently, these adjustments have taken the form of mode-specific constants that favor rail modes over competing bus modes. Prior to 2007, this adjustment was not allowed for cities where fixed-guideway transit did not currently exist. In 2007, the FTA implemented a policy that quantifies the credit that can be applied to new fixed-guideway projects applying for Section 5309 federal funding in cities without existing fixed-guideway transit. This credit is applied post-mode choice, meaning that the credit counts toward the computation of user benefits, but NOT toward ridership on the line. The credits are applied in two ways: by favoring fixed-guideway modes over conventional bus using an FTA-specified mode-specific constant and by discounting the perceived IVTT of the project. In both cases, the credits are applied if certain attributes are expected to be part of the proposed system. The credits are divided into three categories: guideway-like characteristics, span of good service, and passenger amenities. Credits are assigned based on the extent of these attributes and are expressed in terms of transit IVTT (except for the IVTT discount). TABLE A-18 summarizes the FTA’s proposals. The specific values assigned to the proposed service depend on its characteristics. These criteria allow a maximum of 15 minutes of equivalent transit IVTT for trips using park/ride access with no dependence on local bus (i.e., no local bus in-vehicle time) and a 20% discount on the IVTT of the proposed service. The proposed guidance for New Starts/Small Starts policies released in February 2007 state that the maximum values are applied only for drive access trips, without local bus in the path. Therefore, direct walk to premium paths are not eligible for this credit. However, the latest guidance, the FY2010 Reporting Instructions, mention that it is providing case-by-case technical assistance to apply the credit. So there may be a case out there that does allow credit for walk-only/non-local bus-only trips, but our experience is that this only applies to drive access. The minimum credit allowed is 5 minutes of equivalent transit IVTT and no IVTT discount.

Literature and practice Reviews A-43 TABLE A-18. Summary of unmeasured attribute credit. Category Characteristic Description Credits Allowed Guideway-like characteristics Reliability of vehicle arrival Depending on the extent that the vehicle right-of-way is grade-separated and the extent of traffic signal priority or pre-emption along portions of the alignment that are controlled by traffic signals Up to 4 minutes for trips using park/ride access with no dependence on local bus; Up to 2 minutes for all other trips using the proposed project Branding/visibility/ learn-ability Depending on the extent that stations, vehicles, and right- of-way are distinctive, and the system is easy to use Up to 2 minutes for trips using park/ride access with no dependence on local bus; Up to 1 minute for all other trips using the proposed project Schedule-free service Depending on the extent to which service headways are less than 10 minutes in the peak period and less than 15 minutes during the off-peak Up to 2 minutes for trips using park/ride access with no dependence on local bus Span of good service Hours of frequent service Depending on the extent to which weekday service extends beyond the peak period with headways that are less than 30 minutes Up to 3 minutes for trips using park/ride access with no dependence on local bus Passenger amenities Stations/stops Depending on the extent to which these have passenger amenities that relate to safety and security features, protection from the weather, retail activities, comfort, and other features valued by users Up to 3 minutes for trips using park/ride access with no dependence on local bus; Up to 2 minutes for all other trips using the proposed project Dynamic schedule information Depending on the provision of real-time information on vehicle arrivals at stations Up to 1 minute for trips using park/ride access with no dependence on local bus Vehicle amenities Depending on factors such as comfort, and the probability of getting a seat on the proposed service Discount on the weight applied to time spent on the proposed transit service up to 20% Source: FTA, 2007 TCRP Report 118: Bus Rapid Transit Practitioner’s Guide (Kittleson et al. 2007) Chapter 3 of TCRP Report 118 discusses methods to account for BRT characteristics that are not represented by travel time, service frequency, and cost variables. The report presents evidence from several research investigations that suggest that, similar to rail, BRT attracts more ridership than would be explained solely by improvements to frequency and running time. Further, the research suggests that factors attracting additional ridership include identity, passenger information, span of service, and other amenities.

A-44 Characteristics of premium Transit Services that Affect Choice of Mode Consequently, the report recommends forecasting base ridership using existing models or elasticity techniques to reflect the impact of improvements to time, frequency, and cost. TCRP Report 118 recommends increasing the estimate of base ridership by up to 25% to account for non-time/cost attributes. The specific computation is warranted using the attributes of the BRT system to determine whether all or part of the maximum 25% BRT bonus is applied. This is done by computing a score of 0–100 based on the presence or absence of specific BRT components. Running ways contribute up to 20% of incremental bias. Station amenities, vehicle attributes, and service patterns contribute up to 15% each. ITS applications and BRT branding can provide up to 10% each. If the combined effect of the attributes equals 60% or more then, an additional 15% is used to account for the perceived synergy of the components. The final percentage is used to compute the amount of the 25% ridership bonus that is warranted. A list of the components and required characteristics in order to qualify for incremental bias are shown in TABLE A-19. As an example, a fully graded separated busway (20%) with full station amenities (15%), high-quality vehicles (15%), high-quality service (15%), full ITS (10%), and full BRT (10%) branding would earn a total of 85% plus 15% for synergy. The total score would equal 100%, indicating that project ridership would likely be 25% higher than that predicted from time and cost impacts alone. A BRT operating in mixed traffic (0%) with unique, illuminated stations (4%), unique vehicles (5%), all-day, high-frequency service (8%), full passenger information (10%), and BRT branding (10%) would earn a total of 37%, and not obtain any bonus for synergy. Project ridership would likely be 9% (0.25 x 0.37) higher than that predicted from time and cost impacts alone. For an analysis of three BRT corridors, New York City Transit analyzed methods to estimate induced riders that would utilize the BRT system over and above the operational and service improvements (AECOM 2007). The team based their methodology on the report “Additional Rapid Transit Ridership Impacts,” by Herb Levinson (later incorporated into TCRP Report 118).

Literature and practice Reviews A-45 TABLE A-19. Additional ridership impacts of selected BRT components. Component Characteristic Percentage Running ways (max. 20%; not additive) Grade-separated busways (special right-of-way) 20 At-grade busways 15 Median arterial busways 10 All-day bus lanes (specially delineated) 5 Peak-hour bus lanes or mixed traffic lanes — Stations (max. 15%; additive) Conventional shelter — Unique/attractively designed shelter 2 Illumination 2 Telephone/security phones 3 Climate-controlled waiting area 3 Passenger amenities 3 Passenger services 2 Vehicles (max. 15%; additive) Conventional vehicles — Uniquely designed vehicles (external) 5 Air conditioning — Wide multi-door configuration 5 Level boarding (low-floor or high-platform) 5 Service patterns (max. 15%; additive) All-day service span 4 High-frequency service (10 min or less) 4 Clear, single, service pattern 4 Off-vehicle fare collection 3 ITS applications (max. 10%; additive) Passenger information at stops 7 Passenger information on vehicles 3 BRT branding (max. 10%; additive) Vehicles and stations 7 Brochures/schedules 3 Subtotal (maximum of 85) 85 Synergy (applies only to at least 60 points) 15 Total 100 Source: Kittleson et al., 2007 Chicago Transit Authority Circle Line Alternatives Analysis and Metra New Starts Corridor Alternatives Analyses—Chicago (AECOM 2006a) The CTA and Metra developed a New Starts travel demand model that accounts for non- traditional attributes in two ways. First, the attractiveness of rail is accounted for by applying a 15% discount to all CTA and Metra rail running times. This value was calibrated to reflect observed rail ridership patterns.

A-46 Characteristics of premium Transit Services that Affect Choice of Mode Also, the model includes a way to differentiate areas with different levels of walk-to- transit accessibility without the need for a constant based on geography or area type. The Pedestrian Environmental Factor (PEF) is defined as the number of census blocks in a quarter section (one-fourth of a square mile). A higher density of census blocks implies a more regular street network and more local streets. The highest PEFs occur in downtown locations. The factor was designed to have a range of 0–64. The PEFs are used in the auto-ownership model as well as transit path-building and mode choice steps. In the auto-ownership model, the PEF is a variable in the utility equation. The coefficient ranges between 0.018 and 0.435 depending on the number of adults and workers per household. The PEF coefficient is highest for the zero-car household utilities and lowest for the two-car household utilities. TABLE A-20 summarizes their values. TABLE A-20. PEF coefficients—household vehicle ownership model. Adults per Household Workers per Household Utility Zero- Car One- Car Two- Car 1 0 0.075 0.018 -- 1 1 0.075 0.018 -- 2 0 0.435 0.140 0.064 2 1 0.435 0.140 0.064 2 2 0.435 0.140 0.064 3+ 0 0.400 0.102 0.057 3+ 1 0.400 0.102 0.057 3+ 2 0.400 0.102 0.057 3+ 3+ 0.400 0.102 0.057 Source: AECOM, 2006a For transit path-building, the perceived walk times are factored according to the PEF. This reflects the fact that walking is generally easier in urban environments and more onerous in less pedestrian-friendly environments. The walk time factor is lowest for PEFs greater than 50.0 and highest for PEFs less than 30. TABLE A-21 details the walk time adjustments according to each zone’s PEF. These adjusted walk times are used during transit path-building and skimming. TABLE A-21. Walk time adjustments by PEF value. Pedestrian Environment Factor Walk Time Adjustment Low Value High Value 0.0 30.0 3.0 x actual time 30.0 50.0 1.5-3.0 x actual time, linearly interpolated between 3.0 and 1.5 50.0 64.0 1.5 x actual time Source: AECOM, 2006a

Literature and practice Reviews A-47 Use of the PEF in mode choice is intended to capture the impact of different area types on the likelihood of using transit. The PEF is expressed as a utility variable in the mode choice model. The walk-transit variable reflects the fact that the highest walk-transit shares occur in locations where both trip ends are in walk-accessible areas. The walk-transit utility equations are: )40(*25.0),30(ProductionPEF*2.0min1 1 PEFAttractione productionPEF 40(*25.0),30(ProductionPEF*2.0max1 1 PEFAttractione attractionPEF 13*0.2),(*5.0min attractionPEFproductionPEFyWalkUtilitPEF The drive-transit variable reflects the fact that the highest proportion of these trips occur in locations where the attraction end is highly walk-accessible. The drive-transit utility equations are: )40*(25.01 1 PEFAttractione onitAttractiDriveTransPEF 13*)0.2,*5.0min( ctionansitAttraPEFDriveTrtyDriveUtiliPEF Discounted Travel Time Coefficients—Denver and New York City Several cities use an IVTT coefficient for premium modes lower than the coefficient for conventional bus transit. Two cities are briefly highlighted here. The Denver Regional Transportation District’s initial calibration of the mode choice model for the West Corridor light rail transit (LRT) project (Woodford 2007) overestimated the number of bus riders while simultaneously underestimating LRT riders. A discount to the LRT IVTT coefficient was calibrated to observed rail ridership patterns, resulting in a final discount value of 30%. In its Regional Travel Forecasting Model (RTFM), New York’s MTA (AECOM 2006b) uses an IVTT coefficient for commuter rail discounted relative to bus and subway modes. The commuter rail IVTT coefficient was calibrated to observed ridership, resulting in a value 25% less than bus and subway IVTT. TABLE A-22 shows the IVTT values used in Denver and New York City. TABLE A-22. Premium IVTT and wait time coefficient values. Agency Conventional Mode Coefficients Premium Mode Coefficients Mode(s) Value Mode(s) Value Denver Regional Transportation District Bus -0.025 Light rail, future commercial rail extensions -0.0175 New York MTA Bus, subway -0.04306 Commuter rail -0.03222 Source: AECOM, 2006b

A-48 Characteristics of premium Transit Services that Affect Choice of Mode Charlotte New Starts Travel Demand Model—Charlotte, North Carolina (Woodford, 2007) The Charlotte Area Transit System (CATS) evaluated their regional travel model prior to analyzing five transit corridors. During the initial calibration, it was found that the model overestimated park-ride trips on local buses and underestimated park-ride trips on CBD-destined express buses. This matched an overestimation of park-ride trips using shared commercial or church parking facilities, serviced by local buses, and an underestimation of formal CATS lot usage served by express buses. Shadow prices were applied to improve the model’s ability to estimate the number of riders with a CBD destination. Formal CATS park-ride lots received no shadow price. Informal park-ride lots received between 3 and 9 minutes of equivalent IVTT based on the number of parking spaces. These values were determined by calibrating the shadows price until the model results matched park-ride vehicle counts. TABLE A-23 summarizes the shadow prices applied to each lot type. TABLE A-23. Shadow prices applied to park-ride lot type. Lot Type Number of spaces Shadow Price (in equivalent transit IVTT) Formal CATS lot Any No shadow price Informal lot <20 9 minutes 20-70 6 minutes 70+ 3 minutes Source: Woodford, 2007 Southeast Florida Regional Planning Model Version 6.5 Update—Miami/Ft. Lauderdale/ West Palm Beach, Florida (AECOM 2008b) The Southeast Florida Regional Planning Model (SERPM) mode choice model includes mode-specific constants for its two rail services, Metrorail and Tri-Rail. Metrorail is a 22-mile, 22-station heavy rail system in Miami-Dade County. Tri-Rail is a 71-mile, 18-station commuter rail system connecting Palm Beach, Broward, and Miami-Dade counties. The same IVTT coefficient (ranging from -0.150 to -0.200 by trip purpose) is applied to all transit modes. Separate incremental bias constants are applied to Metrorail and Tri-Rail paths, representing the perceived difference in unmeasured attributes between conventional bus and the two premium modes. The incremental bias constants are adjusted until the model results match observed ridership. Similar setups are used in many U.S. cities with premium transit service. The latest calibration effort (for the base year 2000) produced constants ranging in value from 21.8 minutes to 51.5 minutes of equivalent IVTT for Metrorail and 3.9 to 71.5 minutes for Tri-Rail. The full listing of the mode-specific constants is shown in TABLE A-24.

Literature and practice Reviews A-49 TABLE A-24. SERPM v6.5 incremental mode-specific constants (equivalent IVTT). Purpose/Period Mode-Specific Constants Metrorail Tri-Rail Home-based work peak 21.8 3.9 Home-based other peak 29.4 24.6 Non-home-based peak 30.8 37.2 Home-based work off-peak 36.4 13.6 Home-based other off-peak 51.5 71.5 Non-home-based off-peak 25.6 68.6 Source: AECOM, 2008b Lower Manhattan-Jamaica/JFK Transportation Project—New York City The New York MTA recently conducted an analysis of alternatives that would provide a one-seat connection between Lower Manhattan and the Jamaica station. Jamaica is an important station as it serves as the major hub for the Long Island Railroad, connecting Manhattan, Queens (including JFK International Airport) and Long Island. Jamaica station is also an important intermodal station for the New York City Transit system. Jamaica station connects the eastern end of this subway system with local and express buses from Eastern Queens and Nassau County. A stated preference survey was conducted of travelers between Lower Manhattan and Jamaica/JFK International Airport in the fall of 2005. The purposes of the survey were to verify the values of the core variables of MTA’s RTFM, and identify key modal attributes that are likely to materially affect travel behavior in the corridor depending on mode selection (commuter rail vs. subway). Such key attributes include seating availability, transfers, seat type, and fares. The survey asked travelers about their preferences for seating availability and other variables such as transfer and seating types. Seating availability on commuter rail trips of at least 15 minutes was found to be the only statistically significantly factor. Transfer type (whether the transfer took place on the same platform or a different platform) and seating type (whether a bench or forward/reverse) were both found to be statistically insignificant. TABLE A-25 shows the estimated model coefficient, standard error, and t-statistic for seating availability. The coefficient for this variable was estimated for all trip purposes. TABLE A-25. Estimated seating availability coefficients. Commuter Rail Trip Length Between 15 Minutes and 30 Minutes At least 30 Minutes Estimated coefficient 0.4931 0.8332 Error 0.124 0.107 T-statistic 4.0 7.8 Equivalent minutes of IVTT (estimated) 21.9 37.0 Equivalent minutes of IVTT (RTFM) 15.3 25.9

A-50 Characteristics of premium Transit Services that Affect Choice of Mode On the basis of these estimates, seating availability is equivalent to 21.9 and 37.0 minutes of IVTT for 15–30 and 30+ minute commuter rail trips, respectively. If the standard RTFM commuter rail IVTT coefficient is used, seating availability is equivalent to 15.3 and 25.9 minutes. Although recognized as an important characteristic, seating availability was not incor- porated into the mode choice model. One problem was that the medium- and long-trip seating availability coefficients would overestimate commuter rail trips for those lengths. Instead, a con- tinuous variable was preferred, and work was initiated for its development. Another problem was that new software procedures would be needed to estimate the number of seats available for each train at each station. It was decided to test the sensitivity of a continuous seating availability variable before beginning work on the new procedures. Ultimately, the client advanced the project schedule, and was unable to incorporate seating availability into the RTFM. Chicago Transit Authority Smart Card Activity Analysis—Chicago (Mojica 2008a and 2008b) As part of a research collaborative effort with the Massachusetts Institute of Technology, the CTA analyzed changes in travel behavior before and during planned station closings made necessary by the Brown Line Capacity Expansion Project. These changes in the modal shift from rail were revealed by reviewing Smart Card activity and boarding counts. The analysis was able to quantify the trade-offs in in-vehicle and wait time for rail and bus modes. It showed that 1 minute of in-vehicle time in a bus was worth 1.7 minutes in a train, and that 1 minute of wait time for bus was equal to 1.6 minutes of rail wait time. The data were used to develop a binary choice model. The estimated coefficients also showed major differences between bus and rail wait and in-vehicle time, with the rail coefficients 34% to 42% lower than the corresponding bus coefficients (TABLE A-26). Note that the bus and rail wait coefficients were not found to be significantly different from zero. TABLE A-26. Estimated bus and rail time coefficients (binary choice models). Variable Coefficients Relative DifferenceBus Rail In-vehicle time -0.053 -0.031 -42% Wait time -0.133 -0.088 -34% Source: Mojica, 2008a and 2008b Before-and-After Mode Choice Analysis—Dresden, Germany For some time there has been an assumed “rail bias,” that there exists an inherent superi- ority in rail-based over bus-based public transportation alternatives. Axhausen et al. (2001) de- scribe a study done to measure preference of rail over bus, expressed in LOS coefficients, which was done in Dresden, Germany, when a planned replacement of a tram line with bus service was being considered. Choice models (auto, bus, and rail) were estimated using RP and SP data collected before and after the service change. RP data was collected in a one-day travel diary.

Literature and Practice Reviews A-51 Two SP experiments were conducted: (1) inter-mode choice between auto and transit and (2) intra-mode choice within transit, between rail and bus, in terms of differences in levels of service, such as comfort, travel time, and transfer trade-offs. Results of this study showed that there is a small but consistent preference for the rail- based transit, with a lower disutility of IVTT and higher valuation of new and improved vehicles (TABLE A-27). The authors did note that the preference for rail-based transit was stronger for more frequent public transit users. TABLE A-27. Joint RP and SP estimation results. Source: Axhausen et al., 2001

A-52 Characteristics of premium Transit Services that Affect Choice of Mode Transit Agency Interviews One of the objectives of this research project was to discuss the results of the literature and practice reviews to get direct feedback from staff at transit agencies and MPOs. Two rounds of interviews were conducted by Greg Spitz of Resource Systems Group, Inc., during the summer and early fall of 2008. The first round was done to obtain descriptions from various transit agencies and MPOs of methods they are using to perform their premium transit studies, reasons behind those methods, how they are approaching their modeling and forecasting with respect to assigning values to non-traditional attributes, and feedback on the results of the literature and practice reviews. The second round of interviews was conducted later to see if any collaboration would be possible between the needs of this research study and other ongoing survey efforts the transit agencies and MPOs are involved in. This section presents the efforts that were made in contacting transit agency/MPO staff and the outcomes of those efforts. Round 1: Post Literature and Practice Reviews Interviews Once the literature and practice reviews were complete, select staff at various transit agencies were contacted to obtain the perspective of agencies operating and forecasting transit on an everyday basis. Greg Spitz conducted the agency staff interviews during August 2008 with Christopher Chesnut of UTA (Utah Transit Authority), Jeff Busby of CTA (Chicago Transit Authority), Tom Marchwinski of New Jersey TRANSIT, and Rob Alexander of GRTA (Georgia Regional Transportation Authority), and Joe Barr of NYC DOT (New York City DOT). Most agencies seemed very interested in the TCRP Project H-37 topic and were eager to cooperate, and some also participated as members of the project panel. The goals of the first round of interviews were: To hear about recent experiences agencies have had upgrading transit services and implementing premium services To obtain anecdotes to help compare and contrast real-world experiences of transit agencies to that of the literature To get feedback on results of literature and practice reviews • • • Christopher Chesnut, Utah Transit Authority UTA upgraded a local bus to a bus rapid transit in the 3500 South corridor. The local bus had headways of 30 minutes and 2,200 boardings, then increased service to 15-minute headways and 3,200–3,600 boardings; then the service was upgraded to BRT with 4,500 boardings. The service has been open 4 months. There were no changes to parking. The MAX BRT service has different vehicles, with Type 3 doors, different wrap, seating arrangements, etc. The service offers off-board payment, POP system, limited stops roughly every half-mile, and BRT machines at every stop with the same fare. You can board and alight at any door. There is signal priority in the corridor and travel times are roughly 10%–15% faster without the new lanes, while the 3500 South is under construction. There are better covered shelters.

Literature and practice Reviews A-53 There were advertisements at the MAX stations prior to implementation and there were people at each station helping customers for the first few days of operations. The BRT service was branded as “MAX” and wrapped in a logo. Several technologies changed with this service: fare collection, vehicles, and signal priority. When construction is complete, MAX will have its own permanent lane separated. The new lane is expected to raise awareness about the service because it will be 2 miles of exclusive right-of-way in a 10-mile corridor in the most congested area. Bus stops have special signs; there is all MAX marketing throughout the corridor. The new vehicles have nicer windows, are quieter and cleaner, and the boarding process is faster and easier than with local bus vehicles, but there are fewer seats and people are expected to stand. Jeff Busby, General Manager, Strategic Planning, Chicago Transit Authority CTA implemented “X” buses (limited stop express buses) on long corridors without parallel rail service. Bus X49 for example has limited stops every half-mile vs. a typical local bus of 1/8-mile stops. Besides fewer stops farther apart, there is very little difference between the X buses and the local service. Some marketing was done for the X buses on CTA local buses that parallel the new X bus services and on trains all over the city. The trains advertised the links between X buses and the “L” trains. (e.g., connecting to the Blue Line from the Irving Park X80 bus). There were no advertising purchases—very low cost marketing was all that was done on-board CTA vehicles. On the system map, the X buses are indicated slightly differently on the map; X bus stops are actually shown on the map, which is not the case for local bus stops. The X buses have no special branding or “wrap” of any sort. The one distinguishing characteristic is that the destination sign on the bus itself is black on yellow vs. yellow on black for locals, so the X bus can be distinguished by passengers by sight based on this difference, though it is a small one. The X buses also provide more information at the bus stop: there is a drum on the pole with all the stops listed (like a train diagram) and better schedule information. X bus vehicles and technologies are the same as local buses; the only difference is the service with the limited stops, as described above. In the future, CTA is going to implement BRT in some corridors, but the X bus is simply a slightly enhanced limited-stop bus service with no technology or significant structural changes over local buses. As for awareness, transit riders do know about the X bus service and seem very positive toward it. For example, anecdotal observations show that riders will go out of their way to get to the X bus corridor to get to the Blue L line. Conversely, non-riders don’t seem any more aware of the CTA transit service versus prior to the implementation of X bus service (there being no way they would know anything is different). Ridership in X bus corridors has gone up over and above general trends, indicating additional riders/trips. Recently, CTA added revenue hours/frequency over what had existed previously (after X buses had already been implemented). When accounting for this additional

A-54 Characteristics of premium Transit Services that Affect Choice of Mode service, X buses still increased ridership overall in its corridors (including X buses and local bus ridership). Customer satisfaction also went up for corridors with X bus service. One challenge CTA continues to work on is finding the optimal mix between local bus and X bus service. Fifty percent express trips seems to be the minimum ratio required to make an express bus system work in parallel with local service. This is a minimum, as it appears to be more effective if a percentage above 50% express is implemented. However, the express service ratio can only get so high. While express bus works especially well in the rush hour, it can present problems in the middle of the day for those making shorter/more local trips. Too much express service means that local buses can become very crowded in the middle of the day for shorter trip purposes such as shopping. Transit consideration question: increasing party size is clearly a major deterrent for transit, as are complicated trip tours: the higher the party size and more complicated the tour, the less desirable transit becomes. CTA has studied the issue of baggage to understand how that affects transit usage. CTA found that over 50% of their customers are “encumbered” with baggage of some sort, meaning the baggage required customers to use two hands. So, clearly people are using transit with baggage, but probably no more than they can carry themselves in two hands. Finally, another interesting piece of CTA research Mr. Busby mentioned was their ability to track Chicago Card customers’ revealed behavior anonymously after the closure of a number of Brown Line stations due to a major construction renovation. What they found was that there was premium for rail—customers would walk further to it—and that it appears that IVTT on bus was perceived much higher (or possibly customers were unaware of the bus). Rail waiting time was also revealed to be less onerous vs. bus wait times. Tom Marchwinski, Direct of Forecasting, New Jersey TRANSIT New Jersey TRANSIT implemented the River Line LRT service in March 2004. It is the first US diesel-powered DMU LRT and runs 33 miles connecting southern New Jersey to Camden and Trenton and to the Northeast Corridor line to Philadelphia via PATCO service. Most of the corridor had parallel bus lines, and most of these buses are still running, but these bus services were cut back significantly and just run in the peak hours. As noted, the River Line is an LRT service that replaced bus service, meaning travel time is now faster but with nice stations that are spaced farther apart than the previous bus stops. The service is more permanent and includes many new park/ride facilities. The LRT service, while premium, was priced the same as bus or even less in some cases to build ridership. The service includes stations with ticket-vending machines (TVM), phones, a PA system, digital signs to show delays and alerts, platforms, and full signage. Service only runs to about 9:30 p.m. on weekdays due to freight conflicts. Buses run in the later hours (the bus/LRT passes are interchangeable). Upon opening and for the first year there was a major marketing campaign: web sites, newspaper ads, and brochures. There were no TV or radio buys, however. Service was advertised on all Northeast Corridor timetables as it connects to the NEC. This included connection times

Literature and practice Reviews A-55 and a small map. Websites, which still continue (http://www.riverline.com/), were developed with destinations and promotions (e.g., to an aquarium, entertainment center for concerts etc.) Lots of weekend riders were generated due to these attractions and these riders continue to be strong. The service had a logo and was branded: River Line. Some aspects of the branding/look of the service have changed slightly over time, but there have always been logos and nicely painted vehicles. The technology is a first in the US: a self-powered diesel LRT (DMU). The entire line but for the last mile in Camden is exclusive ROW. The line has gates at rail crossings, and the train has many crossings on its ROW. There were also a few new tracks laid for sidings, but most of the ROW was already there. The River Line does mix with traffic in a couple places and there are new tracks in streets in Camden. A major transfer point between the River Line and PATCO was upgraded with a nice walkable and pedestrian-friendly plaza between them as well as facility to transfer to buses. Awareness of the system was increased over previous buses without a doubt. Surveys showed that, after service opened, over 15% of people were riding just to check it out! People just wanted to see it and knew it was there. This curious ridership went down over time, but there was significant awareness of the system as it was the first new rail service in southern New Jersey. Ridership increased for sure. Twenty-five percent came from bus. The rest are new riders from auto trips (50%), 5% were going from NEC rail to PATCO, and the rest of the ridership came from induced new trips. The corridor ridership is up overall, including buses. Ridership met forecasts and grew for the first 3 years then leveled. Now it is going up again due to gas prices, etc. The system is successful but cost a LOT to build ($800 million) and carries 9,000 trips per day. There is still not enough rolling stock in peak, and trains can therefore be packed. There was a LOT of awareness of the project: partly due to ads and partly due to the costs, objections to the line (NIMBYs), etc. Big projects mean people notice them. Grade crossings also mean people notice the system: there are more than 45 on the line. Due to the number of grade crossings, a lot of safety trainings were conducted in schools, which also increased awareness. People are therefore definitely more aware of the service. In fact, the NJ network (public TV) has a picture of the River Line coming through when advertising their TV station identification. There are lots of bikes on the River Line due to an innovative on-board bike hanging system. There are lots more local trips versus commute trips compared to what was expected. Commute trips are still the majority however. Rob Alexander, Georgia Regional Transportation Authority GRTA operates commuter coach (cruiser) buses from suburban Atlanta to the center city and transfer locations with MARTA, the urban transit operator in Atlanta. The service is called Xpress. It is relatively new (within the last 5 years), and no other transit service preceded on all

A-56 Characteristics of premium Transit Services that Affect Choice of Mode routes (except for one). The service began with just two routes and now is up to 27; demand remains very strong; they need more funding and more buses, as they cannot keep up with ridership. As mentioned, there was no preceding service to these routes; however, the service is definitely premium. The coaches are nicely maintained with typical “luxury” coach amenities: reading lights, reclining seats, overhead racks, etc. There has been no purchased advertising except for some small buys in local papers. The fact that people can see the coaches on the road is what gets people seeing them, and the website/phone number is on the sides of the buses. The buses are their “billboards.” The bus service is very customer-service focused with lots of communications with customers via email, the web, and phone. All emails are answered individually; all calls are taken by a person. Delays are put out via email if possible. Newsletters and blog articles are distributed to over 4,000 customers via email and web. GRTA customer-service answers about 500 to 700 customer queries per month. They encourage customers to express their needs. Service went from no prior service to 57-seat MCI coaches/cruisers; premium buses, reclining seats, ac/overhead bins, etc. This is new to metro Atlanta. Half of GRTA customers had never used transit, the other half are experienced in transit from other parts of the country and have higher expectations. The service collects people at P&R lots. Funding from state and counties for maintaining and adding service is an ongoing issue. Buses run in regular traffic with no special treatment of any kind. People become aware of the service by seeing the coaches in traffic and through word-of-mouth from other customers. GRTA does go to some commuter fairs and such events at major employment centers to increase awareness. They have grown to 27 routes from two and are booked solid and having a hard time keeping up with demand. They want to buy another 28 buses and eventually another 32 on top of that for 60 new buses. They are struggling to find the money from the state and local governments. They are trying hard to keep up with the regular demands of the system as noted. Therefore, services such as Wi-Fi are something they would like to do but cannot start until they have their main service goals under control: serving the demand. They would like to set up a Twitter system for delays as well to broadcast information to their riders. Again, these additional attributes are taking a back seat to serving the growing demand. Joseph Barr; Deputy Director, Policy Technology & Management Analysis; New York City Department of Transportation NYCDOT has established two high-profile BRT/Enhanced bus projects. The interview conducted by RSG focused mostly on the Fordham Road BRT, or the “Select Bus Service (SBS)” as it is branded by MTA NYCT. This SBS service is a BRT: the bus lanes are painted a separate color from the regular street, with large signs declaring the lanes as bus lanes. Stops have been reduced slightly to every

Literature and practice Reviews A-57 half-mile over the previous limited service. All ticketing is done off-board the buses at TVMs at the bus stops. TVMs take Metrocards or cash but cannot issue Metrocards. The SBS buses are “wrapped” with a brand logo. There are no ads inside or outside the buses currently, though that may change due to MTA’s financial situation. To promote the service, there was some small advertising done in local papers, but not much. The service has had a fair bit of mostly positive press (not press releases). MTA and NYC DOT also direct-mailed an informational brochure in the SBS corridor, but this was not so much advertising as information. Finally, they placed “customer ambassadors” out at stops in the first couple of weeks to help people adapt to the new system. The buses are equipped with signal priority, off-board fare collection, and on-board bus cameras (though Mr. Barr felt most customers don’t realize there are cameras). The buses are the same type as they always have been, but were thoroughly rehabilitated and cleaned for the new service. New bus shelters are coming to the service soon, and they have better visibility to make people more secure. They are new and larger, too. Buses are told not to wait at time points anymore and to drive the route as fast as possible. Schedules are just headways during different hours of the day (e.g., the bus comes every 5 minutes between 6:00 a.m. and 9:00 a.m.) without time points. Due to these changes, travel time on this route is 14% to 24% faster than the old, limited service. Awareness of the bus corridor and its high priority has definitely increased. This is likely due to the painted lanes (which are hard to miss) and the fact some on-street parking was taken away (which makes non-riders aware that something has changed). Overall ridership is up in the corridor, with a 20% increase in ridership on the SBS over the former limited route (this is a much higher increase than what MTA buses have experienced), though it is not yet clear how much of this increase consists of local riders who switched to SBS. Round 2: Contacts on Potential Collaboration with Other Survey Efforts The second round of interviews took place early in Fall 2008 in anticipation of the start of planning for the research approach. The early start was made to ensure that the project’s survey research could realistically be incorporated into other agencies’ research efforts. The effort therefore required long lead times. The objective was to determine the possibility of collaborating with agencies’ ongoing survey efforts. Agencies were contacted that had received other FTA survey funding. All agencies contacted were interested in the TCRP Project H-37 research, and all were willing to consider ways to cooperate. Agencies contacted included: L.A. Metro Denver RTD Boston MBTA/CTPS New Jersey Transit New York MTA Chicago Pace Georgia Regional Transit Authority Portland Metro • • • • • • • •

A-58 Characteristics of premium Transit Services that Affect Choice of Mode Greg Spitz, RSG, contacted a number of MPO’s and transit agencies based on guidance from the panel and FTA about which agencies would be conducting rider research in the near future. Mr. Spitz also contacted other agencies of panel members, etc., to further explore other potential survey opportunities for TCRP Project H-37. An example of the email that was used (after an initial phone call) to inform transit agencies about the goals of the research and to ask for cooperation is shown in FIGURE A-19. The following paragraphs recap the notes and initial thoughts from each Round 2 interview and summarize how each agency might participate in the TCRP Project H-37 research: LAMTA: Spoke to LAMTA modelers and researchers Chaucie Chu and Robert Farley. LAMTA recently finished the Metro Rapid BRT study but they have additional funds to conduct more surveys, likely in spring 2009. We [the researchers] are working with them to incorporate at least an email question in their future surveys, if not additional conjoint questions. They are concerned about questionnaire length, and therefore want to test long/short survey versions. This likely means an email address is our best method to capture these respondents. Denver RTD: Spoke to Lee Cryer. RTD has recently done two on-board surveys. They did not collect email addresses and were hesitant to try to contact respondents post-hoc via other means (e.g., mail or telephone). They were, however, very cooperative about involving us in a future customer satisfaction study. Susan Henry is the market research person at RTD and she will be surveying about express bus service in fall 2008 or spring 2009. RSG has already sent her an email and will follow up shortly. Boston MBTA/Boston MPO: Spoke with Karl Quackenbush. They have just finished an extensive on-board survey in Spring 2008 which we attempted to be part of, but for which the survey design had already been finalized and was already considered long enough. However, they are conducting a statewide HH Diary survey which we will try and add email questions so that we can be part of that research. Columbus (MORPC): This on-board study is being conducted for MORPC by NuStats. Field will likely be in fall 2008 and we will contact MOPRC to see if we can become involved. We have not yet contacted them but will do so soon so as to try and obtain at least an email question in their questionnaire. New Jersey TRANSIT: We have contacted New Jersey TRANSIT panel member Tom Marchwinski who has been extremely cooperative. While there has been at least one localized bus study where we were considering adding a question, we decided the sample size was too small and the study specific to one area. Instead, we are working with New Jersey TRANSIT to gain permission to use their customer email database from previous broad based surveys. NYMTA: RSG conducted a major OD study for MTA/Metro-North Railroad and had a database of email addresses for these customers. We have requested permission to use these addresses for this study and are hopeful that we will obtain approval over the next month.

Literature and practice Reviews A-59 Dear [Transit Agency], At the bottom of this email is a summary of the TCRP study we’re conducting, Characteristics of Premium Transit Services that Affect Choice of Mode. I was glad to hear that H-37’s research issues resonate with some of the issues you’re confronting in your own modeling. Attached also please find a short PowerPoint deck which discusses MaxDiff research techniques. This is a potential technique we expect to apply when trying to understand and quantify some of the non- traditional attributes of premium transit (reliability, comfort, seating availability, etc.) when RSG conducts our own survey for this study (hopefully with your customers and/or potential customers). We very much appreciate your interest and willingness to cooperate in this effort. As we discussed today, probably the best way to work together on this would be to ask respondents of your upcoming XXXX survey if they would be willing to participate in future research and, if yes, to provide their email address. Your agency would then approve and monitor any subsequent questionnaire RSG might use to survey your customers, but we would take care of all the work and make it as easy and painless for you as possible. In addition, your agency would have full access to the data generated from your customers and possibly learn some new research techniques, etc. I will follow up in the next few weeks and feel free to contact me with any questions or comments. Thanks again for your help and interest. Greg ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ Greg Spitz Director Resource Systems Group Inc. 55 Railroad Row, White River Junction, Vermont 05001 TEL 802.295.4999 ext. 142 FAX 802.295.1006 www.rsginc.com ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ Note: Text is generic, but was personalized for each agency contacted. FIGURE A-19. Email example.

A-60 Characteristics of premium Transit Services that Affect Choice of Mode Summary Scope of the TCRP H37 Study The purpose of this research is to describe the most important factors that differentiate premium transit services from ordinary bus services, and to quantify for practical use the magnitude of these distinguishing features. A successful research effort will both improve the transit industry’s understanding of mode choice determinants and offer practical insights to the forecasting community so that mode choice models and transit path-builders can better represent and distinguish important mode characteristics. Understanding and modeling the real drivers and factors determining travel behavior and eliminating “flat” constants-based model structures will significantly improve the explanatory power as well as potential transferability of travel models. This is a very ambitious and long-term task that includes numerous aspects of travel model improvement. The proposed research is intended to identify the most important “breakthrough” directions with respect to mode choice model systems and consolidate the already acquired experience. 1. Research and practice synthesis – The focus of this synthesis will be to identify (1) distinguishing features of premium transit services, (2) the factors that influence mode choice decisions, and (3) advanced methods to model mode choice relevant to this research effort. This summary will also identify factors or modeling methods that may confound our ability to interpret the mode constants because the constants are correcting for them. 2. Data collection and Analysis – Our data collection effort will, first, seek to understand the extent to which transit alternatives (premium and non-premium) are known and considered as an option. Second, we will use adaptive conjoint analysis and related market research techniques to understand the relative importance of different levels of comfort, convenience, safety and other non-traditional transit attributes in mode choice decisions. Finally, we will conduct segmentation analysis to understand how different market segments respond to these attributes. 3. Advise FTA and practitioners on bringing results into practice – This study will make recommendations on data collection and mode choice model specifications and parameters so that premium transit services can be better distinguished and transit forecasts, in turn, can be refined. Additionally, the need for consistency between choice models and transit path-builders will be addressed. Note: Text is generic, but was personalized for each agency contacted. FIGURE A-19. (Continued).

Literature and practice Reviews A-61 Chicago Pace: Chicago Pace will soon be issuing an RFP for a combined O-D and Customer Satisfaction Study. RSG has been communicating with Pace on this future project and is expected to have good cooperation on adding an email question, if not more questions related to TCRP Project H-37. Portland Metro: Portland Metro is expected to issue an RFP for a study that is very interested in understanding the modeling characteristics of premium transit and how to best characterize premium modes in forecasting. This study is the one where we hope to be able to actually contribute significantly to the questionnaire for the good of both Portland Metro and TCRP Project H-37, as our interests appear to align on this project. Tony Mendoza of Metro has been contacted and was made fully aware of TCRP Project H-37. He is very excited and eager to cooperate on this issue and he hopes to use stated preference and other conjoint techniques, which is important because that is the clearest way to quantify premium transit attributes. The RFP has still not been released, but we are in touch with Metro and will be ready to discuss this more with them once the RFP is made available. Although every agency contacted was extremely cooperative and some even said we could ask a question or two in future research, it became clear that to understand premium transit differences we needed to use a more comprehensive survey than what a question or two could provide. Therefore, the general consensus after speaking with these agencies was that the most realistic way for them to cooperate was to provide email addresses of their riders (and potential riders) to RSG so that we could survey their customers with our own survey instrument developed exclusively to address the objectives of TCRP Project H-37. One suggestion we made to the agencies was to add to agencies’ questionnaires, when feasible, a question on whether respondents have an email address and whether they are willing to participate in future research. The question looks essentially like: May we contact you for future research? Yes, email address______________________________________ No Not only did email lists make it possible to contact an agency’s customers or potential customers easily, it allowed us to send them a comprehensive survey dedicated to the needs of TCRP Project H-37. It was usually easier for the agency to provide emails than to somehow incorporate new questions into a major research initiative, which typically had its own objectives and therefore could not be easily combined without adding to already time- and space- constrained questionnaires.

A-62 Characteristics of premium Transit Services that Affect Choice of Mode Summary Based on the findings in the literature review and actions taken in practice, it is clear that typical mode choice model specifications lack important features differentiating transit services and this omission affects the quality of forecasts and our ability to represent the merits of one alternative over another. The fact that creative actions have been taken in practice underscores that this issue is not new by any means, but the increasing variety of techniques to address, to some extent, these shortcomings is encouraging. It is the goal of this research to both refine our understanding of transit mode choice behavior and refine these strategies or come up with new strategies for improving mode choice models and transit path-builders. The literature review focused on both the awareness of transit services and the features of transit services. Though the literature on transit awareness is relatively thin, a few studies have shown that simply assuming perfect knowledge of the transit system is wrong, and may result in models dramatically overstating the potential market for transit in terms of which travelers see transit as a choice in models. Overstating the potential market for transit in applied mode choice models means that calibrated constants and estimated coefficients will be biased in order to calibrate the model. Understanding this bias, particularly the differences in awareness of premium vs. non-premium transit, is a useful step in understanding transit mode choice and explaining the components of calibrated mode choice model constants. The literature review helped identify over 90 transit service attributes, as previously shown in TABLE A-1. Many transit agencies conduct periodic customer satisfaction surveys, and the surveys from four transit agencies were reviewed and cross-referenced with the attributes identified in research literature. One goal of this research was to quantify the relative importance of these attributes; however, this can be done to varying degrees for only half of the 90+ attributes, as shown in TABLE A-17. There is clearly some overlap among certain attributes, but the fact remains that extensive research does not appear to have been done covering the importance of all of these service features. Based on the research and studies reviewed for this appendix, the most important service features are identified in TABLE A-2. Notably, the non-traditional attributes that seemed to be the most important include level-of- service variables (e.g., reliability and service priority), seat availability, on-board comfort (e.g., seats, smoothness of ride), station cleanliness, and information services. In practice there have been several attempts to account for and model accurately differences in observed ridership on different modes, other than typical model specifications that include mode-specific constants and the standard variables. These techniques include asserting mode-specific constants, valuing travel time differently for different modes, and enhancing the specification to include non-traditional variables such as reliability. The goal of this research is to inform and build on these techniques.