Bus Rapid Transit Practitioner's Guide (2007)

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.

Bus Rapid Transit Practitioner’s Guide Bus Rapid Transit Practitioner’s Guide Page 3-1 Estimating BRT Ridership CHAPTER 3. ESTIMATING BRT RIDERSHIP INTRODUCTION AND SUMMARY Estimating BRT ridership is an important task. Realistic and reliable ridership forecasts are essential in sizing system design features, developing service plans, estimating capital and operating costs, performing alternatives analysis and cost- benefit comparisons, and making investment decisions. This chapter reviews current BRT ridership experience, summarizes salient ridership research efforts, describes travel demand models and elasticity methods for estimating BRT ridership, and gives guidelines for applying modal preference (bias) factors to BRT systems. Key findings and guidelines are as follows: • BRT ridership forecasts are needed for the base year, the opening year, the year when ridership reaches maturity, and a design year usually 20 years into the future. FTA’s Proposed Interim Guidance and Instructions, Small Starts Provision of the Section 5309 New Starts Program, issued June 5, 2006, suggests that opening year forecasts will be required for projects defined as Small Starts (less than $250 million total cost and $75 million federal contribution). For larger projects, both 20-year and opening-year forecasts will be required. • Ridership estimates should be provided for peak and off-peak conditions by line segment and by station boardings and alightings. • On-board travel surveys should capture key traveler information (e.g., trip origins, destinations, purposes, and frequencies and socioeconomic characteristics). This information provides an important input to various demand estimation procedures. A CBD employee survey is desirable to provide origins and travel modes for downtown workers. • Ridership can be estimated by the traditional four-step process (i.e., trip generation, trip distribution, mode choice, and trip assignment) where BRT operates on a new right-of-way (such as a busway). Household travel surveys can provide the basic information needed for modeling and analysis, but data from on-board surveys also should be gathered in order to have sufficient data representing transit users during model development. > The “pivot point” application of the incremental logit mode choice model is well-suited for estimating BRT ridership, especially when analyzing a new alignment. > Travel paths should use acceptable weights for in-vehicle and out-of- vehicle travel times. Network coding should treat BRT as a separate facility in terms of travel times and stop locations. • Travel time, service frequency, and cost elasticities can be used for smaller- scale projects where BRT would operate along existing bus routes. An on- board survey can provide information about desired travel patterns as well as demographic and socioeconomic information. Allowance should be made for “new” trips (i.e., trips diverted from automobiles, trips not Ridership forecasts are needed for BRT projects to obtain FTA New Starts and Small Starts funding. Use of existing transit rider origin- destination surveys can help determine BRT trip patterns. Elasticities can be used to estimate ridership for smaller BRT projects.

Bus Rapid Transit Practitioner’s Guide Estimating BRT Ridership Page 3-2 Bus Rapid Transit Practitioner’s Guide made before, and trips made with greater frequency). Population and employment growth should be taken into account. • BRT’s unique physical and operating features must be recognized in the travel demand estimation process. Salient studies of aggregate and disaggregate customer response to new BRT systems (or upgraded express bus service) have found the following: > The attractiveness of BRT systems, not unlike that of new rail systems, has been greater than might be expected on the basis of reductions in travel times and costs. > All things being equal (i.e., newness, component quality, system configuration and completeness in terms of all the elements of rapid transit, origin-to-destination travel times, reliability, and costs), BRT systems are likely to attract levels of ridership similar to those of rail- based systems. Studies of ridership elasticities for arterial street BRT in Boston, Los Angeles, and Vancouver (BC) indicate that actual ridership was up to about 20% more than that computed by travel time and service analysis frequencies. Accordingly, a 25% increase is a suggested upper limit for full-featured BRT above that obtained by elasticity computations. Recent practice has applied mode-specific “bias constants” equivalent to up to 15 minutes of in-vehicle travel time for rail rapid transit (i.e., for modeling purposes, the impedance for a trip using rail would be up to 15 units less than that computed using the unadjusted impedance function). The BRT (and rail) bias constants used should depend upon the quality and extent of the features available for each transit alternative that is evaluated. In this chapter, judgments were made as to the likely impacts of various BRT features on ridership. These preferences reflect the informational advantages of the unique identity of a system with simple route structure and schedules, the superior waiting and transferring environments of stations as opposed to bus stops, and the comfort/ride quality of better, more modern vehicles and exclusive transit running ways. The differences are far greater between local bus systems and generic rapid transit than they are between, for example, light rail and BRT running in the same environment with the same station and running way configurations, basic route pattern, and schedule. Therefore, the ridership forecasting approach used for one rapid transit mode should be used for the other, subject to the caveat of system comparability. The operable guidance for forecasting is “be conservative and consistent.” The sections in this chapter document the aspects of estimating ridership response to BRT features and provide guidelines for keying BRT features to ridership estimates. Examples of ridership estimates are given in Chapter 5. RIDERSHIP EXPERIENCE In the past 10 years, a large number of BRT systems have opened in the United States and Canada. Information has been assembled on ridership growth, the sources of this growth, the relevance of demand elasticities, and rider attitudes. Relevant findings follow. With similar attributes, BRT systems can attract ridership comparable to rail transit ridership. Ridership for most major transit projects, including BRT, is forecast using disaggregate choice models based on a logit function and a linear measure of (dis)utility. The utility function includes a constant, often referred to as a “bias” constant, which accounts for all of the “unmeasured attributes” that contribute to individual choice. These unmeasured attributes contribute to the perceived desirability of one mode compared to another. Thus, a quantity added to or subtracted from the bias constant may be said to reflect the unmeasured attributes of BRT.

Bus Rapid Transit Practitioner’s Guide Bus Rapid Transit Practitioner’s Guide Page 3-3 Estimating BRT Ridership Ridership Growth Ridership experience with BRT in six major urban areas is summarized in Exhibit 3-1. In most cases, corridor ridership has grown faster than the reduction in transit travel time, suggesting demand/travel time elasticities over 1.0. In other words, although increases in service frequencies contribute to the ridership gains, other factors appear to be at work as well. The data also indicate that a large portion of BRT ridership consists of new transit trips, not trips diverted from other transit routes. EXHIBIT 3-1 Ridership Experience with BRT Location % Corridor Ridership Gain Time Period Maximum % Reduction in Travel Time % BRT Ridership that is New Transit Trips Los Angeles 40 3 years 25 >30 Miami 85 5 years 30 >50 Brisbane (Australia) 60 2 years NA >45 Vancouver (BC) 30 2 years 16 >25 Boston 100 18 months 20-30 >30 Oakland 20* 1 year 17 >30* * Offset to secular decline SOURCE: CBRT (1) Expanded service and improved frequency also enhance ridership during weekends. Exhibit 3-2 shows that ridership along the South Miami-Dade busway corridor increased more than 70% on weekdays and 150% on weekends from 1996 to 2003. Most of this growth reflects improved coverage as well as the presence of the busway. EXHIBIT 3-2 Ridership Growth over Time: South Miami-Dade Busway Corridor Time Period 1st Quarter 1996 3rd Quarter 2003 % Change Average weekday 7,600 13,000 +70 Average weekend (Saturday + Sunday) 6,000 15,000 150 SOURCE: South Miami-Dade Busway Corridor Case Study (2) The same phenomenon was apparent along Boston’s Washington Street Silver Line. As shown in Exhibit 3-3, combined Saturday and Sunday traffic grew more than 90% as compared to 80% growth in weekday travel. Prior Modes Previous travel modes of BRT riders in Adelaide (Australia), Boston, Los Angeles, Oakland, Pittsburgh, and Vancouver are shown in Exhibit 3-4. According to the exhibit, new transit trips (i.e., trips made by former automobile drivers and pedestrians and by riders who did not make the trips before) represented approximately 20% to 33% of the trips in Adelaide, Boston, and Los Angeles. Former rail rapid transit riders represented 22% of the BRT riders in Boston and 13% of the BRT riders in Oakland. In Boston, the Silver Line BRT service provides more direct access to Dudley Square and downtown Boston than the Orange Line (rail). BRT systems have been observed to attract 20% to 33% new riders.

Bus Rapid Transit Practitioner’s Guide Estimating BRT Ridership Page 3-4 Bus Rapid Transit Practitioner’s Guide EXHIBIT 3-3 Boston: Washington Street Ridership Growth over Time 0 2,000 4,000 6,000 8,000 10,000 12,000 14,000 16,000 1997 2001 2002 2003 Weekday Saturday Sunday SOURCE: MBTA Silver Line (3) EXHIBIT 3-4 Prior Transportation Mode of BRT Riders % Using Prior Mode BRT System Bus Subway Drive Auto Walk Did Not Make Trip Other Adelaide (Australia) 70 — — — 24 6 Boston: Silver Line 45 22 3 — 17 13 Los Angeles: Wilshire-Whittier 67 — — — 33 — Oakland: San Pablo 55 13 19 — 9 4 Pittsburgh: East Busway Extension 82 — 7 — 11 — Pittsburgh: West Busway 56 — 34 4 — 6 Vancouver (BC): 98B 72 — 24 1 — 3 SOURCE: TCRP Project A-23A Interim Report (4), Massachusetts Bay Transportation Authority (MBTA), and Alameda-Contra Costa Transit District (AC Transit) Rider Characteristics Exhibit 3-5 and Exhibit 3-6, respectively, compare the demographic characteristics of riders on the Silver Line BRT system in Boston and the park-and- ride/transitway system in Houston with the characteristics of riders on the respective local bus systems. The characteristics of riders on the premium bus systems appear to have more in common with the general perception of the characteristics of rail transit users than with the general perception of the characteristics of local bus users. EXHIBIT 3-5 Characteristics of MBTA Silver Line Riders Customers % 1995 (Route 49) % 2003 (Silver Line) Origin in South End 29 48 Ages 18-24 3 15 Household income > $80,000 per year 8 15 SOURCE: MBTA

Bus Rapid Transit Practitioner’s Guide Bus Rapid Transit Practitioner’s Guide Page 3-5 Estimating BRT Ridership EXHIBIT 3-6 Characteristics of Houston RTCR Riders Houston METRO Service % Riders with Annual Household Income > $50,000 % Riders with Annual Household Income > $75,000 % Riders with 2 or More Household Vehicles RTCR park-and-ride services 70 50 61 Local bus 11 – 16 NOTE: RTCR = Rubber-Tired Commuter Rail (BRT) SOURCE: 2002 Houston METRO on-board survey In both areas, the improved bus service attracts more high-income riders than the corresponding local bus service. Houston’s “rubber-tired commuter rail” service attracts half its riders from households with incomes of more than $75,000; more than 60% come from households with two or more vehicles. The difference in the characteristics of those who ride the express bus compared to those who ride local buses mainly reflects the design of the service. The express bus gathers riders only in more affluent suburbs and operates in an express mode to the downtown area, with no intermediate stops in areas with lower-income populations. More than 35,000 riders use the system each weekday. Boston’s Silver Line covers the exact same area as the previous Massachusetts Bay Transportation Authority (MBTA) route 49 local service, but it does so faster and attracts riders from the Orange Line (rail rapid transit). Attitude and Preference Surveys Attitude and preferences surveys of both transit riders and non-riders regarding different BRT components can be an important input to estimating the impact of different BRT components on ridership attraction. Several transit agencies have conducted rider surveys for existing BRT services in order to design new and extended service. Other transit agencies are beginning to conduct such surveys before services are planned, designed, and implemented. The purpose of preference surveys is to identify which BRT components are most important to potential users and which would contribute the most to a decision by riders and non-riders to use such a service. It is important to administer such surveys to both riders and non-riders, as the premium attributes associated with BRT are intended to attract potential choice riders to use the service. Types of Surveys Two types of surveys could be applied in BRT planning and design: • Potential riders and non-riders could be queried on the relative importance of different BRT components before the new BRT service is implemented. • Surveys could ask riders after they have taken a new BRT service which components of the service most influenced their decision to ride the new service and what potential enhancements could be made to the service. These surveys typically use a numerical rating scale, with ratings extending from “not at all important” to “extremely important.” Some surveys have been conducted where riders rate the overall importance of different components from “excellent” to “very poor,” with the percentage rating for different attributes reported. Attitude/preference surveys can help transit agencies estimate BRT components’ impacts on ridership. Surveys can be conducted before and after BRT service is implemented.

Bus Rapid Transit Practitioner’s Guide Estimating BRT Ridership Page 3-6 Bus Rapid Transit Practitioner’s Guide Results from Past Surveys Several transit agencies have conducted attitude surveys to identify the relative importance of different BRT components. Surveys undertaken in Los Angeles, Oakland, and Washington, D.C., are highlighted in this section. Los Angeles In Los Angeles, a survey was undertaken after the initial Metro Rapid Wilshire-Whittier and Ventura BRT lines were opened to measure the importance that riders and bus operators place on various attributes of Metro Rapid service and potential enhancements to the service. Exhibit 3-7 summarizes the responses. The highlights are as follows: • Operators’ highest-rated current attributes were simple routes (9.2), traffic signal priority (8.9), and schedules (8.6). • Customers’ highest-rated current attributes were service intervals (9.8), simple routes (9.4), and time until next bus display (9.2). • Operators’ highest-rated potential enhancements were off-vehicle fare payment (9.0) and exclusive bus lanes (8.5). • Customers’ highest-rated potential enhancements were feeder network (8.7) and multiple-door entry and exit (8.5). EXHIBIT 3-7 Los Angeles: Metro Rapid Attribute Importance Ratings Attributes Operators and Customers Operators Customers Current Attributes Simple Routes 9.3 9.2 9.4 Schedules 8.8 8.6 8.9 Service Intervals 9.3 8.5 9.8 Less Frequent Stops 8.8 8.4 9.0 Level Bus Entry/Exit 8.6 8.0 9.0 Color-Coded Buses and Stations 8.3 7.5 8.8 Traffic Signal Priority 8.8 8.9 8.9 Time Until Next Bus Display 8.1 4.7 9.2 Potential Enhancements Exclusive Bus Lanes 7.9 8.5 7.7 High-Capacity Buses 7.9 7.6 8.0 Multiple-Door Entry and Exit 8.4 8.0 8.5 Off-Vehicle Fare Payment 8.4 9.0 8.3 Feeder Network 8.0 5.9 8.7 NOTE: Attribute Importance Ratings are based on a 0-10 scale with 10 = “extremely important” and 0 = “not at all important.” SOURCE: A Qualitative Study of Metro Rapid and Associated Alternatives (5) Oakland A rider survey was conducted after Oakland’s San Pablo Rapid BRT service opened. Exhibit 3-8 summarizes the ratings obtained for a variety of performance measures. The exhibit shows the percentage of respondents who rated each measure as “excellent,” “good,” “fair,” “poor,” and “very poor.” The highest rating was for the ease of identifying the right bus. Other measures that at least 75% of respondents rated as “excellent” or “good” were wheelchair securement, travel time, quality of new buses, location of bus signs, and service frequency.

Bus Rapid Transit Practitioner’s Guide Bus Rapid Transit Practitioner’s Guide Page 3-7 Estimating BRT Ridership EXHIBIT 3-8 Oakland: Performance Measures Survey Percentage of Respondents Rating the Performance Measure Performance Measure Excellent Good Fair Poor Very Poor Rapid Bus service overall 39.3 43.6 14.8 1.2 1.2 Easy to identify the right bus 45.8 36.5 14.5 1.7 1.5 Wheelchair securement 42.4 37.8 16.6 1.9 1.3 Travel time on the bus 37.2 40.3 19.2 1.9 1.4 Quality of new buses 39.9 37.2 17.4 3.0 2.5 Location of bus signs 35.5 41.6 18.3 2.8 1.9 Frequency of buses 34.1 40.9 19.3 3.8 1.8 Reliability 30.3 42.0 23.0 3.3 1.4 Routes go where I need to go 34.7 36.6 21.8 4.7 2.3 Quality of bus shelters 27.6 41.7 24.1 4.5 2.0 Cleanliness 26.7 42.1 23.2 5.5 2.5 Personal safety on buses 26.0 42.2 24.4 4.7 2.7 Driver courtesy 29.6 38.8 24.2 4.6 3.6 Information at bus stops 27.2 37.8 22.3 9.4 3.3 Availability of seats 21.2 39.4 28.3 8.3 2.9 Value for fare paid 23.1 33.5 27.7 9.7 6.0 SOURCE: AC Transit presentation Washington, D.C. Service improvements desired by bus riders and non-riders in the Washington, D.C., area are set forth in Exhibit 3-9 and Exhibit 3-10. The improvements most desired by riders were on-time performance, more frequent service, and a longer service span. The improvements most desired by non-riders were better information, better shelters, and more convenient stops. These desired service improvements are clearly the service features that are or can be supported by BRT. SOURCE: WMATA Regional Bus Study, 2003, as reproduced in TCRP Web-Only Document 32 (6) EXHIBIT 3-9 Service Improvements Desired by Bus Riders in Washington, D.C.

Bus Rapid Transit Practitioner’s Guide Estimating BRT Ridership Page 3-8 Bus Rapid Transit Practitioner’s Guide SOURCE: WMATA Regional Bus Study, 2003, as reproduced in TCRP Web-Only Document 32 (6) EXHIBIT 3-10 Service Improvements Desired by Infrequent Bus Riders and Non- Riders in Washington, D.C. Research Findings Several research investigations have analyzed the ability of express bus/BRT service to attract riders relative to the ability of rail transit to attract riders: • A landmark study by McFadden et al. (7) found that, where travel times, costs, transfer requirements, and system quality are equal, rail- and bus- based rapid transit systems are likely to have the same passenger attraction. • Pushkarev and Zupan (8) found that both new busways and new rail rapid transit lines experienced substantial increases in ridership: > Shirley Busway - Washington, D.C. 104% > Lindenwold Line (rail) - New Jersey 56% > Skokie Swift (rail) - Chicago 54% > BART (Transbay) (rail) - San Francisco 51% • Ben-Akiva and Morikawa (9) indicated that, when quantifiable service characteristics (travel time, cost, transfers, etc.) are equal, riders show no preference for rail transit over quality bus alternatives for CBD-oriented work trips. • Currie (10) stated that BRT systems should be able to generate ridership equal to rail when the total trip attributes of both alternatives (travel times, costs, ride quality, minimal transfers, and quality of stations and facilities) are the same. Conclusions from Aggregate Evidence The preceding examples suggest BRT ridership responses that are more similar to what happens when new rail systems are introduced rather than what happens with relatively simple changes in local bus service frequency, travel times, and service span. The examples suggest that the identity, information, and amenity advantages of BRT in addition to improvements in span of service, frequency, routing, and travel times are important in attracting riders. Four past research investigations evaluated the relative ridership generation propensity of BRT and LRT.

Bus Rapid Transit Practitioner’s Guide Bus Rapid Transit Practitioner’s Guide Page 3-9 Estimating BRT Ridership The combined effects of improved travel times, service frequencies, and BRT features on BRT ridership are summarized for four BRT services in Exhibit 3-11. (The information in the exhibit was summarized for the BRT planning study Bus Rapid Transit Plans in New York’s Capital District [11] and used in forecasting BRT ridership for the Route 5 corridor in Albany, NY.) Exhibit 3-11 shows that 10% to 21% of the ridership increases were in addition to those attributed to travel time and service frequency improvements. The greatest gains were in Boston, where extensive physical changes and urban design improvements were made along the streets used by BRT. EXHIBIT 3-11 Impact of Various Factors Beyond Level of Service on BRT Ridership Los Angeles Metro Rapid Vancouver Boston Ridership Increase Ventura Blvd Wilshire- Whittier Blvd B-Line #98 Silver Line 2,850 riders1 20,660 riders1 4,0002 2,2903 Weekday 26% 33% 29% 30% Due to headway changes 6% 8% 9% 7% Due to travel time changes 10% 12% 6% 2% Due to other changes 10% 13% 14% 21% 1 SOURCE: TCRP Report 90 (12) 2 SOURCE: APTA Intermodal Operations Planning Workshop (13) 3 SOURCE: MBTA counts Based on these findings, it is likely that a full-featured BRT service operating on a fully segregated running way with specialized (or stylized) vehicles, attractive stations, and efficient fare collection practices would have a 25% gain in base ridership beyond gains from travel time and service frequency improvements. RIDERSHIP ESTIMATION OVERVIEW Ridership estimation procedures should recognize the unique aspects and needs of BRT. BRT operates in an assortment of running ways that range from mixed traffic to grade-separated busways. In many cases, BRT services are even laid out on existing bus routes. The range of operating environments suggests that several ridership estimation approaches may be appropriate. Approaches include applying regional travel demand and mode choice models, using pivot-point procedures (incremental logit models), and applying service elasticities. The pivot-point approach is especially desirable where BRT will operate on a new alignment such as a busway. The methods that are used should be reliable, produce reasonable results, and be easy to comprehend by transit planning and operations personnel. Data collection requirements and costs should be kept to a minimum. Surveys are needed to provide a clear picture of existing travel patterns and provide inputs for model development and calibration. An on-board rider survey is essential to indicate where passengers board and alight; to identify passenger origins, destinations, and trip purposes; and to obtain passengers’ socioeconomic characteristics. A CBD (or other major activity center) employee survey can provide useful information on employee travel modes and trip origins and destinations. A full-featured BRT service with separate running way could have a 25% gain in ridership beyond gains associated with travel time and frequency improvements.

Bus Rapid Transit Practitioner’s Guide Estimating BRT Ridership Page 3-10 Bus Rapid Transit Practitioner’s Guide Equitable treatment of all modes of travel in terms of travel times, costs, path- building, and network assumptions is essential. “Path-building” for use in models should give proper weights to each travel time component. Transit networks should clearly differentiate various BRT and local transit services that operate along the same street or in the same corridor. APPLICATION OF TRAVEL DEMAND ESTIMATION MODELS BRT ridership should be estimated from the traditional four-step demand estimation process when major investments are anticipated (e.g., when BRT will operate on new right-of-way). Household travel surveys are needed to provide the basic information for modeling and analyzing. The modeling process is appropriate on a system (or corridor) scale especially for long time horizons where future growth is anticipated. Providing realistic estimates of current and future employment is essential. Analysis should be conducted for both peak and off-peak conditions. Key Steps The steps and data flow for the four-step demand estimation process are shown in Exhibit 3-12 and discussed below: 1. Trip Generation. This step estimates the number of trip ends produced by and attracted to a given travel analysis unit or zone. Inputs are generally the number of households and jobs, stratified by characteristics such as income, auto ownership, and type of job. This step models the trip frequency decision. 2. Trip Distribution. This step estimates travel flows by linking trips “produced” and “attracted” based on zone-to-zone “impedances” (i.e., generalized cost estimates derived from a transportation system network description). 3. Mode Choice. This step models the mode choice decision. It estimates the share of trips made between each origin-to-destination pair based on demographic characteristics (e.g., income) and the characteristics of the competing modes in terms of times and costs. 4. Assignment. This step models path choice. It estimates the flow of person or vehicle trips on/through each element of the transportation network (i.e., link, lines, stations, and termini). Inputs are origin-to-destination trip matrices or trip tables and a transportation system network description. Appropriate “path-building” for BRT is important in using regional modeling to develop BRT ridership projections. While most planning agencies use the traditional four-step trip-based process, many agencies in larger metropolitan areas are considering use of tour-based or activity-based processes. Advanced models based on these processes may be used when available for BRT analysis. Most modern mode choice models are based on a theory of user utility first applied by McFadden. This approach posits that an individual makes a choice among alternatives based on the utility of each alternative relative to the combined utility of all other alternatives considered. Utility is represented as a function that is a linear combination of the measured attributes of each alternative plus a constant that reflects the value associated with the unmeasured attributes.

Bus Rapid Transit Practitioner’s Guide Bus Rapid Transit Practitioner’s Guide Page 3-11 Estimating BRT Ridership Trip Generation Land Use, Socioeconomic Characteristics Trip Ends Trip Distribution Mode Split Assignment Detailed Netw ork Description, O/D Times and Costs Total Person Trip Tables Trip Tables by Mode Line, Link Flow s SOURCE: Sam Zimmerman EXHIBIT 3-12 Four-Step Travel Demand Estimation Process Mode Choice The choice of travel mode can be forecast in several ways depending on the nature of the project and available data. The generally accepted method for projecting choice of mode in the four-step process is to apply a discrete choice model estimated using disaggregate data of revealed behavior. This model most often takes the form of a mode-split model based on a logistics (or “logit”) function. While developed with disaggregate data, these models are generally applied using aggregate data developed in the previous steps of the four-step process. This logit formulation estimates the probability of the choice of each of the various modes for any given trip depending on each mode’s relative desirability for that trip. Modes are relatively more desirable if they are faster, cheaper, or have other more favorable features than competing modes. Stated another way, the share of trips between two points is a function of the utility of a given mode divided by the sum of the utilities of all modes (expressed in exponential terms). Thus, the more utility that a mode has for a potential traveler, the larger its share of all trips will be. The general form of the model is shown in the following equation: A logit model can be used to estimate mode choice.

Bus Rapid Transit Practitioner’s Guide Estimating BRT Ridership Page 3-12 Bus Rapid Transit Practitioner’s Guide ∑ = = n x xdemoofutility mdemoofutility e emP 1 )( (3-1a) where: P(m) = the probability of using mode m for a given trip n = number of modes available for a given trip In application, the model is expressed in “disutilities” or impedances. The general equation becomes: ∑ = = n x xdemoforimpedancea mdemoforimpedancea ij ij ij e emP 1 )( )( )( (3-1b) where: P(mij) = the probability of using mode m for a given trip between i and j a = calibration parameter n = number of modes available for a given trip The exponential expansion is negative. Therefore, the higher the impedance of a specific mode is, the lower the probability that the mode will be chosen and the lower its resulting mode share. The formulation shown in Equation 3-1b is referred to as “multinomial logit.” It is used when the data suggest that all modes are “independent,” so that a change in the probability of choice of any one mode (m) will affect the choice probabilities for all other modes (1…n) proportionally. In practice, choices of mode are not fully independent. Instituting a significant new transit facility (e.g., BRT or light rail) is more likely to attract those who are already using some form of transit. The choice of transit, as opposed to the choice of an auto-based mode, reflects all the transit options available to a user. The form of the model used to estimate these sets of choices (e.g., walk-to- transit vs. drive-to-transit and local service vs. premium service) is known as “nested logit.” The impedance of a higher-level choice (e.g., transit vs. auto) is expressed as a combination of the impedances of modes included in subordinate sets or “nests” (e.g., local service and premium service) and the relative choice probabilities. In either formulation, the impedances are defined by a linear combination of factors that reflect the mode’s attractiveness, the characteristics of trip-makers and their households, and other factors that represent the environment in which the trip is made. The factors directly related to the attributes of the mode of interest include in-vehicle travel time, out-of-vehicle travel time (e.g., walking, waiting for the initial vehicle, and waiting while transferring), and out-of-pocket costs (such as fares, tolls, and parking). Factors related to trip-makers and their households can include income, auto ownership, number of workers, and so forth. Income often is reflected by converting out-of-pocket costs to equivalent minutes of in-vehicle time based on a proportion of the hourly wage. Environmental factors affecting choice of mode include such items as CBD destination, availability of sidewalks, and density of development.

Bus Rapid Transit Practitioner’s Guide Bus Rapid Transit Practitioner’s Guide Page 3-13 Estimating BRT Ridership Even when all the factors listed previously are included in the impedance function for a given mode, analysis typically reveals that a quantity needs to be included to reflect the unmeasured attributes. This term is often referred to as the modal bias constant. The attributes of transit services that these constants are thought to reflect include attributes such as reliability, comfort, passenger amenities, station features, and service branding. These factors are calibrated for each mode for each trip interchange. Exhibit 3-13 illustrates a typical logit-based mode choice model that could be utilized. A possible formulation for impedance of each mode is as follows: [ ] bxaxWaxaxampedanceI nn +++++= ...332211 (3-2) where: W = income of traveler x1 = in-vehicle travel time x2 = out-of-vehicle travel time x3 = out-of-pocket costs xn = other measures a1...an = coefficients b = modal bias constant Further information on disaggregate travel demand/mode choice modeling can be found in Discrete Choice Analysis Theory and Application to Predict Travel Demand (14), Urban Travel Demand (15), and A Self-Instructing Course in Disaggregate Mode Choice Modeling (16). EXHIBIT 3-13 Example Logit Mode Choice Model Formulation Use of a bias constant reflects unmeasured mode choice attributes. ∑ = m impedancea impedancea ij m ij m ij e emP )( )( )( (3-3) where: P(m)ij = the probability of a trip from origin i to destination j using mode m Impedanceij = mmmij mm ij mm ij m EDCTOVBTIVA +×+×+× $ = generalized cost or disutility (impedance) = a measure of travel difficulty Am, Bm, Cm, Dm = model coefficients (differentiated by mode in some cases) mijTIV = in-vehicle time between zones i and j for mode m mijTOV = out-of-vehicle time between zones i and j for mode m mij$ = out-of-pocket cost between zones i and j for mode m Em = modal bias constant for mode m (may be same for all transit modes) a = calibration parameter e = base of the natural logarithms

Bus Rapid Transit Practitioner’s Guide Estimating BRT Ridership Page 3-14 Bus Rapid Transit Practitioner’s Guide INCREMENTAL LOGIT MODEL (PIVOT-POINT PROCEDURE) Pivot-point procedures that apply the incremental logit model are extremely useful in forecasting BRT ridership when they are accompanied by on-board surveys that capture key traveler information. They have been used in places as diverse as the United Kingdom; China; York, ON; Philadelphia; and Tucson. They have several important advantages: • They are observed (measured) mode shares. • They require describing only those system components that are anticipated to change. • They require data only for the influence area of the route or corridor under study. Thus, the analysis requires much less effort than developing a full-scale travel demand model, and it produces results that conform to the base case scenario. The pivot-point procedure estimates changes in mode choice relative to a base year condition. The predicted relative changes are applied to a base matrix to determine future demand (ridership). More specifically, the future mode share is a function of the existing mode share and the changes in utilities for a specific mode as compared with the changes in utilities for all modes being analyzed. (Further discussion is contained in ten publications listed in the references [17–26]). The formulation of the incremental logit model is as follows: ∑ = ∆ ∆ × × = k i u i u i i i i eP ePP 1 ' (3-4) where: Pi = baseline probability of using mode i Pi’ = revised probability of using mode i ∆ui = the change in utility for mode i k = number of travel modes available Manheim (23) has suggested the following simplifications: 1. Two modes i and j ij uu i j i e P P P ∆−∆         + = 0 0 ' 1 1 (3-5) where: P0i and P0j = initial mode share for modes i and j ∆ui and ∆uj = the change in utility for mode i and j 2. Only changes in some levels       −+ = ∆− 111 1 0 ' 0 t u t P e P t (3-6) where: P0t = initial transit mode share P0t’ = future transit mode share ∆ut = change in transit utility The incremental logit model draws its coefficients for utilities from available models for the area under consideration. Where such models are not available, The incremental logit model is also known as the pivot-point procedure. The pivot-point procedure is useful for forecasting BRT ridership when accompanied by a transit rider survey.

Bus Rapid Transit Practitioner’s Guide Bus Rapid Transit Practitioner’s Guide Page 3-15 Estimating BRT Ridership coefficients may be “borrowed” from other sources. Alternatively, the illustrative coefficients shown in Exhibit 3-14 may be utilized. These coefficients represent the mid-range of U.S. experience with mode choice models. EXHIBIT 3-14 Illustrative Coefficients in the Mode Choice Model Variables Coefficients Attribute Units HBW HBO NHB In-vehicle time for (most) transit modes Minutes -0.020 -0.010 -0.020 In-vehicle time for commuter rail Minutes -0.016 -0.008 -0.016 All out-of-vehicle time Minutes -0.040 -0.020 -0.040 Drive-access time Minutes -0.040 -0.020 -0.040 Transfers Number -0.100 -0.050 -0.100 Fares (cents) Cents -0.003 -0.0015 -0.0015 Transit-access logsum Utiles 0.6 0.6 0.6 SOURCE: Discussion Piece #9 (27) The general steps in applying the model are as follows: 1. Define the influence area for the traffic analysis zones that would be directly affected by the new BRT service. 2. For each affected zone-to-zone pair, define the existing transit service level (in terms of in-vehicle time, wait time, walk time, travel time, etc.) and the likely changes as a result of the proposed BRT service. 3. Estimate the change in transit share by the pivot-point process. 4. Convert pivoted transit shares to a zone-to-zone trip table. 5. Assign trips to the proposed BRT line to derive ridership estimates. This process can be completed for future years by applying growth factors or using available future-year trip matrices (19). Transit path-building should find for each zone-to-zone interchange the best single path available for transit system walk access and for transit system drive access. Ideally, the best paths should reflect combined headways (where several transit lines on the same mode service common boarding and alighting locations) and should avoid multiple-path effects across different transit modes (27). Illustrative impedance factors are given in Exhibit 3-15. The weights given to transfers depend upon the attractiveness and convenience of transfers. EXHIBIT 3-15 Illustrative Impedance Weights for Path Selection Impedance Units Weight In-vehicle time for (most) transit modes Minutes 1.0 In-vehicle time for commuter rail Minutes 0.8 All out-of-vehicle time Minutes 2.0 Drive-access time Minutes 2.0 Transfers Number 2.0-5.0 Fare (cents, peak/off-peak) Cents 0.15/0.075 SOURCE: Discussion Piece #9 (27) APPLICATION OF ELASTICITY FACTORS As previously discussed, ridership changes resulting from BRT service can be estimated by introducing BRT travel times and service frequencies into mode-split models. Alternatively, it may be desirable to apply various travel time and service elasticities based on estimated changes in service span, frequencies (or bus miles), There are five key steps in applying the incremental logit model. Elasticity factors can be applied where BRT is overlaid on existing routes and for small-scale BRT investments.

Bus Rapid Transit Practitioner’s Guide Estimating BRT Ridership Page 3-16 Bus Rapid Transit Practitioner’s Guide and travel times. Application of elasticities is generally appropriate where BRT service is overlaid on existing bus routes and there are relatively small-scale investments. Elasticity Methods Ridership elasticity is defined as the change in ridership corresponding to a 1% change in fare, travel time, or service frequency. It is normally computed in three ways: 1. Shrinkage Factor. The shrinkage factor has been used as a “rule of thumb” in estimating the ridership effects of fare changes. It is the simplest method to use and gives a reasonable approximation for small fare changes. The percentage increase in ridership is equal to the percentage change in an attribute (e.g., travel time) times the appropriate elasticity factor. The equations are as follows: 1 12 1 12 1 1 )( )( X XX R RR X X R R E − − = ∆ ∆ = (3-7a) or 1 121 12 )( X XXER RR − += (3-7b) where: E = elasticity R1 = base ridership R2 = estimated future ridership X1 = quantity of base attribute (such as travel time or frequency) X2 = quantity of future attribute 2. Midpoint (Linear) Arc Elasticity. This method is commonly used in estimating ridership changes and is used in Chapter 5. It is defined as follows: FR XEXE RXERXER 1 12 1211 2 )1()1( )1()1( = +−− +−− = (3-8) where: E = elasticity R1 = base ridership R2 = estimated future ridership X1 = quantity of base attribute (such as travel time or frequency) X2 = quantity of future attribute F = multiplier 3. Log Arc Elasticity. The log arc elasticity most closely approximates the “point elasticity.” It is defined as follows: 112 log)log(log2 10 RXXER +−= (3-9a) or 112 ln)ln(ln2 RXXEeR +−= (3-9b) Elasticity computations can involve shrinkage factor, midpoint arc elasticity, or log arc elasticity methods.

Bus Rapid Transit Practitioner’s Guide Bus Rapid Transit Practitioner’s Guide Page 3-17 Estimating BRT Ridership where: E = elasticity R1 = base ridership R2 = estimated future ridership X1 = quantity of base attribute (such as travel time or frequency) X2 = quantity of future attribute A comparison of these elasticity computation methods is shown in Exhibit 3- 16. For small changes (± 10%), the three methods give similar results. However, for large changes, results from the shrinkage factor method diverge considerably. EXHIBIT 3-16 Elasticity Values for Different Methods of Computation Fare Change (%) Log Arc Elasticity Midpoint Arc Elasticity Shrinkage Factor -50% -0.300 -0.311 -0.46 -30% -0.300 -0.303 -0.38 -10% -0.300 -0.300 -0.32 +10% -0.300 -0.300 -0.28 +30% -0.300 -0.302 -0.25 +50% -0.300 -0.311 -0.23 +100% -0.300 -0.311 -0.19 SOURCE: TCRP Web Document 12 (28) Application Application of elasticities requires estimating the likely base ridership along the BRT route. This base ridership reflects a portion of the total existing route or corridor ridership. An on-board survey of bus riders along the proposed BRT route or corridor can assist in allocating existing ridership between BRT and existing bus service. This survey should provide origin-to-destination and station- to-station travel patterns as well as rider characteristics. It can be adjusted to future years based on anticipated growth in the corridor. Base Ridership Estimates Ridership diversion from existing routes should reflect travel patterns, comparative travel times and service preferences, where BRT is located, and whether BRT replaces an existing bus route. General guidelines are given in Exhibit 3-17. When BRT replaces a single local service, all existing ridership can be allocated to the BRT service (Option 1). The more common circumstance is where BRT and local service will operate on the same street (Option 2). The ridership allocation between BRT and local service can be based on judgment (including experience elsewhere); it can reflect division of ridership equally between the two services; or it can (preferably) be based upon origin-to-destination and boarding/alighting patterns, market research, and/or relative travel times. Exhibit 3-18 gives possible allocations based on various relationships between BRT and local bus running times. To maintain reasonable headways between BRT and local bus service on the same street, it may be appropriate to initially allocate ridership about equally between the two services. This has been the experience of several existing BRT systems. Moreover, equal headways are desirable at major boarding points such as downtowns. Thus, a “default” allocation of 50% seems reasonable. Identify base ridership in order to apply elasticities.

Bus Rapid Transit Practitioner’s Guide Estimating BRT Ridership Page 3-18 Bus Rapid Transit Practitioner’s Guide EXHIBIT 3-17 Guidelines for Allocating Base Corridor Ridership to BRT and Local Services Service Before Service After Initial Allocation of Service/Ridership between BRT and Local Service Travel Time Savings Service Frequency Changes 1. Single local service Single BRT service All service on street allocated to BRT All time savings allocated to BRT in applying elasticities All frequency changes allocated to BRT; use BRT service frequencies in applying elasticities 2. Single local service BRT plus local service on same street 1. Judgment 2. Equal allocation 3. Based on patterns of boarding and alighting, where available, and relative travel times Time savings allocated to each type of service in applying elasticities Use a portion of BRT trips SOURCE: Estimated EXHIBIT 3-18 Examples of Mode Shares Based on Relative BRT and Local Service Running Time Ratios on Same Street Allocation Method Equation (a) 21 2 tt t + Equation (b) 21 2 tt t + Equation (c) 21 2 tt t ee e + Equation (d) 21 1 tt t ee e −− − + t1 = BRT minutes and t2 = local service minutes Relative Travel Times Results t1 t2 Equation (a) Equation (b) Equation (c) Equation (d) 1 1.0 0.50 0.50 0.50 0.50 1 1.5 0.55 0.60 0.62 0.62 1 2.0 0.59 0.67 0.73 0.73 1 2.5 0.61 0.71 0.82 0.82 1 3.0 0.63 0.75 0.88 0.88 1 4.0 0.67 0.80 0.95 0.95 SOURCE: Estimated A more general BRT ridership allocation equation is as follows: )()()( 111 BApRBpRApR +=+ (3-10) where: p = percentage of base ridership attracted to BRT A = ridership growth due to time savings (and possible frequencies computed by elasticities) B = increase in base ridership resulting from special features of BRT R1 = base bus ridership on street (or in corridor)

Bus Rapid Transit Practitioner’s Guide Bus Rapid Transit Practitioner’s Guide Page 3-19 Estimating BRT Ridership Elasticity Computations Preferably elasticities should be applied on a station-to-station basis. An approximate value can be obtained by looking at aggregate time savings and ridership. Exhibit 3-19 gives typical midpoint arc elasticity values that could be used in estimating ridership. It should be noted that there is a considerable range in reported elasticities. Therefore, these values should be modified as appropriate to reflect local experiences. EXHIBIT 3-19 Typical Midpoint Arc Elasticities Item Travel Time Bus Miles Bus Frequencies Application New routes replace or complement existing routes Service expansion Greater frequency of existing routes Range -0.3 to -0.5 0.6 to 1.0 0.3 to 0.5 Typical -0.4 0.7 to 0.8 0.4 SOURCE: Patronage Impacts of Changes in Transit Fares and Services (29) and TCRP Report 99 (30) Elasticity data for in-vehicle travel times can be obtained from The Demand for Public Transportation (31). This document and similar U.S. information suggest that the in-vehicle travel time elasticity for home-based work trips (as affected by dedicated exclusive bus lanes) should be in the range of -0.5 to -0.7. General elasticity values of -0.3 to -0.5 have been reported both in the United States and United Kingdom. As an example, assuming that travel times decrease from 12 to 10 minutes as a result of BRT operation, the following changes in ridership are anticipated based on an elasticity of -0.35 and a base ridership of 1,000. By the shrinkage factor method: %8.5058,1 12 )1210)(000,1)(35.0( 000,12 +== −− +=R By the midpoint arc elasticity method: %6.6066,1 )12)(135.0()10)(135.0( )000,1)(10)(135.0()000,1)(12)(135.0( 2 +== +−−−− +−−−− =R The equation for estimating BRT ridership from changes in both in-vehicle travel time and service frequency is as follows: [ ] )1()1(1 2113 xFaFRR +−+= (3-11a) where: R1 = base ridership F1 = multiplier for travel time elasticity F2 = multiplier for service frequency elasticity a = proportion of BRT ridership that would save time by boarding the first bus that arrives at a combined BRT/local stop Examples of elasticity calculations are provided.

Bus Rapid Transit Practitioner’s Guide Estimating BRT Ridership Page 3-20 Bus Rapid Transit Practitioner’s Guide x = increase in ridership resulting from special BRT features (0 < x < 0.25) R3 = estimated future ridership If a = 1, the equation reduces to: )1(2113 xFFRR += (3-11b) The relative increase in ridership is: [ ] )1()1(1 21 1 3 xFaF R R +−+= (3-11c) Thus, the relative increase in ridership is independent of the initial ridership value. When BRT is overlaid on local bus routes, there may be improved frequency at BRT stations as a result of the combined service, and some proportion of riders may save time by taking the first bus (BRT or local) that arrives. This proportion of riders is represented by a in Equation 3-11a. Estimates of the proportion of riders saving time for a combined BRT/local route are given in Exhibit 3-20. If the individual headways are 10 minutes and the time saved by taking the first bus that arrives is 20 minutes, then, from Exhibit 3-20, a = 0.25. EXHIBIT 3-20 Estimated Proportion of Riders Saving Time for Various BRT and Local Headways Individual BRT and Local Headway* Total Time Savings 8 Minutes 10 Minutes 12 Minutes 5 minutes 0.80 1.00 1.00 10 minutes 0.40 0.50 0.60 20 minutes 0.20 0.25 0.30 30 minutes 0.13 0.17 0.20 40 minutes 0.10 0.12 0.15 * These values are the values of a in Equation 3-11a, and a = BRT headway ÷ (2 x time savings on entire route). SOURCE: Computed ESTIMATING ADDITIONAL RAPID TRANSIT RIDERSHIP IMPACTS Transit riders want to reach their destinations safely, quickly, and reliably. This objective is best met by bus and rail rapid transit that operates as a premium mode and offers riders the following: • A clearly identifiable running way, with a sense of permanence and minimum traffic interferences • Safe, secure, and convenient access to attractive yet functional stations • Clean, comfortable, climate-controlled vehicles that are easy to board and exit • Passenger information systems at stations and on vehicles, which give “next station” announcements and vehicle arrival times • A long service span, with frequent service throughout the day • A simple, understandable service pattern • A clear system image and identity

Bus Rapid Transit Practitioner’s Guide Bus Rapid Transit Practitioner’s Guide Page 3-21 Estimating BRT Ridership An open question is what features BRT must have in order to qualify as “premium.” Is branding sufficient? Can operations in mixed traffic afford a degree of reliability sufficient for riders to perceive the operations as premium, and, if not, what proportion of the service must be on a restricted guideway to achieve “premium” status? What features are required at BRT “stations” for riders to perceive the service in a manner similar to rail? Some suggested answers follow. Current practice suggests that a modal bias constant in the range equivalent to 10 to 12 minutes of in-vehicle travel time is appropriate to account for the characteristics of rail transit service that are not represented in impedance functions that include only travel time, service frequency, and cost. A few studies based upon travel time elasticity computations have suggested that full-featured, “complete” BRT could attract up to 25% more riders than that obtained by applying elasticity factors. This additional ridership reflects “new” trips as a result of BRT. Using these findings as a guide, a travel time bias constant equivalent up to 10 minutes of in-vehicle time may be considered in forecasting ridership for BRT systems, depending upon the extent and quality of the BRT system. A “complete” BRT system could also increase the base ridership up to 25% more than that obtained from elasticity computations. This increase is in addition to the ridership gains resulting from elasticity computations. Each BRT component will account for a portion of the 25% increment. An estimated distribution of the additional ridership impacts, grouped by the estimated maximum percentages for each component, is shown in Exhibit 3-21. These estimates were developed by the research team. Where site-specific data from preference surveys suggest other percentages, the site-specific data should be used. Transit agencies are encouraged to collect local data and/or derive percentages from customer surveys and share their findings with other transit agencies. EXHIBIT 3-21 Estimated Additional Ridership Impacts of Selected BRT Components Component Maximum % Running ways 20% Stations 15% Vehicles 15% Service patterns 15% ITS applications 10% Branding 10% Subtotal 85% BRT component synergy (when subtotal is 60 or more) 15% Total 100% SOURCE: Estimated by research team Because a quality running way provides the basic underpinning of BRT, it is estimated to account for 20% of new ridership. Stations, vehicles, and service patterns are each estimated to account for 15%. ITS applications and branding are each estimated to account for 10%. Another 15% is suggested for the synergy of all components when the subtotal exceeds 60%. Exhibit 3-22 gives a breakdown of the (estimated) percentages for various types of treatments for each component. Except for running ways, the percentages are additive, depending upon the number of features provided per component. Added BRT features have an impact on ridership beyond those impacts associated with travel time savings and service frequency improvements. This impact could be as high as 25%. Site-specific data from preference surveys can be used to identify incremental BRT ridership impacts. Multiple BRT components may create synergy.

Bus Rapid Transit Practitioner’s Guide Estimating BRT Ridership Page 3-22 Bus Rapid Transit Practitioner’s Guide EXHIBIT 3-22 Additional Ridership Impacts of Selected BRT Components Component Percentage 1. Running Ways (not additive) Grade-separated busways (special right-of-way) At-grade busways (special) Median arterial busways All-day bus lanes (specially delineated) Peak-hour bus lanes Mixed traffic 20 (20) (15) (10) (5) — — 2. Stations (additive) Conventional shelter Unique/attractively designed shelter Illumination Telephones/security phones Climate-controlled waiting area Passenger amenities Passenger services 15 — 2 2 3 3 3 2 3. Vehicles (additive) Conventional vehicles Uniquely designed vehicles (external) Air conditioning Wide multi-door configuration Level boarding (low-floor or high platform) 15 — 5 — 5 5 4. Service Patterns (additive) All-day service span High-frequency service (10 min or less) Clear, simple, service pattern Off-vehicle fare collection 15 4 4 4 3 5. ITS Applications (selective additive) Passenger information at stops Passenger information on vehicles 10 7 3 6. BRT Branding (additive) Vehicles & stations Brochures/schedules 10 7 3 Subtotal (Maximum of 85) 85 7. Synergy (applies only to at least 60 points) 15 Total 100 NOTE 1: Applies to a maximum of 10-min travel time bias constant (e.g., percentage of 10 min) NOTE 2: Applies to a 25% gain in ridership beyond that obtained by travel time and service frequency elasticities SOURCE: Estimated by research team Exhibit 3-23 gives an example of the estimated additional ridership for a high- level BRT system (with busways, off-vehicle fare collection, special vehicles, etc.) and for a minimal BRT system (without those components). In this example, the high-level BRT system would have a 9.5-minute bias constant as compared to a 4.3- minute bias constant for the minimal system. The increase in base ridership is in addition to that obtained from elasticity computations. The increases in base ridership for the high-level and minimal systems shown in Exhibit 3-23 are 24% and 11%, respectively. GUIDELINES The most complex ridership forecasting approaches are used for the detailed alternatives analyses and project design activities associated with large, costly applications. At the other end of the scale are simple “sketch planning”

Bus Rapid Transit Practitioner’s Guide Bus Rapid Transit Practitioner’s Guide Page 3-23 Estimating BRT Ridership approaches—often using elasticities, growth factors, and other simple techniques “borrowed” from other cities—appropriate for smaller, less complex, and less risky applications. EXHIBIT 3-23 Illustrative Examples of Additional Ridership Estimates Com- ponent System 1 (High-Level) System 2 (Minimal) Running Ways Grade-separated busway 20% All-day bus lanes 5% Unique, attractively designed 2% Unique, attractively designed 2% Illumination 2% Illumination 2% Telephones/security phones 3% Telephones/security phones 0% Stations Passenger amenities 3% Passenger amenities 0% Uniquely designed vehicles 5% Uniquely designed vehicles 5% Wide multi-door access 5% Wide multi-door access 0% Vehicles Low-floor vehicles 5% Low-floor vehicles 0% All-day service span 4% All-day service span 4% High-frequency service 4% High-frequency service 4% Clean, simple service pattern 4% Clean, simple service pattern 4% Service Pattern Off-vehicle fare collection 3% Off-vehicle fare collection 0% Passenger information at stops 7% Passenger information at stops 7% ITS Applica- tions Passenger information on vehicles 3% Passenger information on vehicles 0% Vehicles and stations 7% Vehicles and stations 7% BRT Branding Brochures and schedules 3% Brochures and schedules 3% Subtotal 80% 43% Synergy 15% 0% Total 95% 43% Bias (10 minutes x Total) (in minutes) 9.5 4.3 Elasticity increment (0.25 x Total) 0.24 0.11 SOURCE: Estimated by research team Regardless of the type of BRT application being analyzed, the most conservative, reasonable ridership forecasting approach available should be used. In most cases, this conservative approach will involve ridership elasticity-based “growth” factors and/or mode choice models derived from statistical analyses of detailed demand survey data for the existing conventional local bus system. Pivot- point mode split estimates will be useful. If, however, comparisons are to be made with rail-based rapid transit alternatives or if it is desired to estimate the upper bound of an envelope of ridership expectations, then a more aggressive approach can be used. The guidelines presented below for this situation differentiate between conventional local bus systems and full-featured BRT applications. For the purpose of the guidelines, full-featured BRT is defined as follows: • The system has permanently integrated rapid transit elements as well as a unique identity and quality brand image. • The system operates on dedicated transitways, either totally independent from the street system or physically separated in arterial or freeway rights- of-way, for the majority of its corridor. • The system has all-day service levels that permit passengers to arrive randomly at stations and avoid experiencing waiting times perceived to be excessive (maximum headways of 15 minutes in the off-peak and 10 minutes in the peak).

Bus Rapid Transit Practitioner’s Guide Estimating BRT Ridership Page 3-24 Bus Rapid Transit Practitioner’s Guide • The system has permanent stations with a high design quality, a high level of amenities, and a unique BRT identity. • The system incorporates high-quality vehicles that are configured for the BRT services offered and markets served and have a unique BRT identity. Sketch Planning For “sketch planning” purposes, the “upper end of the envelope” ridership forecasts for full-featured BRT systems should use existing local bus ridership in BRT corridors and the elasticities cited in the literature (e.g., The Demand for Public Transportation [31]) for rapid transit systems, most often rail-based, rather than elasticities cited for conventional local bus systems. In the case that the requisite rapid transit travel times and service frequency elasticities are available for a given city, these should be utilized. If only local bus system elasticities are available for the given city, these can be “factored” using the procedures described earlier. The “BRT growth” factors can be applied above and beyond elasticity factors reflecting travel time and frequency changes to the local bus system. These factors should vary from 1.05 to 1.25 times the existing all-day corridor demand, depending on the nature of the BRT application and the extent of features provided. In the case of an integrated package of improvements where there is no dedicated running way but some type of BRT “brand identity,” then the special BRT factor should be 1.05 to 1.15. If the BRT application is similar to the full-featured system described above, the factor would be closer to 1.25. For applications that are neither a full-featured, integrated system nor a simple package of bus service and facility improvements, the BRT “growth factor” would be proportional to the features included but lie between 1.05 and 1.25 as previously noted. Detailed Alternatives Analyses Past practice has applied bias constants for rail rapid transit systems of no more than 12 minutes of equivalent in-vehicle travel time. Accordingly, a bias constant of up to an equivalent 10 minutes of in-vehicle travel time could be considered for full-featured BRT. Guidelines for applying bias constants to BRT systems follow: 1. Where the results of travel model calibration and validation efforts using real ridership data suggest that customer response to the travel times and out-of-pocket costs of new rail systems will be different from and more positive than those for conventional local bus systems, BRT alternatives of similar content and quality to the empirically observed rail systems should be treated the same as the rail alternatives. > The same mode choice model structure and “calibration” coefficients and constants should be utilized for BRT. The suggested bias constant ranges up to 10 minutes for BRT (and up to 12 minutes for rail-based transit). 2. Where neither rail-based transit nor BRT exists in a given metropolitan area, proposed BRT alternatives of content and quality similar to proposed rail-based alternatives should be treated the same as the rail-based alternatives. > The same mode choice model structure and “calibration” coefficients and constants should be utilized for BRT. The suggested bias constant The precise value will depend on the extent of BRT features. Mixed-traffic BRT alternatives (e.g., streetcar and rapid bus) would be afforded proportionally less favorable treatment than full-featured, higher quality BRT operating on independent, grade- separated running ways.

Bus Rapid Transit Practitioner’s Guide Bus Rapid Transit Practitioner’s Guide pihsrediR TRB gnitamitsE 52-3 egaP ranges up to 10 minutes for BRT (and up to 12 minutes for rail-based transit). 3. Where the BRT alternative is different (lower) in content and overall quality than proposed rail-based alternatives for valid technical reasons, then modal bias constants and impedance (generalized cost) coefficients should be adjusted to be a reasonable and proportional average (depending on quality and content) of those for the conventional local bus system and those for the rail-based modes, again keeping in mind the maximum 10-minute BRT in-vehicle travel time advantage noted above. > The mode choice model modal bias constant and impedance (generalized cost) coefficients used to estimate ridership for BRT should fall in between that obtained from a valid calibration for the existing local bus system and that obtained for a rail-based system. Where BRT falls in the continuum would depend on the nature of the respective systems. 4. Where the current rail system is old, in poor repair, and unreliable and has real safety and security issues, the coefficients and constants used for full- featured BRT alternatives and any proposed rail transit alternatives should be the same even if there is model calibration evidence that “all things being equal, customers prefer bus to rail.” Results should be checked for reasonableness, whatever method is used. REFERENCES 1. Diaz, R.B., M. Chang, G. Darido, E. Kim, D. Schneck, M. Hardy, J. Bunch, M. Baltes, D. Hinebaugh, L. Wnuk, F. Silver, and S. Zimmerman. Characteristics of Bus Rapid Transit for Decision-Making. FTA, Washington, D.C., 2004. 2. Miami-Dade South Busway Corridor Case Study, University of South Florida, Center for Transportation Research. 3. Herzenberg, A. MBTA Silver Line. Presented at Financial Future of Transit in the New York Region. New York University, Wagner School of Public Administration, Oct. 2004. 4. Kittelson & Associates, Inc., Herbert S. Levinson Transportation Consultants, DMJM+Harris, and OURCO, Inc. TCRP Project A-23A Interim Report. Unpublished. 2004. 5. MSI International. A Qualitative Study of Metro Rapid and Associated Alternatives. Jul. 2002. 6. TranSystems Corporation, Planners Collaborative, Inc., and Tom Crikelair Associates. TCRP Web-Only Document 32: Elements Needed to Create High- Ridership Transit Systems: Interim Guidebook. Transportation Research Board of the National Academies, Washington, D.C., 2005. 7. McFadden, D.L., A. Talvitie, S. Cosslet, I. Hasan, and F.A. Reid. Demand Model Estimation and Validation. Urban Travel Demand Forecasting Project, Phase I Final Report, Vol. 5. University of California, Berkeley, June 1977. 8. Pushkarev, B.S., and J.M. Zupan. Public Transportation and Land Use Policy. Indiana University Press, Bloomington, 1977. 9. Ben-Akiva, M., and T. Morikawa. Comparing Ridership Attraction of Rail and Bus. Transport Policy, Vol. 9. Elsevier Science, Ltd., Oxford, Apr. 2002. 10. Currie, G. The Demand Performance of Bus Rapid Transit. Institute of FTA is continuing to evaluate the use of bias constants in BRT and rail New Starts proposals.

Bus Rapid Transit Practitioner’s Guide Estimating BRT Ridership Page 3-26 Bus Rapid Transit Practitioner’s Guide Transportation Studies, Department of Civil Engineering, Monash University, Melbourne, Australia, 2004. 11. Falbel, S., H. Levinson, K. Younger, and S. Misiewicz. Bus Rapid Transit Plans in New York’s Capital District. Journal of Public Transportation, 2006 Bus Rapid Transit Special Edition. National Center for Transit Research, University of South Florida, College of Engineering, Tampa, 2006. 12. Levinson, H., S. Zimmerman, J. Clinger, S. Rutherford, R. Smith, J. Cracknell, and R. Soberman. TCRP Report 90: Bus Rapid Transit: Vol. 1, Case Studies in Bus Rapid Transit, and Vol. 2, Implementation Guidelines. Transportation Research Board of the National Academies, Washington, D.C., 2003. 13. APTA Intermodal Operations Planning Workshop. TransLink Welcoming Session. Vancouver, BC, Aug. 9-11, 2004. 14. Ben-Akiva, M., and S. Lerman. Discrete Choice Analysis Theory and Application to Predict Travel Demand. MIT Press, Cambridge, 1985. 15. Domenich, T., and D. McFadden. Urban Travel Demand: A Behavioral Analysis. North-Holland, New York, 1975. 16. Horowitz, J.L., F.S. Koppelman, and S.R. Lerman. A Self-Instructing Course in Disaggregate Mode Choice Modeling. Final Report. UMTA, Washington, D.C., 1986. 17. Semi-Independent Forecasts of Ridership and User Benefits for New Starts Projects. FTA, Washington, D.C., June 2006. 18. Kumar, A. Use of Incremental Form of Logit Models in Demand Analysis. Transportation Research Record 775. TRB, National Research Council, Washington, D.C., 1980, pp. 21–27. 19. Ho, E. DRRA Route 55 Corridor Ridership Estimates - Summary. Gallup Corporation. 20. Daly, A., J. Fox, and J.G. Tuinenga. Pivot-Point Procedures in Practical Travel Demand Forecasting. 45th Congress of the European Regional Science Association. European Regional Science Association, Amsterdam, August 2005. 21. City of Tucson Alternatives Analysis: Incremental Logit Methodology. Hexagon Transportation Consultants, Inc., San Jose, Apr. 2005. 22. Integrated Transport Economics and Appraisal Division. Transport Analysis Guidance: Variable Demand Modelling - Key Processes. Department for Transport, London, June 2006. 23. Manheim, M.L. Fundamentals of Transportation Systems Analysis. Vol. 1: Basic Concepts. MIT Press, Cambridge, 1979. 24. Tucson Major Transit Investment Study: Model Methodology and Forecasting Results. Draft. Hexagon Transportation Consultants, Inc., San Jose, Oct. 2005. 25. Analysis and Augmentation of MWCOG’s Transit Models: Executive Summary and Three Technical Work Papers. Barton-Aschman Associates, Washington, D.C., March 1983. 26. Martin, W.A., and N.A. McGuckin. NCHRP Report 365: Travel Estimation Techniques for Urban Planning. TRB, National Research Council, Washington, D.C., 1995. 27. Discussion Piece #9: Semi-Independent Forecasts of Ridership and User Benefits for New Starts Projects. FTA, Washington, D.C., June 2006.

Bus Rapid Transit Practitioner’s Guide Bus Rapid Transit Practitioner’s Guide Page 3-27 Estimating BRT Ridership 28. Pratt, R.H., Texas Transportation Institute, Cambridge Systematics, Inc., Parsons Brinckerhoff Quade & Douglas, Inc., SG Associates, Inc., and McCollom Management Consulting, Inc. TCRP Web Document 12: Traveler Response to Transportation System Changes: Interim Handbook. TRB, National Research Council, Washington, D.C., 2000. 29. Mayworm P.A., A.M. Lago, and J.M. McEnroe. Patronage Impacts of Changes in Transit Fares and Services. Report USDOT/UMTA UPM 33. Ecosometrics, Sept. 1980. 30. KFH Group, Inc. TCRP Report 99: Embracing Change in a Changing World: Case Studies Applying New Paradigms for Rural and Small Urban Transit Service Delivery. Transportation Research Board of the National Academies, Washington, D.C., 2004. 31. Balcombe, R., R. Mackett, N. Paulley, J. Preston, J. Shires, H. Titheridge, M. Wardman, and P. White. The Demand for Public Transportation, A Practical Guide, Report 593. Transportation Research Laboratory, London, 2004.

Bus Rapid Transit Practitioner’s Guide Estimating BRT Ridership Page 3-28 Bus Rapid Transit Practitioner’s Guide This page is intentionally blank.