Fixed-Route Transit Ridership Forecasting and Service Planning Methods (2006)

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.

Reason No. Agencies Responding Agencies Responding (%) New routes 31 86 Route changes affecting 25% or more of a route 24 67 New mode/new type of service 24 67 The next 5 or 10 years 23 64 The next fiscal year 22 61 Route changes affecting less than 25% of a route 16 44 Minor adjustments to route segments 12 33 Scheduling changes 11 31 Other 5 14 Threshold No. Agencies Responding Agencies Responding (%) Formal 13 41 Informal 8 25 None 11 34 Total responding 32 100 TABLE 4 THRESHOLD FOR TRIGGERING RIDERSHIP FORECAST TABLE 3 REASONS FOR FORECASTING RIDERSHIP INTRODUCTION This is the first of two chapters presenting the results of a sur- vey of transit agencies regarding ridership forecasting. The survey was designed to elicit information on methodologies in use in a variety of situations, level of satisfaction with these methods, and suggestions for improvements. This chapter analyzes results related to data inputs, fore- casting methodologies, organizational issues, and the use of forecasting methods for specific scenarios. A wide variety of circumstances can generate the need for a ridership forecast, suggesting that a variety of tools and techniques may be needed. To address this issue, the survey provided seven spe- cific scenarios and asked how each agency would forecast ridership under each scenario. TYPOLOGY: TIME, GEOGRAPHIC SCOPE, AND EXTENT OF CHANGE Ridership forecasting varies from informal to formal or from simple to complex. Near-term changes are more likely to be evaluated informally, whereas most long-range transporta- tion plans use a traditional four-step model. Changes affect- ing one or two routes or route segments do not receive the same level of analysis as a systemwide restructuring. Minor scheduling or route adjustments rarely call for the use of a formal model; however, the introduction of new modes such as light rail or BRT almost always does. There is an inverse concern regarding the appropriateness of a particular method for a particular purpose. Traditional four-step travel models were not designed to measure the results of incremental changes to the transit network, are far too time consuming to use for such a purpose, and would be unlikely to yield an accurate answer because they are not sen- sitive to this level of change. Back-of-the envelope methods may be insufficient for forecasting the ridership impacts of a package of service changes. The survey asked agencies under what circumstances they would prepare a ridership forecast (Table 3). A majority of the agencies reported that they would forecast ridership for a new route, major route changes, a new mode or type of ser- vice, for the next 5 or 10 years, and for the next fiscal year. Minor service changes or scheduling changes were much less 8 likely to generate a ridership forecast. The most common “other” response was a fare change. Table 3 suggests that there may be a threshold in terms of the scale of service change that would trigger a ridership forecast. Table 4 shows that 66% of respondents have either a formal or informal threshold. Four agencies noted a threshold of a 25% change in miles, hours, or riders, whereas three reported 10%. Other factors that would require a ridership forecast include the need for board approval and significant cost impacts. DATA INPUTS Ridership forecasting can rely on various factors, including ridership at different levels, origin/destination information, demographic and land use factors, and economic trends. Myriad data sources are available for use. This section describes the factors and data sources used as input, with par- ticular attention paid to origin/destination data. CHAPTER THREE RIDERSHIP FORECASTING METHODOLOGIES

9of ridership forecasting (29%). Nearly one-quarter of respond- ing agencies (23%) do not consider origin/destination data. The findings in this section suggest that a wide variety of data sources are used in ridership forecasting. Certain sources may be very important across all forecasts, whereas others may be useful only for particular types of forecast. Subsequent sections explore how data are used in forecasting procedures. ANALYTICAL TECHNIQUES Ridership forecasting techniques can differ by mode, time frame, and scope of the change being analyzed. This section presents agency responses regarding analytical techniques used to forecast ridership. Most agencies use more than one method of forecasting ridership, depending on the scope of the change and the pur- pose of the forecast (Table 9). The majority of responding agencies use different forecasting methods for long-range and short-range forecasts (Table 10). Interestingly, multimodal agencies are slightly more likely to use the same methodol- ogy for all modes (Table 11). Factors No. Agencies Responding Agencies Responding (%) Existing route or route segment ridership 31 89 Ridership on similar routes 30 86 Existing system ridership 28 80 Demographic factors in the service area 27 77 Land use within the affected service area 25 71 Origin/destination information 24 69 Economic trends within the service area 21 60 Other 10 29 Total responding 35 100 Factors Primary Use Existing route or route segment ridership Change in route Ridership on similar routes New route or corridor Change in route Existing system ridership Annual budget forecast Long-range plan Demographic factors in the service area Change in route Land use within the affected service area Change in route New route or corridor Origin/destination information Major new service Four-step travel model Economic trends within the service area No consensus TABLE 6 PRIMARY USES FOR INPUT FACTORS TABLE 5 FACTORS CONSIDERED AS INPUTS TO FORECASTING METHODOLOGY Data Sources No. Agencies Responding Agencies Responding (%) Ridership data from the farebox 30 86 Ridership data from recent ridechecks 28 80 Existing land use 25 71 Census demographic data 23 66 Origin/destination data from on-board surveys 22 63 Forecast land use 19 54 Ridership data from APCs 14 40 Origin/destination data from models 14 40 Economic forecasts 11 31 Economic trends 10 29 CTPP demographic data 9 26 Other 11 31 Total responding 35 100 APC = automated passenger counter; CTPP = Census Transportation Planning Package. TABLE 7 DATA SOURCES FOR RIDERSHIP FORECASTS TABLE 8 ROLE OF ORIGIN/DESTINATION DATA IN RIDERSHIP FORECASTING Role No. Agencies Responding Agencies Responding (%) Considered, but not a major part 15 43 Major part 10 29 Not considered 8 23 Depends on time frame/level of detail 2 6 Total responding 35 100 Table 5 reveals that agencies use a wide variety of factors as inputs to ridership forecasts. Ridership factors are most fre- quently mentioned, followed by demographic characteristics, land use, origin/destination data, and economic trends. Several factors were mentioned in the “other” category, including travel time, fares, congestion levels, automobile ownership, land use changes, and market research survey results. If a factor is involved for some types of changes or fore- casts but not others, the agencies indicated this on the survey. Roughly half of the agencies noted the type of change or forecast for which a particular factor is used. Table 6 high- lights the primary uses for each factor. The most often used data sources include ridership data from the farebox and from recent ridechecks, existing and forecast land use, census demographic data, and origin/ destination data from on-board surveys, as shown in Table 7. APCs have made inroads and are now used at 40% of responding agencies; however, APCs are still the least likely source of ridership data among those listed. In the “other” cat- egory, three agencies mentioned household travel surveys. The use of origin/destination data was of particular interest, and a question addressing this was included on the survey. As Table 8 shows, the most common response was that origin/ destination data are considered (43%), but are not a major part

10 The most frequently used techniques are qualitative in nature: professional judgment and rules of thumb/similar routes. At least half of responding agencies use elasticities and a traditional travel demand model to forecast ridership. Table 12 shows techniques included in agency methodology. Three of the “other” agency responses reflect trend line analy- sis; others mentioned Institute of Transportation Engineers trip generation rates, GIS, and an unspecified agency model. It may be useful to distinguish among the qualitative tech- niques in Table 12, because they are referred to later in this chapter. “Similar routes” forecasts ridership on a given route based on the experiences on other routes with similar service areas and frequencies. An analyst might base a ridership forecast for a new crosstown route on the productivity of other crosstown routes or develop a ridership estimate for an extension of a route to a mall based on ridership on other routes at similar malls. “Rules of thumb” codify past experi- ence in general rules. Examples can take the form of “new routes generate x riders per revenue hour,” or “route exten- sions to suburban residential developments generate y riders per 100 households.” “Professional judgment” relies on the judgment and experience of the analyst and is the most sub- jective qualitative technique. For example, an analyst might use professional judgment to adjust a ridership estimate developed by means of another technique upward or down- ward depending on the presence or absence of schools, retail centers, high-density residential development, or the general character of the neighborhoods along a particular route. If a technique is used for only certain types of changes or forecasts, the agencies indicated this on the survey. Roughly half of the agencies noted the type of change or forecast for which a particular technique is used. Table 13 highlights the primary uses for each factor. Professional judgment and rules of thumb/similar routes are most often used for route, ser- vice, and schedule changes. Service elasticities are used for these types of changes as well, whereas fare elasticities are used for fare changes. The four-step travel model is used most often for major new service. Several agencies reported using a range of service elastici- ties, as suggested in national studies (9,11). Service elasticities were different depending on existing service frequency, service area density, time of day, or analyst judgment. For agencies reporting a single service elasticity value, this value was as low as 0.2 (in New York City) and as high as 0.5. Reported fare elasticities varied from 0.175 to 0.35. Calculating ridership using passenger boardings (unlinked ridership) is common among transit agencies. Unlinked rid- ership is the reporting standard for the NTD, and APCs and fareboxes register boardings. For modeling purposes, knowl- edge of the number of linked trips is often desirable. Table 14 shows that transit agencies are much more likely to forecast ridership in terms of unlinked trips than linked trips. Methods No. Agencies Responding Agencies Responding (%) Multiple 23 66 Single 12 34 Total responding 35 100 Methods No. Agencies Responding Agencies Responding (%) Different by time frame 22 71 Same by time frame 9 29 Total responding 31 100 Methods No. Agencies Responding Agencies Responding (%) Different by mode 9 45 Same by mode 11 55 Total responding 20 100 TABLE 9 SINGLE VERSUS MULTIPLE METHODS OF RIDERSHIP FORECASTING TABLE 10 FORECASTING METHODS: SHORT-RANGE VERSUS LONG-RANGE FORECASTS TABLE 11 FORECASTING METHODS: MULTIMODAL AGENCIES Forecasting Technique No. Agencies Responding Agencies Responding (%) Professional judgment 29 83 Rules of thumb/similar routes 28 80 Service elasticities 22 63 Four-step travel demand model 18 51 Econometric model 7 20 Regression analysis 7 20 Other 7 20 Total responding 35 100 TABLE 12 FORECASTING TECHNIQUES USED BY TRANSIT AGENCIES TABLE 13 PRIMARY USES FOR FORECASTING TECHNIQUES Forecasting Technique Primary Use Professional judgment Any route or service change Rules of thumb/similar routes Any route or service change Change in route Headway/schedule change Elasticities Any route or service change Fare changes Headway/schedule change Four-step travel demand model Major new service Econometric model No consensus Regression analysis No consensus

11 The use of qualitative methods such as similar routes analysis or professional judgment is widespread among tran- sit agencies for route, schedule, and fare changes. Service elasticities are the major quantitative method in use. Several transit agencies are satisfied with the use of qualitative tech- niques, noting their accuracy and simplicity of use. ORGANIZATIONAL ISSUES Transit agencies have different structures. This section explores where the ridership forecasting function is located within an agency and whether it is a dedicated function or part of a planner’s overall responsibilities. This section also considers the time and effort required to prepare a ridership forecast and how forecasts are distributed. Responsibility for Ridership Forecasts Fourteen transit agencies reported more than one lead depart- ment in preparation of ridership forecasts. The transit agency’s planning department is the most common location for the ridership forecasting function, as shown in Table 15. Among the “other” departments are Strategic Planning and Policy, Transit Research Section (under the Marketing Department), and Business Development. Twelve of the 20 agencies that reported a single lead department for ridership forecasting named Transit Planning. Table 16 shows a fairly even split in terms of whether rid- ership forecasting is the responsibility of a dedicated person or group. The results suggest that this responsibility is some- what more likely to be part of general duties for all but major changes. Time and Effort Required A range of estimates were given for the time and effort required to prepare ridership forecasts. Table 17 shows that simple or short-range forecasts can generally be completed in 3 days or less, whereas complex or long-range forecasts can take much longer. The wide time range in long-range forecasts reflects the method used: trend line analysis takes much less time [one day or less was reported by seven respondents (Table 17)] than a four-step model run. How Forecasts Are Used Ridership forecasts are nearly always distributed and used internally. Most responding agencies also share the fore- casts with their boards. Table 18 shows that it is less com- mon to distribute these forecasts to other groups. Four respondents cited local stakeholders among “others” who receive ridership forecasts, whereas three mentioned the FTA. RIDERSHIP FORECASTING UNDER A VARIETY OF SCENARIOS Survey results in the previous sections shed light on how transit agencies go about the process of forecasting rider- ship. However, the very different types of circumstances under which a ridership forecast is needed or desired can be lost in an aggregation of overall responses. To better under- stand how ridership forecasts are generated and used, the survey included seven scenarios and asked transit agencies Department or Agency No. Agencies Responding Agencies Responding (%) Transit planning department 22 65 Transit operations planning department 9 26 Transit budget/finance department 8 24 MPO 6 18 Transit operations department 3 9 Other 4 12 Total responding 34 100 MPO = metropolitan planning organization. TABLE 15 DEPARTMENT OR AGENCY RESPONSIBLE FOR RIDERSHIP FORECASTS Structure No. Agencies Responding Agencies Responding (%) Part of general duties 13 38 Depends on scale/extent of forecast 13 38 Dedicated person or group 8 24 Total responding 34 100 TABLE 16 STRUCTURE OF RIDERSHIP FORECASTING FUNCTION Time No. Agencies Responding Agencies Responding (%) Simple or Short-Range Forecasts Less than one day 8 32 One to three days 12 48 Two weeks or longer 5 20 Total responding 25 100 Complex or Long-Range Forecasts One day or less 7 47 One to three months 3 20 Longer than three months 5 33 Total responding 15 100 TABLE 17 TIME REQUIRED FOR RIDERSHIP FORECASTS Type of Trip No. Agencies Responding Agencies Responding (%) Unlinked 24 71 Both linked and unlinked 10 29 Total responding 34 100 TABLE 14 RIDERSHIP FORECASTS: LINKED VERSUS UNLINKED TRIPS

12 to describe how they would forecast ridership under each scenario. Responses included data to be used and method- ologies. This section summarizes data and techniques men- tioned by at least 10% of respondents under each scenario. A complete list of responses is included in Appendix B. Each section also provides verbatim responses from selected agencies as examples of approaches to ridership forecasting. The case study chapter (chapter five) includes all responses to these scenarios from the six case study agencies Percentages in the scenario tables are based on answers from all 36 responding agencies. One agency indicated that it would not forecast ridership under any of the scenarios, and others indicated that they would not forecast ridership for certain scenarios. “Would not analyze” characterizes these responses in the table for each scenario. Scenario A: Half-Mile Rerouting of Existing Route to Serve a New Shopping Center The most common approaches under this scenario were to evaluate similar conditions in terms of shopping centers else- where in the service area and to evaluate similar routes and previous service changes of this nature. Current route rider- ship is important, as is consideration of the impact of this detour on existing through ridership. Agencies also reported the use of trip generation rates and professional judgment. Table 19 summarizes responses. Examples of specific responses include: We would use the GIS to provide an integrated comprehensive market analysis using Census demographics, APC ridership by stop, and other land use data as available to compare this service with our current same type of service and project from there using professional judgment. Size of shopping center, demographics of current route ridership, level of current route ridership, proximity of transfers to/from other routes that have ridership that would be attracted by the shopping center, [and] number of existing riders adversely affected by the deviation. Impact on existing customers—travel time, access, egress, fare, etc. Trip generation/distribution based on size, type of shopping center. Prior experience. Quick spreadsheet analysis. Scenario B: Extension of Existing Route for One Mile to Serve a New Residential Development As with Scenario A, the most common approaches under this scenario were to evaluate similar routes and previous service changes of this nature and evaluate similar conditions in terms of residential developments elsewhere in the service area. The socioeconomic and demographic profile of the area is useful to know, as is the population and population den- sity. Several agencies noted route productivity as a consider- ation; most would expect the same level of productivity for the extension, but would consider the productivity of similar route segments or of similar previous changes. Trip genera- tion rates and professional judgment were also cited as tools in developing a ridership forecast. Perhaps the most interest- ing response was to assume that the extension would meet minimum performance standards in terms of boardings per revenue hour or other factors, with the implication that if per- formance fell short, the extension would be discontinued. Table 20 presents the responses. Response No. Agencies Responding Agencies Responding (%) Similar conditions/area 13 36 Similar routes/service change 11 31 Current route ridership 9 25 Consideration of through ridership 8 22 Trip generation rate 6 17 Professional judgment 5 14 Would not analyze 5 14 TABLE 19 RIDERSHIP FORECASTING FOR SCENARIO A: REROUTING TO SERVE A NEW SHOPPING CENTER Response No. Agencies Responding Agencies Responding (%) Similar routes/service change 12 33 Similar conditions/area 11 31 Socioeconomic/demographic data 7 19 Route productivity 6 17 Trip generation rate 5 14 Assume minimum performance standard 5 14 Would not analyze 5 14 Population/population density/no. households 5 14 Professional judgment 5 14 TABLE 20 RIDERSHIP FORECASTING FOR SCENARIO B: ROUTE EXTENSION TO SERVE NEW RESIDENCES Distribution and Use No. Agencies Responding Agencies Responding (%) Internally 33 97 To board members 23 68 To the MPO 10 29 To elected officials 9 26 To others 10 29 Total responding 34 100 MPO = metropolitan planning organization. TABLE 18 DISTRIBUTION AND USE OF RIDERSHIP FORECASTS

13 Examples of specific responses include: Would not prepare a specific forecast, but would ensure that the new development has a sufficient number and density of residents to ensure that it could support service meeting our transit service guidelines. Depending on the type of service, our guidelines call for (1) a minimum density of 20 to 30 residents per hectare (8 to 12 residents per acre) or 20 to 25 jobs per hectare (8 to 10 jobs per acre) over a minimum developed area of 10 hectares (four acres); (2) a road and pedestrian access system that permits safe access and efficient operation of transit service; (3) a minimum of 175 to 250 total person trips (by all modes) per additional bus service hour. Other factors such as the socioeconomic characteristics of the community and existence of travel demand management pro- grams may also be considered in applying these guidelines. I’ve used two different methodologies. Most commonly, I per- form an analysis that compares community and service charac- teristics with similar parts of the route network. Typically, I’ll consider headways, span of service, residential and employment densities, and family incomes. If there is a major generator along the extension, I’ll consider it separately. For example, if we were considering an extension to a community college, I’d look at the number of students and apply a mode split. That mode split would vary depending on where students are coming from and whether we can coordinate bus schedules with class times and whether we will offer any fare incentive to customers. The second approach I’ve used is a small sketch planning model that I originally developed in the 1990s. It considers residential and employment densities within TAZs along the route, family income, headways, and average travel time from each TAZ along the route to several different types of traffic generators (malls, hospitals, community colleges, etc.). The relative importance of each variable is then calibrated to achieve maximum consistency between projected and actual boardings on established routes. Doing this, I found that the model is about as reliable as if I make an informed guess. Accordingly, I seldom use it. Scenario C: Change in Headway from 12 to 10 Minutes During Peak Hours A typical approach to forecasting ridership in response to a change in headway is to use elasticities. Several respondents mentioned route productivity as a factor. One application of productivity is to use historical or comparable productivity changes elsewhere as the basis for the ridership forecast. Another is to forecast ridership changes only if the route’s current load is above the maximum load factor. This scenario had the highest number of responses indicating that an agency would not perform a ridership forecast for this type of change (see Table 21). Examples of specific responses include: Work with MPO and use service elasticities from regional trans- portation model to forecast ridership increases. We would generally not conduct such a forecast due to the inelasticity of our ridership. Over 70 percent are transit depen- dent and are not generally swayed if frequencies change by such a small amount in either direction. However, we would gener- ally assume that any additional service hours would generate the same number of passengers per hour. Estimate the cost of more frequent service. From ride counts, obtain existing boardings on the route during the AM peak. From our transit assignment model, obtain the average ‘weighted’ travel time for customers who use this route (note that weighted travel time is the time for a customer’s trip from the beginning of the trip at the origin to their final destination with each time component weighted; e.g., wait time weighted by a factor of 1.5, walk time weighted by 2.0, etc.). Use elasticity model to estimate the number of new customers attracted due to the percentage decrease in their overall weighted travel time; e.g., TWTT [total weighted travel time] = 60 minutes, reduction in travel time will be 1 minute savings multiplied by wait weight of 1.5 = 1.5 minutes; therefore, per- centage savings in travel time = 1.5/60 = 2.5%. Our AM peak weighted travel time elasticity is 1.5; therefore, number of new customers = number of existing customers * 1.5 * 2.5% (note 2.5% is negative because it represents a travel time savings). Compute number of customers gained per dollar spent. If greater than agency threshold of 0.23, then recommend for six month trial; [otherwise] do not recommend. Note: more frequent service can also be recommended without any ridership forecast if current loads on the buses exceed agency standards. Scenario D: Implementation of New Crosstown Route to Enhance Service Area Coverage and Provide More Direct Connections An examination of other crosstown routes is the most com- mon response. Evaluating transfer data and how connecting routes are used is also important. Respondents also men- tioned the need to understand the demographics in the area to be served. Productivity was cited as the best metric to use in comparison with other routes and areas. Other approaches included using a four-step travel model (be- cause this would be a new route), considering similar con- ditions or areas, evaluating trip generators and land use within one-quarter mile of the proposed route, and assum- ing that the new route would meet minimum performance standards for a cross-town route. Table 22 summarizes responses. Response No. Agencies Responding Agencies Responding (%) Elasticities 12 33 Route productivity 10 28 Would not analyze 8 22 Professional judgment 4 11 Similar routes/changes 4 11 TABLE 21 RIDERSHIP FORECASTING FOR SCENARIO C: HEADWAY CHANGE

14 Examples of specific responses include: Would model using regional transportation model, but would interpret the results based on comparison with existing compa- rable routes given the inaccuracy of the regional model at the route level. Population along proposed route, demographics of population, traffic generators along route, convenient transfers to other routes. Review of all generators, attractions, service frequency, span, fares, competitive/complementary services in area, demograph- ics, employment. Scenario E: Implementation of New Mode such as Bus Route Transit This scenario presents the most drastic change to the exist- ing transit system and calls for the most formal analytical techniques to forecast ridership. Nearly half of all respon- dents indicated that they would rely on the four-step travel model. Several who mentioned that they would not analyze this type of change noted that there are no plans for a new mode of transit service and thus there would be no need to analyze ridership impacts. Many agencies would hire a con- sultant to develop a ridership forecast. Examination of travel time changes and application of appropriate elasticities were also mentioned (see Table 23). Examples of specific responses include: Code BRT service, modify models to add new mode, and use updated four-step model. If BRT was being examined, we would likely use elasticities to examine how current ridership would be impacted based on incremental improvements over regular bus service. If a new rail line was being examined, the rigor of the analysis would be based on how the forecast is being used. For conceptual design purposes, ridership would likely be developed using rule of thumb methods. For projects beyond conceptual design, we would likely use the MPO’s four-step model. For a simple feasibility analysis (not one requiring environ- mental clearance) we would calculate additional service hours on the corridor, including any time savings resulting from the BRT improvements. We would then review the number of pas- sengers per service hour on the existing service, and assume that the additional service hours would at least meet the exist- ing threshold. Then, we would review the ridership trend analysis for any of the other BRT corridors we have imple- mented in order to make assumptions of similar performance. We have experienced a six percent increase in ridership on our first BRT corridor. That increase would be compared to other proposed corridors to determine whether similar increases could be anticipated. For BRT projects that would require envi- ronmental clearance [Environmental Impact Statement], we would perform a series of model runs using the countywide travel demand model. Scenario F: Prediction of Next Year’s Ridership as Part of Budget Process Most agencies forecast next year’s ridership using trend line analysis, with some consideration for expected service and fare changes and professional judgment. A few agencies do not forecast ridership one year ahead. Table 24 summarizes responses. Examples of specific responses include: Service evaluation uses an historical trend methodology compar- ing the ridership trends between consecutive months over time and disaggregating by weekday, Saturday, and Sunday. Service elasticities are used when evaluating service changes prior to implementation of the methodology. For FY 99 through FY 06, percentage differences between forecast and actual annual system ridership have ranged from 0.01 percent to 0.85 percent. We use an econometric model for this. Response No. Agencies Responding Agencies Responding (%) Similar routes/changes 15 42 Transfer data/connecting routes 8 22 Socioeconomic/demographic data 6 17 Productivity 5 14 Would not analyze 5 14 Four-step travel model 4 11 Similar conditions/area 4 11 Evaluate trip generators/land use within 0.25 mile 4 11 Assume minimum performance standard 4 11 TABLE 22 RIDERSHIP FORECASTING FOR SCENARIO D: CROSSTOWN ROUTE Response No. Agencies Responding Agencies Responding (%) Four-step travel model 17 47 Would not analyze 7 19 Hire a consultant 6 17 Analyze travel times 4 11 Elasticities 4 11 TABLE 23 RIDERSHIP FORECASTING FOR SCENARIO E: NEW MODE SUCH AS BUS ROUTE TRANSIT Response No. Agencies Responding Agencies Responding (%) Trend line 21 58 Service level changes 13 36 Fare changes 5 14 Professional judgment 4 11 Would not analyze 4 11 TABLE 24 RIDERSHIP FORECASTING FOR SCENARIO F: RIDERSHIP FORECAST FOR NEXT YEAR

15 Generally a fairly rough estimate based on change in overall ser- vice level plus some adjustment for expected population changes (regional plan assumed 1.5 percent population increase per year). Our budget estimates have two very separate components. We generally apply an underlying system growth rate that is usually a conservative continuation of the previous year’s trend. In this system, that is usually about one percent. We then add in the sep- arate calculation of the impacts of any service changes we have programmed for the coming year, discounted to allow for a start- up period. We normally factor for a three-year start-up curve— 50% of projected in the first year, 75% in the second, and 90% in the third. Scenario G: A 10-Year Ridership Forecast as Part of a Long-Range Plan This scenario shows a split between formal and informal rid- ership forecasting techniques (see Table 25). As in Scenario F, the need to consider planned service changes was cited, and several agencies do not prepare a 10-year forecast. Examples of specific responses include: Based on service levels, impact of any fare changes, and growth/loss rate trends from recent years. We would start with this year’s ridership and change it as needed for any planned improvements, service reductions, fare changes, or anticipated economic changes based on professional judgment. However, we would factor in any model-based projections from our MPO if we are introducing major new service. Develop ridership trend and develop a target based on demo- graphic trends and professional judgment. Also work with MPO and use transit inputs into regional transportation model. SUMMARY Analysis of how transit agencies prepare ridership forecast- ing for seven scenarios supports and amplifies other survey responses. The findings included: • A wide variety of data sources are used in ridership forecasting. The most often used data sources include ridership data from the farebox and from recent ridechecks, existing and forecast land use, census demographic data, and origin/destination data from on- board surveys. APCs have made inroads, but are the least likely source of ridership data among those listed. Origin/destination data, although frequently consid- ered, are not a major component of ridership forecast- ing for a majority of respondents. • The planning department is the most likely home for the forecasting function within a transit agency. However, it is not unusual for multiple departments to be involved in different levels of ridership forecasting. • Simpler, less formal approaches are used for route-level and other small-scale service changes. The examples show that some of these “simpler” approaches have grown more sophisticated as GIS databases are used to assess demographic characteristics and identify similar routes and as APCs and ongoing programs improve the accuracy of ridership data. • Use of elasticities is widespread for changes to existing service, particularly frequency changes. • More formal methods, including use of the four-step travel model are used when either the change or the time frame is beyond the scope of the current system; for example, introduction of a new mode and forecast- ing over the next 10 years. The next chapter summarizes agencies’ assessments of their ridership forecasting methods. Response No. Agencies Responding Agencies Responding (%) Four-step travel model 16 44 Trend line 12 33 Service level changes 8 22 Would not analyze 5 14 TABLE 25 RIDERSHIP FORECASTING FOR SCENARIO G: 10-YEAR RIDERSHIP FORECAST