SWUTC Research Project Description
Title of Project: Predicting Incremental Ridership Due to Buses On Shoulder Implementation
Project Number: 476660-00073
Principal Investigator:
Randy Machemehl
(512) 471-4541
P.I. Affiliation: University of Texas at Austin
Project Monitor:
Surinder Marwah
Capital Metropolitan Transit Authority
Austin, TX
(512) 369-7047
Project Status: Active
Date Started: 9/1/09
Estimation Completion Date: 8/31/10
Estimated Cost - Current Fiscal: $58,000
Estimated Cost - Total Planned: $58,000
Project Summary:
Project Abstract:
“A BOS” [Bus-Only-Shoulder] is a street or highway shoulder constructed, modified, or enhanced to support bus traffic. Generally, buses are permitted to use shoulders during peak traffic periods to bypass congested sections of freeways or arterial roadways. BOS systems are now operating in many cities both inside and outside the United States. Due to the typically low costs of BOS implementation and quantifiable time savings of bus passengers, BOS implementations usually produce positive economic benefits. Where BOS systems have significant operating histories, numbers of bus riders have increased, however, transit systems have generally experienced ridership increases due to many factors including perceptions of congestion and fuel price. This work will develop a methodology for estimating additional riders that would be attracted to BOS implementations. The planned approach will use both ridership change histories and stated preference survey data to produce complementary predictive models.
Project Objectives:
This work will develop a methodology for estimating additional riders that would be attracted to BOS implementations. The planned approach will use both ridership change histories and stated preference survey data to produce complementary predictive models.
Task Descriptions:
Task 1: Summarize transit ridership estimation techniques that have been used to assess BOS impacts.
Transit agencies currently operating BOS facilities have been identified as part of the work already completed for Capital Metro. Some information regarding procedures used to estimate ridership changes has been obtained, however, the following list of cities known to have BOS will be re-contacted to further investigate any forecasting methods used. These are chosen due to the fact that they indicated in previous contacts that they did use some type of forecasting procedure.
Minneapolis, MN
Seattle, WA
San Diego, CA
Cleveland, OH
Cincinnati, OH
Miami, FL
A critical summary of methods used, if any, will be developed.
Task 2: Gather observed ridership changes associated with BOS implementations.
The same metro areas listed in Task 1 will be asked to provide before-after ridership data describing the impacts of BOS. Special emphasis will be placed upon Minneapolis and Seattle since these two cities have, by far, the longest history of BOS operations and they are US cities. Their data will be especially valuable because it will provide short and long term ridership change histories.
Task 3: Normalize observed ridership data removing seasonal and other time based trends.
A recently completed doctoral dissertation by Dr. Ashley Haire (faculty supervisor: Randy Machemehl) gathered transit ridership histories for over 250 US transit systems as part of an effort that produced ridership forecasting models for those systems individually and for geographic regions. Dr. Haire found significant time-wise (seasonal, monthly and other) patterns within those data and developed procedures for removing those trends from the ridership data. Her procedures will be applied to the data acquired in Task 2 so that ridership changes likely due to BOS implementation can be identified separate from other influences. Dr. Haire’s work provides clean ridership data for all the cities listed in Task 1, so availability of her work will greatly expedite this task.
Task 4: Gather candidate predictor variables.
The normalized ridership change data of Task 3 will provide the “predicted” variables [Yi’s in the hypothesized equation form Yi = f(X i ,1 . . . .X i,n)]. The predictor variables (Xi’s) will be chosen to maximize the predictive power of the relationships subject to availability. The likely predictive variables will be measures of area wide or corridor specific congestion, transit system characteristics, transit usage, and metro area demographics and socioeconomics. The same predictor variables will be specifically or at least conceptually represented in the stated preference (SP) survey scenarios of Task 6. Having the same predictors included in both the observed and SP models will facilitate comparison and transferability among the two modeling approaches.
Task 5: Develop ridership change forecast models based on observed data.
Observed data ridership forecasting models will be develop using the products of Tasks 3 and 4. Analyses of relationships between predicted and predictor variables will be used to hypothesize model forms and simplicity of use will be a primary objective throughout the development process.
Task 6: Develop and administer stated preference survey.
As noted in Task 4, scenarios will be described using the predictor variable information represented in the models developed in Task 4. These are likely to include area wide or corridor specific congestion, transit system characteristics, transit usage, and metro area demographics and socioeconomics. The survey will be administered in three steps. First, a pilot application will be conducted to gather sufficient information to characterize variability in the sampled population. These estimates of variability will be used to estimate required sample sizes for the primary survey. Second, the primary survey will be administered to acquired the sample sizes estimated in the first step. Finally, the more precise variability data provided by the primary survey will be used to estimate the precision of the resulting primary survey data. Survey administration will be WEB based and targeted to groups that provide cross-sections of the potential rider population.
Task 7: Develop ridership change forecast models based on stated preference survey.
Results of the SP survey developed through Task 6 will be analyzed and employed to develop predictive models that complement the observed data predictive models of Task 5. While the observed data models represent behavior of travelers to a variety of conditions, some of which are easily characterized and some are simply unknown, the SP based models will be based upon carefully characterized scenarios. However, since conceptually the same predictor variables are contained in both models, they will complement each other and can be easily compared.
Task 8: Documentation
Work performed through the seven tasks will be documented in a comprehensive final report.
Index Terms: