Sketch Model to Forecast Heavy-Rail Ridership

1. Sketch Modelto ForecastHeavy-Rail Ridership Len Usvyat1, Linda Meckel1, Mary DiCarlantonio2, Clayton Lane 1 – PB Americas, Inc. 2 – Jeffrey Parker & Associates, Inc.

2. Research Goals • Ridership estimation is key to new transit projects • Catch-22 of cost versus outcome: • Four-step models are difficult and expensive • Outcome may prove project unfeasible • Agencies need an easy-to-apply, inexpensive sketch level tool

3. This Tool versus Others • Geared to heavy-rail stations outside of downtown • Computer software needed to calculate the results: • MS Excel • ArcGIS (not required) • Other tools our team developed: • Light Rail model • Commuter Rail model

4. What else is out there? • Pushkarev & Zupan: densities for various transit modes • Cervero: drawbacks of four-step model and importance of sketch-level tools • The FDOT TBEST model: patronage by stop • TCRP Report 16: commuter & light rail models • Update to TCRP Report 16 published in 2006: models for commuter and light rail • ARRF: FTA suggested tool

5. Data Collection • Transit agency-specific boardings (2004-2006) for all U.S. heavy-rail transit stations (N=474) • MPO data for demographic data (same year) • U.S. CTPP data for the year 2000, where no MPO data is available

6. Heavy-Rail Systems Included • 10 cities • 11 systems (NYC Subway excluded) • 32 heavy rail lines • 474 stations

7. Variables Analyzed • Station area demographics • Station specific transportation attributes • Corridor demographic characteristics • Metro area demographic and transportation attributes

8. Station Area Demographics • 1/4, 1/2, 1, and 2-mile radii around each station: • Employment (all jobs) • Population and population 16+ years old • Population divided by employment • Average household size • Median household income • Average number of vehicles per household • Zero-car households, households with cars • Zero-car households divided by households with cars • One-car households, two-car households, three-car plus households

9. Station Specific Transportation Attributes • Whether a station is a secondary downtown (yes or no) • Number of buses connecting to the station • Connection to other rail system (yes or no) • Number of rail lines connected to this station • Availability of parking (yes or no) • Whether it is a terminal station (yes or no) • Indication if the station is a special transit attractor • Cash and commuter fare to downtown (yes or no) • AM, PM, Midday peak and weekend headway (minutes) • Hours transit system is operated per 24-hour period • Distance, time, and speed to primary & secondary downtown • Distance to nearest station (miles)

10. Corridor Demographic Characteristics • Zero-car households divided by households with cars • Households with cars along the corridor • Zero-car households along the corridor • Total employment along corridor • Total population along corridor • Total employment divided by total population along the corridor

11. Metro Area Demographic and Transportation Attributes • Central Business District (defined in by Lane) • CBD area, ratio of CBD area to total metro area • CBD employment and employment density • CBD employment divided by total metro area employment • Metropolitan area (as defined by MPO) • Metro area • Metro employment • Metro population • Median household income in the metro area • Vehicles per household in the metro area

12. Analysis • Variable review • Continuous • Discrete • Categorical • Normality of independent and dependent variables (natural log) • Correlation coefficients • Multiple linear regression

13. How to define a CBD? • Log of employment density (“ED”) by TAZ • Mean and standard deviation of ED for the entire region • Map TAZs whose ED is at least 1.5 and 2 standard deviations above the mean • “CBD area”: contiguous TAZs whose ED is 2 and 1.5 standard deviations above the mean

14. Station Area Coverage • Non-exclusive versus exclusive station areas • Non-overlapping radii do not double count boarding drivers • Correlation is significantly improved

15. Station Area “Donuts” • Variation in density around the station and boardings • Correlations with boardings p<0.05

16. Heavy-rail station boardings = -972 + 1,625 * [if this is a terminal station, 0 if not] + 1,346 * [if this is a secondary downtown, 0 if not] + 1,710 * [if this is a special attractor station, 0 if not] + 70 * [number of buses connecting to this station] + 884 * [if there is parking available, 0 if not] + 2,271 * [if there is connection to other rail, 0 if not] + 115 * [distance to downtown, in miles] - 2,792 * [ln (midday headway in minutes)] + 0.024 * [CBD density, in employees per square mile] + 0.224 * [employment within 0.25 miles of the station] + 0.133 * [employment within 0.25 to 0.5 miles of station] + 0.219 * [population within 0.25 to 0.5 miles of station] + 5,938 * [empl within 2 miles of entire line div by pop]

17. Results • Station level: r2=0.61 • Line level: r2=0.70 • Metropolitan Area level: r2=0.81

18. Applying the Model

19. Next Steps • Census tracts • Downtown station ridership • Age of the system (time dependent analysis) • Size of parking lots • Other station characteristics (elevators, cleanliness, station attendants, underground/aboveground) • Time of day analysis • Taking transfers into account • Buses versus bus routes

20. Conclusion • Follow the application guidance • Calibrating the model • Means and standard deviation • Urban character is crucial • Linear model drawbacks • New tools from FTA or other agencies • LRT and Commuter Rail models are available in TRR 1986 • Try it out and let us know your outcomes!

21. Questions Len Usvyat 215-209-1239 usvyat@pbworld.com