Sketch Models to Forecast Commuter and Light Rail Ridership: Update to TCRP Report 16

Document Type

Journal Article

Publication Date


Subject Area

land use - planning, ridership - commuting, ridership - forecasting, mode - rail, mode - tram/light rail


Transit users, Transit riders, Transit projects, Sketch planning models, Ridership, Reverse commuting, Regression analysis, Regression, Railroad commuter service, Patronage (Transit ridership), Light rail transit, Commuter rail


Ridership potential is among the most valuable attributes to understand about a proposed light or commuter rail line during early stages of project development, yet few nationally relevant sketch-level tools exist for feasibility analyses. Research was done to develop a nationally applicable, sketch-level ridership forecasting tool for light rail and commuter rail. The study collected current ridership, demographic, and transportation system data from 17 U.S. regions, including 58 commuter rail corridors, 22 light rail corridors, and 1,218 stations, and it tested 163 possible explanatory variables. The effort yielded two multivariable regression equations that show close relationships between actual and predicted values, with adjusted R-squared values of 0.97 for commuter rail and 0.92 for light rail. The models also validate well to existing rail systems in six regions and successfully predict actual line ridership with adjusted R-squared values of 0.84 for commuter rail and 0.47 for light rail. The new models improve on previous tools by introducing sensitivity to reverse-commute trips and special transportation hubs or ports and by introducing transportation-system variables to measure the impacts of faster operating speeds, lower fares, or shorter midday headways. The models require readily available data and use simple form, making the tool technically accessible to a large number of transportation planners. An apples-to-apples comparison with similar sketch models developed nearly a decade ago in TCRP Report 16 shows that the new tools perform consistently better, explaining between 31% and 213% more variation in observed rail boardings among the same validation data sets.