Bike Share Station Tool
How Was Ridership Forecasted?
The traditional process for bike share ridership forecasting applies simple per station or per bike averages to the number of stations or bikes in a proposed system. What’s the downside? It doesn’t account for unique demographic, land use, built environment or other factors.
Our research shows that the effects of bike sharing station networks are extremely important to ridership levels. These effects have a robust, statistically significant relationship within systems, across systems, independent of other variables, and when controlling for other demographic and spatial variables. Other critical factors in estimating bike sharing ridership include population density, retail job density and income levels.
By forecasting ridership at the station level using our ridership forecasting model, we were able to determine the appropriate size of each station and select an appropriate number of bikes and stations for the network as a whole. These estimates supported the analysis of potential membership size, capital and operating costs and revenue levels.
When asked to relocate some stations based on stakeholder response to a draft set of station locations, we used the model to quickly decide on appropriate locations for the new stations; to evaluate the effect of the relocated stations on ridership, cost and revenue projections; and to keep the Sacramento Air Quality Management District on their grant application schedule.
What is our Ridership Forecasting Model?
Through the use of the Environmental Protections Agency’s (EPA) Smart Location Database, other spatial and demographic data sets, and actual bike share ridership numbers observed by bike share systems in Denver, Minneapolis, and Washington D.C., we developed an innovative, regression-based, ridership forecasting model. This model makes station-level forecasts that are significantly more sensitive to real-world conditions than other models.
Through our original involvement in the development of the EPA’s database, a geographic Census data resource that includes 90 attributes shown to be related to travel behavior, we were able to use our specific insight of how to use the data along with the ridership numbers to develop the forecasting tool. We can now apply the model to any community and provide more accurate ridership and revenue projections for proposed bike-share systems.