Transit agencies need up-to-date ridership forecasts to estimate fare revenue and make regular service changes. However, ridership forecasting is not easy, particularly for mid-size and small-size transit agencies that do not have the resources to develop complex models, considering the many factors that affect ridership levels. 

To help  transit agencies in this need, this project will develop an open-source forecasting tool for estimating ridership. The forecasting model would use regression techniques and will include variables such as service levels (e.g., vehicle revenue miles), population, employment, gas prices, and telecommuting. Data needed to create the forecasting model will be compiled and will include historical route-level ridership and service provision data that transit agencies regularly collect. Other publicly available data from the US Census Bureau (e.g., population), the US Bureau of Labor Statistics (e.g., employment/unemployment levels), and the Energy Information Administration (e.g., gas prices) will also be obtained. Using this data, a model will be developed as a script using software such as the open-source statistical program R. The model outputs will be ridership response (elasticities) to changes in different internal and external factors, which would be used for ridership forecasting. The forecasting model will be web-based and can be used by any bus-based transit agency to create forecasts specific to that agency's region. The Nashville Public Transit system and the University of California Transit system at Davis will implement the developed tool. A user guide and training videos will be prepared that will describe in more detail how to use and customize the tool.

The proposed open-source tool will offer several benefits to the transit agencies. First, the model will forecast ridership at the route level, which will provide agencies flexibility to forecast ridership based on changes in specific areas in the city that might affect only one or a few routes. Second, these forecasts can be updated annually with minimal effort/time and with little to no cost. Lastly, the tool would be customizable, which will allow agencies to use additional variables (e.g., reliability) based on their local data availability.