Rail ridership estimates are a critical indicator for the potential viability of commuter rail expansion. Estimating ridership on extension services can be difficult when the new service exceeds the longest line in the current system. In these cases, it is useful to have controls on the potential market. This paper presents the development of a station-level ridership model to analyze alternatives serving a corridor that extends the MBTA Boston to Lowell commuter line to Nashua, Manchester, and Concord, New Hampshire.
Commuter rail ridership data analysis shows that the commuter rail mode share increases as distance from the Central Business District (CBD) increases. This presents a challenge for models estimated from station-level data to incorporate a travel time or distance variable because the positive correlation between distance and ridership can often result in positive coefficients on distance variables or variables that serve as a proxy for distance, such as speed or travel time. In this project the proposed extension exceeds the service extent of the current commuter rail system so it was critical to avoid unreasonable distance and time parameters in the forecasting model, which would overstate the travel market at outlying stations.
The American Community Survey based Census Transportation Planning Products (CTPP) Journey to Work (JTW) dataset captures the travel market implicitly. JTW trips to the Boston CBD were associated with the nearest commuter rail station to establish the station level commuter travel market. Two sets of models were estimated: one using JTW; the other using a population buffer around the station to define the travel market. The JTW model was preferred because it explained much of the ridership variability (adjusted R-squared values of 0.758) and is driven by the actual travel market. Accounting for non-work travel was done by incorporating socio-economic variables and station attributes into the model. This paper describes the model development process and presents estimation results and forecasts from both sets of models.