Estimating Fare Noninteraction and Evasion with Disaggregate Fare Transaction Data

Public transportation authorities rely on electronic fare transaction records for revenue collection, service planning, and performance measurement. When passengers make public transportation trips without interacting with the fare system, demand is underreported and fare revenue is lost. In some cases, this issue is studied through costly manual surveys that cover a small portion of stops and times. Based on disaggregate fare transaction data, this research introduces a framework and stochastic model for estimating fare noninteraction and evasion on systems without automatic passenger counting. The model produces estimates by stop and time of day that could be used by agencies to effectively target fare inspection and off-board validation resources, as well as to improve the accuracy of scaling for inferred origin-destination matrices. Applying the model to morning peak trips originating at inbound surface stations of Boston's Green Line, it is estimated that the authority loses $3,600 per weekday to fare evasion seen in about one-third of noninteractions and 9% of all trips. A card-specific parameter estimated by the model sheds light on personal attitudes toward noninteraction.