On behalf of the internal audit group at a large regional transportation agency, we designed a stratified statistical sample for determining the true rate of toll evasion on electronic Radio Frequency Indentification (RFID)-based toll collection systems. The system is owned by a multi-state consortium, thus subject to interagency settlements, and billing for uncollected tolls had varying degrees of success. At the same time, the agency was actively considering gateless tolling, and thus needed to understand potential revenue losses from different types of toll evasion. Each electronic transaction selected for audit required laborious manual retrieval and review of video footage; at the same time, toll evasion is presumed to be a rare event, making its true rate difficult to determine by statistical sampling. Through the use of statistics, we minimized the number of events that required reviewing but retained confidence in the results, and helped them develop upper and lower bounds for toll evasion rates and resulting revenue losses.
