A joint panel binary logit and fractional split model for converting route-level transit ridership data to stop-level boarding and alighting data. (September 2020)

• Record Type: Journal Article
• Authors: Rahman, Moshiur; Yasmin, Shamsunnahar; Eluru, Naveen
• Is Part Of: Transportation research. Volume 139(2020)
• Page Start: 1
• Page End: 16
• Publication Date: 2020-09
• Subjects: Transit ridership -- Alighting -- Boarding -- Bus stop -- Route-level -- Joint model -- Binary; Transportation -- Research -- Periodicals; 388.011
• DOI: 10.1016/j.tra.2020.06.015
• Languages: English
• ISSNs: 0965-8564

Abstract

Highlights: The study relates route-level transit ridership to stop-level ridership. A panel joint model of binary logit and fractional split model is proposed. Stops with ridership vs. no ridership are examined by using binary logit model. Fractional split model is used to identify the proportion of route-level ridership. The result shows the feasibility of getting stop-level data with route-level data. Abstract: Detailed ridership analytics requires refined data on transit ridership to understand factors affecting ridership (at the stop and/or route-level). However, detailed data for stop-based boarding and alighting information are not readily available for the entire bus system. Transit agencies usually resort to compiling ridership data on a sample of buses operating on the various routes. We propose an approach to infer stop-level ridership for transit systems that only compile route-level ridership information. A joint model structure of binary logit and fractional split model is proposed to estimate stop-level ridership data sourced from route-level ridership. The model is developed for the Greater Orlando region with ridership data for 8 quadrimesters (four-month time periods) from May 2014 through December 2016. In the presence of repeated data measures, panel version of the joint econometric models for boarding and alighting are estimated. The development of such an analytical framework will allow bus systems with only route-level ridership data to generate stop-level ridership data. The model results offer intuitive results and clearly supports our hypothesis that it is feasible to generate stop-level ridership with route-level ridership data. For transit agencies with ridership data at the stop-level, the proposed model can also be employed to understand how various stops along a route interact with one another toward affecting route-level ridership contributions.