Unveiling route choice strategy heterogeneity from smart card data in a large-scale public transport network

• Published in: Transportation research. Part C, Emerging technologies Emerging technologies, 2022-01, Vol.134, p.103467, Article 103467
• Main Authors: Arriagada, Jacqueline, Munizaga, Marcela A., Guevara, C. Angelo, Prato, Carlo
• Format: Article
• Language: English
• Subjects: Passenger behavior; Public transport; Route choice; Smart card data
• ISSN: 0968-090X; 1879-2359
• DOI: 10.1016/j.trc.2021.103467

•Public transport passengers use at least two type of route choice strategies.•48.8% of passengers use common lines or aggregated strategy for route choice.•51.2% of passengers use disaggregated strategy for route choice.•The latent class models allow capturing heterogeneity in route choice strategy.•Smart card data allows studying the passenger route choice behavior. We contribute to the understanding of public transport passenger route choice behavior by developing and applying methods that capture behavioral strategies by making use of smart card data. We begin by proposing the classification of possible route choice behavioral strategies in two groups: disaggregated strategies and aggregated strategies. In the former, the alternatives correspond to itineraries, which are fixed sequences of stops and public transport lines. In the latter, common line alternatives are considered, which are combinations of itineraries defined under given criteria. Almost all route choice models use consideration sets composed only of itineraries, while public transport assignment models for strategic analysis mostly use a version of the common lines approach. We postulate that this dichotomy is inappropriate and that, instead, heterogeneity exists in the route choice strategy, both between users and across contexts. With the aim of verifying this hypothesis, we first propose an indicator function constructed as the difference between expected and observed trips for a given behavioral assumption. We apply then the indicator to a case study based on smart card data from the city of Santiago, Chile, from which we find evidence of heterogeneity. We identify individuals that follow either an aggregated or a disaggregated strategy, as well as others who seem to be using a combination of both strategies. We further analyze the heterogeneity hypothesis using an integrated discrete choice and latent class approach, which we apply to the same case study. This approach involves estimating path-size logit models built with alternatives from disaggregated and aggregated strategies, as well as a latent class model built from a combination of both. It also addresses methodological challenges related to the definition of the consideration set and the correction of endogeneity. Results confirm the heterogeneity hypothesis, suggesting that the mean probability that passengers belong to the class that uses a disaggregated strategy for route choice is 51.2%, and that this heterogeneity marke