A Dynamic Choice Model to Estimate the User Cost of Crowding with Large-Scale Transit Data

Abstract

Abstract Efficient mass transit provision should be responsive to the behaviour of passengers. Operators often conduct surveys to elicit passenger perspectives, but these can be expensive to administer and can suffer from hypothetical biases. With the advent of smart card and automated vehicle location data, operators have reliable sources of revealed preference (RP) data that can be utilized to estimate transit riders’ valuation of service attributes. To date, effective use of RP data has been limited due to modelling complexities. We propose a dynamic choice model (DCM) for population-level longitudinal RP data to address prominent challenges. In the DCM, riders are assumed to follow different decision rules (compensatory and inertia/habit) and temporal switching between decision rules based on experience-based learning is also formulated. We develop an expectation–maximization algorithm to estimate the DCM and apply our model to estimate passenger valuation of crowding. Using large-scale data of 2 months with over four million daily trips by an Asian metro, our DCM estimates show an increase of 47% in passenger’s valuation of travel time under extremely crowded conditions. Furthermore, the average passenger follows the compensatory rule on only 25.5% or fewer trips. These results are valuable for supply-side decisions of transit operators.