Immigrant Mode Choice in Canadian Cities: Mixed Logit with GPS Panel Data and Joint RP-SP Estimation

Abstract

This study investigates how immigrant status and integration level shape transportation mode choice in Canadian cities. The analytical framework combines four elements: a continuous multidimensional integration index, mixed logit models with systematic immigrant-linked taste heterogeneity, joint revealed–stated preference estimation, and a GPS-based custom mobile app panel dataset of approximately 80,000 trip observations and 622 stated preference scenarios from 100 participants in Toronto and Montréal. A multi-stage data processing pipeline transforms raw GPS traces into estimation-ready choice sets with decomposed transit journey attributes. The mixed logit specification reveals that sampled immigrants exhibit approximately 66% lower sensitivity to in-vehicle travel time than Canadian-born residents, while higher integration is associated with reduced transit use. Counterfactual simulations demonstrate that access time reductions are approximately three times more effective than fare reductions in shifting car commuters to transit. Five-fold cross-validation yields 80–82% prediction accuracy across specifications.