Built environment's main and moderating associations with travel mode choice across trip purposes

Limited research has simultaneously investigated both the main and moderating effects of the built environment on transport mode choice across various trip types. To fill this gap, we estimate the parameters of cross-classified random intercept and random slope logit models employing Bayesian inference. We use data on 97,564 trips taken by 76,819 residents of Montreal, Canada, and develop three models to analyze three specific types of trips: commuting, school, and non-work. Our results show that the built environment of a neighborhood affects the likelihood of car travel, not just among its residents but also among visitors. For instance, doubling the POI density of a neighborhood has a dual effect: a 2.60 % reduction in the likelihood of residents opting for a car for non-work travel, and a 4.20 % decrease for those traveling to the neighborhood for non-work purposes. The built environment also moderates the effect of sociodemographic variables on mode choice; for instance, though higher-income individuals generally tend to use their automobiles to commute, public transit or proximity to downtown reduces their use of private cars. The estimated main impact of the built environment is relatively homogeneous for commuting and non-work journeys but slightly different for school trips.