Modeling spatial and social interdependency effects on commuting mode choice

• An autoregressive mode choice model is constructed to capture interdependency effects. • Interdependencies derive from both spatial proximity and attitudinal similarity. • Direct policy impacts can be separated from indirect effects of interdependency. • Social interactions account for 40 % of total policy effects on public transit use. • Travel time and bus stops proximity boost transit use more than travel costs changes. In daily life, individuals are influenced by the behaviors of others. The question of how far-reaching this social influence extends to travel behaviors has received significant attention in recent decades, through the capture of dyadic interaction effects that may exist among individuals. Along these lines, in the current study, we apply a Spatial-Attitudinal Probit Model (SAPM) that assumes an autoregressive lag structure in the utilities underlying individuals’ travel mode choice, specifically focusing on the choice between car and public transportation for commuting trips. Notably, the magnitude of interdependency among decision agents is measured by a global weight matrix, accounting for a dual source of influence: (1) spatial proximity, measured as the Euclidean distance between individuals’ residential locations, and (2) attitudinal similarities, specifically perceptions regarding sustainable mobility and environmental awareness. To our knowledge, this represents the first application of an autoregressive travel mode choice model accounting for both geographical and attitudinal proximity as simultaneous sources of interaction. The dataset for our analysis includes 2,347 observations, corresponding to the one-way commute trips of 2,347 individuals, as reported during a survey conducted between October 2019 and January 2020 in the metropolitan area of Cagliari, Italy. Our results reveal the significant role of social autoregressive parameters and the presence of interdependency effects among individuals’ commute mode choices. The utilization of a social lag structure allows for the separate identification of direct and indirect effects of explanatory variables. Notably, around 40% of the total effect is attributed to the indirect effects arising from individuals’ social interdependencies. This finding holds important implications for evaluating and planning potential future measures aimed at increasing public transit usage.