Abstract:In this study, we propose a trivariate latent-class modeling framework to jointly study ridehailing usage frequency, vehicle ownership, and expectations to change vehicle ownership. The proposed model, in addition to accounting for parameter heterogeneity through outcome-variable-specific latent segmentations, allows for an insightful behavioral interpretation of the relationships among the variables that indicate membership in the latent segments associated with each outcome variable. Our results point to more nuanced relationships between the three variables of interest and the external factors associated with them than what most other studies in the literature have revealed so far. More specifically, we see a less straightforward relationship between age and ridehailing usage frequency, for which other studies have generally pointed to a negative relationship. Our results reveal two latent clusters of similar average age who show drastically different ridehailing usage frequency. Furthermore, although we see evidence of a negative association between vehicle availability and ridehailing usage frequency, our latent class framework again reveals two clusters with similar vehicle availability but different ridehailing usage, pointing to the influence of other factors such as attitudes and built environment in differentiating their ridehailing usage. Regarding the relationship between ridehailing usage and expectations to change vehicle ownership, our results show that, of the two clusters with similar vehicle availability and age, the one with higher ridehailing usage is less likely to expect an increase in vehicle ownership within the next three years. This result shows some promise for the future impact of ridehailing in containing increases in car ownership.