Revisiting built environment and vehicle kilometer traveled: Does car ownership matter?

Although the literature has extensively evaluated the relationships between the built environment (BE), car ownership, and vehicle kilometers traveled (VKT), there is a lack of research on the causal effect of car ownership on VKT. Moreover, little is known about whether and how the BE and VKT connections differ between car and non-car owners. Two machine learning models, namely double machine learning and gradient boosting decision trees, are applied to fill the above gaps based on large-scale survey data from Changchun, China. The findings show a causal relationship between car ownership and VKT. The average treatment effect of car ownership accounts for 17.18 % of the difference in VKT between car and non-car owners. All BE factors at both residential and work locations exert nonlinear associations with VKT. Moreover, the associations differ between the two groups. The findings offer implications for refined policymaking and BE planning.