Investigating Contextual Variability in Mode Choice in Chicago Using a Hierarchical Mixed Logit Model

Abstract

In this paper, a hierarchical random-coefficient mixed logit model is applied to quantify variability in commuters’ mode choice in the Chicago metropolitan area, especially concerning the contextual variability by the traits of census tract of residence. It is found that individual mode choice behaviour varies considerably across residential locations. Moreover, the contextual effects are found to modify the marginal utility of mode choice. Especially, in-vehicle travel time and gasoline cost are significant covariates of census tract traits (such as percentage of blue-collar residents, ethnicity). Furthermore, random variation is present even after both contextual and individual traits are controlled for, suggesting intrinsic randomness in individual mode choice. The hierarchical structure of quantifying contextual variability proves to be a useful tool in capturing intrinsic heterogeneity in mode choice. The study findings have important implications for integrated land use and transport planning especially at the geographical levels below that of the region.