This study has two objectives. The first objective is to investigate the role that relative level-of-service (LOS) characteristics of travel modes play in explaining the heterogeneity in mode split for the work trip. The investigation is carried out through a comparative analysis between mode split models with and without relative LOS variables. The analytical tool used in developing the models is a relatively new statistical procedure for non-parametric analysis of data, referred to as Classification and Regression Trees. The results show that relative LOS characteristics explain a modest portion of the mode split heterogeneity. The results also show that mode split appears to vary with relative LOS at a higher rate for trips in corridors where the transit service is superior to or comparable with the auto service than for trips in corridors better served by auto. In contrast, auto ownership appears to vary with relative LOS at a higher rate for trips in corridors better served by auto.

The second objective of this study is to investigate interactions between mode split for the work trip and the mobility dimensions of residential location, employment location and auto ownership. This is done through an exploratory analysis, in which tabulations and plots of each dimension are examined separately, resulting in a multi-dimensional picture or the interactions. Travellers that have different mobility patterns are observed to have distinct characteristics. Specifically, households with different compositions or at different stages in life cycle have different mobility patterns. The observations suggest that mode choice appears to be a household decision made together with other mobility decisions, and that these decisions are highly dependent on household composition and/or stage in life cycle.

The study discusses the implications of the findings in transportation planning. Also, the study recommends some modifications to the modelling practice of mode split in light of the results of the investigation.