Abstract
The rapidly-evolving landscape of transportation technologies requires that we re-examine our transportation modeling tools and their ability to reflect the likely impacts of these technologies across society. Traditional auto ownership models, among other components of travel demand models, are designed based on the assumptions that one’s transportation alternatives are known and deterministic. However, this assumption is in tension with decades of empirical evidence that documents the influence of various constraints on decision making. Further, the effects of these constraints on choice outcomes, may vary systematically; access to a given transportation alternative within the choice set may be tied to a set of decision-maker characteristics.
In this study, the researchers implement a choice set formation model framework to measure potential variations in auto ownership outcomes, with emphasis on constrained travelers. The model specifications test various factors that have been found in past empirical studies to influence households’ ownership choice alternatives, including financial constraints. Results show that the choice set formation model specification with constraint variables offers improvements in model performance, particularly for constrained subgroups, relative to the conventional model specifications. The choice set formation model offers reductions in prediction error rates, with 37% and 29% lower error rates for Black households and households with lower educational attainment, respectively.
These results demonstrate the reality of considerable differences in model performance across subgroups. While further study is warranted to understand implications for transportation demand modeling practices, this article argues that the use of conventional model specification may result in harmful consequences for constrained travelers when their travel behaviors and outcomes are poorly captured or distorted.
