Modeling constrained destination choice for shopping: a GIS-based, time-geographic approach

For accurate prediction, a shopping destination choice model must rely on a realistic representation of available opportunities. Thus, spatio-temporal constraints are indispensable in specifying a realistic choice set. Failure to take into account such constraints results in misspecification of shopping alternatives, erroneous inferences, and misunderstanding of individual travel behavior. However, due to lack of data, heavy computational burden, and algorithm complexity, spatio-temporal constraints have often been ignored in shopping destination choice modeling. In our study, we use the potential path area, which is the projection of a space–time prism onto a plane, to determine an individual’s destination choice set given their spatio-temporal constraints. In this way, the destination choice set is a more realistic representation of the shopping alternatives available to the individual. We select 616 shopping trips from a travel survey conducted in five counties of the Louisville KY-IN MSA in 2000, and aggregate shopping opportunities to traffic analysis zones (TAZs) for analysis. Multinomial logit models are estimated to understand the determinants of shopping destination choice. The results are compared to those of a conventional, unconstrained destination choice model.