Abstract
Conventional dynamic discrete choice models have not satisfactorily represented individual heterogeneity due to the insufficient utilization of individual attributes and behavior data. This paper therefore attempts to establish a new dynamic travel mode choice model simultaneously incorporating cross-sectional and longitudinal heterogeneity by decomposing taste parameter of travel level-of-service over time. Throughout an empirical analysis sing revealed preference panel data, it is confirmed that the established model can flexibly describe choice behavior dynamics with a higher accuracy. By introducing the above-defined taste structure into dynamic GEV model, it is shown that the closing timing of panel survey affects the model estimation results as well as initial condition of panel survey.
