Abstract
Travel behavior changes over time and such changes might be not the same across individuals, either. In the context of stated choices, such behavioral dynamics might become more complicated. This paper proposes a new DGEV (Dynamic Generalized Extreme Value) model with heterogeneity for not only SP data, but also the combined SP/RP data to further improve the predictability of the stated preference (SP) data. The DGEV model is used to simultaneously capture initial conditions, state dependence, and future expectation as well as time-varying tastes. Heterogeneity is measured with respect to the observed individual tastes to levels of travel services. An empirical analysis is conducted using a 4-wave panel data of travel mode choice collected in Hiroshima City in 1987, 1990, 1993 and 1994, where a new transit system was targeted. It is confirmed that the proposed models are effective to capture heterogeneous dynamic in travel mode choice behavior.
