A disaggregate model system of vehicle holding duration, type choice and use

Duration models are still very uncommon in transportation analysis, although they have been extensively applied in some other fields (medicine, labour market economics). The model system developed in this paper contains several submodels, one of them being a duration model for the time elapsed between two vehicle transactions. In the first wave of a new car drivers panel, questions were asked about the respondents' present and previous vehicle. The answers to these questions are used to estimate a model system with four components: (i) models for vehicle holding duration (in months); (ii) models for the choice of vehicle type; (iii) regression equations for annual kilometrage; (iv) regression equations for fuel efficiency. For the duration model several assumptions on the form of the hazard function are tested, as well as functional forms which account for heterogeneity and time-varying covariates. The type choice model is a logit model for the new car conditional on replacement. The regression equations are estimated using selectivity correction terms. In the paper we discuss the specification of the model, describe the dataset used and present the main estimation results. Finally, the estimated models are used in predicting the next wave (1 yr ahead) and in various policy simulation runs. We hope to extend this model system in the future to a dynamic simulation system which covers all types of vehicle transactions for a multi-year period.