Vehicle ownership or transaction models in the literature are used for a variety of purposes and their applications are considerably increasing. The recent dramatic increase of fuel price has even further stimulated the need for such models. Additionally, with increasing concerns about environmental issues and fuel consumption, a reliable vehicle ownership model has become a vital tool for national, regional and local agencies.

Some of the fundamental and most complex decisions made by households include the relatively long-term decisions on work and home location and automobile ownership. There are complex interdependencies among these three decisions which jointly determine household mobility that can influence the entire household travel pattern.  One can postulate that job and home location choices can influence individual’s vehicle ownership and utilization behavior. Therefore, transportation system cannot be accurately analyzed if the correlations between these influential elements are ignored. To a great extent such decisions are made either by the household or influenced by the household structure rather than made exclusively by individual household members.  Therefore, it is also important to incorporate group decision making process within such modeling framework.

One can argue that the household vehicle fleet size in a target year can be estimated as a function of current total number of vehicles in the fleet plus all the vehicles that will be purchased, minus those vehicles that will be disposed out of the household fleet of vehicles. Therefore, understanding the timing of the transaction is an important element for predicting the number of vehicles within a household fleet.  While typically, modeling the transaction time is usually ignored in household vehicle ownership models.

This study concentrates on presenting the results of developing a system of hazard-based equations in which timing of the residential relocation, job relocation and vehicle transaction are the endogenous variables. This study utilizes data from the Puget Sound Transportation Panel Survey (PSTPS) along with the land-use, network skims, and built environment attributes from the Seattle Metropolitan Area. The PSTPS is a panel survey that covers 10 waves starting from 1989 until 2002 and covers Seattle and its surrounding areas. In this study, the interaction between the residential relocation and the job relocation is initially modeled at a lower level of the proposed framework while the impact of the correlated job and residential relocation on vehicle transaction timing is modeled at an upper level of the modeling framework. Timings of the endogenous variables are estimated by a hazard-based duration formulation in a system of simultaneous equations. Moreover, unobserved heterogeneity is included among the exogenous variables through a normal distribution in all of the equations. Two different approaches are utilized in order to estimate the coefficients of the models. First, coefficients are estimated in a hierarchical system of equations, and second, the coefficients are estimated in a joint simultaneous system of equations. Then the results of both approaches are presented, compared, and discussed.