In the United States, transportation uses approximately 26% of the nation energy consumption, and contributes 27% to the total greenhouse gas emissions. Given the important role of private transportation, it is urgent to develop effective and innovative quantitative methodologies to support public authority decision making. Although a substantial body of the literature investigates household vehicle ownership decisions - vehicle holding, type and usage; the majority of the existing studies focuses on only one of these three decisions, is often limited to specific geographic areas and is not calibrated with the most recent travel survey data available.

This dissertation proposes a modeling framework that is able to incorporate all the three decisions simultaneously, and takes into account the correlation across the discrete variable (vehicle holding) and the continuous variable (miles traveled). In this integrated discrete-continuous choice model, a multinomial probit model is used to estimate household vehicle holding decision, while a multinomial logit model is adopted to estimate the vehicle type decision. The vehicle usage decision variable is integrated with the discrete variables by adopting an unrestricted correlation pattern between the discrete and the continuous variables. Since the problem has no closed-mathematical form, I use estimation techniques based on Monte-Carlo simulations and numerical computation of multivariate normal probabilities to derive the solutions.

Though a number of studies have demonstrated that unordered behavioral models outperform the ordered mechanisms for vehicle holding decisions, those comparative studies were only conducted for the discrete decisions concerning vehicle ownership. Therefore, an ordered discrete-continuous model structure is developed, in which an ordered probit replace the multinomial probit for the vehicle holding decisions. Both the unordered and ordered structures are estimated and validated on the 2009 National Household Travel Survey data. Ordered models are in general preferred to unordered models for the lower computational costs to derive the analytical solutions. However, results from operational data show that the unordered discrete-continuous models always outperform the ordered ones in terms of both statistical goodness of fit and predication capabilities.

The proposed modeling framework is then applied to the entire nation and a system of national vehicle ownership models is derived. The models are calibrated using the 2009 National Household Travel Survey data, each combining four regions (Northeast, Midwest, South and West) and three area types (urban, suburban and rural). In addition, the models are applied to the 2009 American Community Survey data for six randomly selected counties/areas. The prediction results for the six counties/areas demonstrate the prediction capability of the models calibrated. The national models are valuable both for the national level (to evaluate federal policies) and small areas (that lack local household travel survey data). The results also demonstrate that the integrated discrete-continuous framework has good prediction capabilities in modeling household vehicle ownership decisions.

Lastly, the dissertation estimates a discrete-continuous model for the Washington D.C. Metropolitan Area and analyzes the impact of improved bus and metro services on household ownership and use decisions in that area. The 2009 National Household Travel Survey data and the General Transit Feed Specification data are integrated, and then both spatial and temporal measurements of transit services are created on the Census Tract level. The results show that improved transit is a significant factor in household vehicle ownership choices and that the proposed methods are able to effectively predict changes in vehicle ownership and usage with respect to the transit improvements.

In conclusion, the dissertation contributes to both the theoretical analysis and the practical applications of the household vehicle ownership problem. The results provide decision makers with advanced quantitative methods that are able to effectively analyze policies, aiming at promoting greener travel behavior and at mitigating energy consumption and emissions.