Estimating spatial interdependence in automobile choice with multilevel data

Traditionally, researchers studying transportation choice have used data either acquired from household surveys or broad, region-wide aggregates. Problems with survey collection processes are well documented, including consumer resistance, rising costs and limited regional coverage. Aggregate data alone may not offer enough variation or provide sufficient observation of the choice making process. I apply a technique commonly used in health outcomes research, the use of census proxies, to the field of transportation research, facilitating the analysis of any geographic region while avoiding the necessity of conducting additional independent surveys. Specifically, in Chapter 2 I use the 2000 Bay Area Travel Survey (BATS) and the contemporaneous U.S. Census file to evaluate the use of census proxies in applied transportation research. My results indicate that a vehicle choice model composed of proxies effectively approximates survey model parameters, and is attractive in terms of traditional selection criteria and prediction.;The spatial proximity between individuals has been found to affect their likelihood of exhibiting herd behavior in the selection of travel mode choice. In Chapter 3, I use spatial econometrics to evaluate whether consumer interaction influences automobile choices in BATS, and discuss the impact of uncontrolled spatial effects on statistical inference. I find evidence for clustering in the proportionate ownership of several different auto types after controlling for potential confounders, and that spatial effects influence the probability of selecting certain car types.;In Chapter 4, I search for spatial effects in Fresno County Department of Motor Vehicles (DMV) micro-level registration records augmented with socioeconomic and demographic information from the United States census. The resulting dataset is essentially a synthetic survey composed of vehicle characteristics and proxies for household attributes. Again, I demonstrate that auto type ownership is spatially clustered. I address the confounding role of self-selection in area wide auto purchases by accounting for the decision to switch type ownership. In effect, I control for the possibility that individuals with like preferences congregate in certain regions, and estimate the effect of consumer interaction on the probability to select a given car type.