ESTIMATING REGIONAL AUTO OWNERSHIP MODELS USING THE 1990 CENSUS PUBLIC USE MICRODATA SAMPLE (PUMS)

Disaggregate (household-level) auto ownership choice models have typically been estimated using large-scale cross-sectional household travel surveys. These travel demand models will typically stratify housholds into households owning zero, one, or two-or-more vehicles within the household. This basic market segmentation is critical in the estimation and application of a regional set of disaggregate travel demand models. An alternative regional data set for estimating disaggregate auto ownership choice models is the 1990 Census Public Use Microdata Sample (PUMS). The PUMS are two disaggregate files of individual 1990 Census records (household and population characteristics) of either 1% of an area's households, or 5% of an area's households (the 1% and the 5% Sample). Disaggregate auto ownership choice (logit) models have been estimated based on the PUMS data files for the nine-county San Francisco Bay Area. Disaggregate validation is reported at the Public Use Microdata Area (PUMA) and by other market segments. These models are also compared to disaggregate models based on the 1981 and 1990 Metropolitan Transportation Commission (MTC) household travel surveys. Strengths and weaknesses of both approaches - PUMS versus household travel surveys - are discussed. The primary weakness of the PUMS is the lack of neighborhood characteristics, such as land use density or accessibility measures, at a fine enough geographic level (i.e., regional travel analysis zone). Transferability of the model estimation methodology to other metropolitan regions is discussed.