The development of a system for using Nationwide Personal Transportation Survey (NPTS) data to estimate regional or local travel behavior--vehicle and person trips and miles of travel--is detailed. This system can be used by state or municipal transportation planners. The census tracts were classified into groups, or clusters, that tend to be homogeneous for individual travel behavior. These census tract clusters were based on household income, employment rate, number of household vehicles, and area type (urban, suburban, or rural). NPTS data were used to estimate driving characteristics for each of the clusters derived in the classification step. How well the goal of estimating regional or local travel characteristics was achieved by using standards computed from an independent survey from Baton Rouge, Louisiana, and by using independent data from three add-on components of the NPTS itself was assessed. Estimates computed from the NPTS data by using the census tract cluster method were compared with estimates computed from the standards and with estimates computed from the NPTS data by using competing methods in which households were classified by size of the metropolitan statistical area, census division, or census region. It was found that in most cases the census tract clustering method predicted travel better than the other methods, with small sample sizes generally being the cause when the census tract method was not the best.
Abstract