This report proposes a technique for estimating the spatial variation of average household vehicle-miles traveled (VMT). The dependent variable, VMT, is estimated for block groups in the metropolitan regions of two cities - San Francisco, CA, and Boston, MA. The independent variables were obtained from the U.S. Census and include variables such as commute time to work and percentage of workers using public transit. Model-predicted values for zip-code-level VMT demonstrate a correlation coefficient of 0.90 with values imputed from Massachusetts state vehicle inspections data. These findings are proposed as evidence that "snap-shot" estimates of urban and regional variations in household VMT may be possible through the manipulation of freely available Census data. However, the results do indicate spatial autocorrelation and future estimates must take into account such spatial anomalies in estimating VMT.
Abstract