Abstract:Estimation of congestion costs, presumed to be one of the largest external costs of automobile travel, is typically based on a single value of time delay for motorists in metropolitan areas. However, the estimation may be wrong if the profiles of motorists are different at different times of day. This study uses a survival model to examine the demographic and socioeconomic profiles of motorists at different times of day at congested locations in southern California, by using on-road remote-sensing measurements and license plates images obtained in 2007 and 2008 by the California South Coast Air Quality Management District (SCAQMD). More than 80,000 vehicles were observed from fifteen selected study sites over fifteen days. Their plates, through anonymized registration records, revealed addresses at the census block group level, which have homogenous profiles by construction. Motorists’ profiles at different times of day display large variation, however, according to extended Cox model with a non-parametric baseline hazard, which is used to accommodate both the time-invariant and time-varying effects of the covariates. This study thus proposes a new approach to examine heterogeneity among motorists.