The purpose of the National Motor Vehicle Crash Causation Survey (NMVCCS) was to collect information on the vehicles, the roadways, and the environmental conditions as well as the human behavioral factors that are likely to contribute to crash occurrence. The data was collected on crashes involving light vehicles, during the period January 2005 to December 2007. The primary focus of the survey is on the events immediately prior to a crash as well as on the associated factors as described by the occupants and witnesses, reported by the police, and assessed by the NMVCCS researchers. One of the special features of NMVCCS was to collect information at the crash scene itself, thus enabling the researcher to obtain first-hand information while it was still relatively undisturbed. Due to the nature of the targeted information and the method of operation, this survey required a complex sample design to get national representation at a reasonable cost. Both time and location of a crash were considered in the selection of crashes to attain high efficiency of the on-scene survey. A two-dimensional sampling frame reflecting on both space and time of crash occurrence was used in sampling crashes from among those occurring between 6 a.m. and midnight. A probability-based sampling procedure was performed in multiple stages. In order to make the NMVCCS sample nationally representative, the inclusion probability in each sampling stage was taken into account in developing the weights. While doing so, appropriate adjustments were made for the crashes from which information could not be collected due to the operational difficulties and other challenges. This report provides details of the sampling procedure and the related operational challenges. The analytical details of the estimation methodology are also discussed. Some national estimates and their standard errors, based on the data collected during the first year of its operation, are presented. (Author/publisher)
Abstract