The purpose of this research was to explore the relationship between roadway and roadside accident rates for Washington State highways to improve the Washington State Department of Transportation's (WSDOT's) process of modelling roadway and roadside accident rates and to arrive at possible improvements in the efficiency of WSDOT's safety project programming process. The project tested the use of the seemingly unrelated regression estimation (SURE) model to model the roadway and roadside simultaneously. The theoretical advantage of the SURE approach is that it does not impose any a priori assumptions on the explicit linkage between roadway and roadside accident rates, and there is no theoretical support for explicit linkage, either. The data used to derive this model were from a random sample of 500 one-mile sections from the Washington State highway system. Traffic data included traffic volumes, truck compositions, annual average daily traffic, traffic speeds, and other relevant information. Geometric data included lane, shoulder, median, curve, and intersection information. Historical weather data such as monthly precipitation and temperature were collected from the National Oceanic and Atmospheric Administration database. In comparing significant explanatory variables between the roadway accident rate and roadside accident rate models, very few variables were common. This confirms that it is preferable to specify separate functional forms for roadway and roadside accident rates. Empirical results indicated that correlation between roadway and roadside accident rates was insignificant, indicating that efficiency gains from the SURE model would be minimal. The important finding from a programming standpoint is that jointly modelling the roadway and roadside simultaneously would not result in significant efficiency improvements in comparison to the current state-of-the-practice in Washington State. (Author/publisher)
Abstract