University of Vermont Transportation Research Center

Document Type


Publication Date



The 2010 Highway Safety Manual (HSM) developed by AASHTO provides predictive equations for quantifying the safety effects of planning and designing roadway alternatives. These equations have been developed based on data sets that are nearly 20 years old from a small number of states, so they must be calibrated to conditions in a specific state in order to ensure that the predicted number of crashes on a state’s infrastructure are accurate. The purpose of this project was to develop calibration factors (CFs) and updated safety-performance functions (SPFs) for the undivided, two-lane, two-way rural road (TLTWRR) predictive models in the HSM for Vermont. Calibration-factor calculations were conducted for the entire state together, and also for two sets of geographic divisions, to investigate the effects of regional variations on crash prediction in Vermont. The calculated calibration factor for undivided TLTWRR segments (2U) in Vermont was found to be 0.298, while the regional CFs varied from 0.214 to 0.367. A trend toward decreasing crash rates over time may be reflected in the relatively low CFs calculated for Vermont’s 2U segments. Calculated CFs for TLTWRR Intersections in Vermont were 0.448 for both three-leg (3ST) and four-leg (4ST) stop- or yield-controlled intersections, and 0.568 for the four-leg signalized (4SG) intersections. These CFs are similar to the average of the states reviewed in this study. Re- estimation of SPFs for the 2U, 3ST and 4ST site types resulted in new SPF equations that outperformed the use of CFs, as measured by the Freeman-Tukey (FT) R2 measure. The re-estimated SPF for the 4SG site type was not statistically viable. Therefore, the default equation in the HSM should continue to be used with CMFs and the CF to estimate predicted number of crashes at 4SG sites.