The predictive models in the Highway Safety Manual (HSM) is based on the Safety Performance Functions (SPFs), which is a statistical regression model based on observed crash data from similar facility types and estimates the predicted average crash frequency the base conditions. To account for differences between the base conditions and the specific conditions of the facility site, accident modification factors (CMFs) are utilized to adjust the prediction to account for the geometric design and traffic control features of the specific site.

SPFs in the HSM were developed using historic crash data collected over a number of years at sites of the same facility type in different states. Because the SPFs provided in the HSM are developed using data from other states it is more than likely that they cannot be transferred directly to other locations and times. Thus HSM’s predictive model often needs to be calibrated to capture local state or geographic conditions. Moreover, accident frequencies for similar facility types can also vary from one jurisdiction to another, since their locations differ in climate, driver population and characteristics, accident reporting threshold, accident reporting practices and other contributing factors.

To let the SPFs better accommodate the local data, two strategies are usually taken:

     • The first strategy is to calibrate SPFs provided in HSM so that the contents of HSM can be fully leveraged.

     • The second strategy is to estimate location-specific SPFs regardless of the predictive modeling framework in the HSM.

The main objective of this research project is to either calibrate the SPFs provided in the HSM using New Jersey (NJ) data or develop new NJ-specific SPFs for at least twenty different facility types.