Quasi-induced exposure: Issues and validation

Central to roadway safety is the ability to determine roadway users or roadway sections that exhibit characteristics that could be noted as less safe. Estimating crash rates is one of the most common ways to assess the relative risk of road users or road facilities. The traffic safety community can then act to improve road safety by applying the knowledge gained from such studies. The frequency of crashes for any given roadway, driver, and environmental conditions can be used in the numerator for calculating such crash rates. These frequencies can be determined with acceptable accuracy from existing databases. However, accurate estimates of a drivers exposure for the same variables are difficult or impossible to be obtained from the available data. This creates a problem not only in finding the denominator to develop crash rate calculations but also in performing statistical tests to determine the significance of the variables in question. In order to overcome this problem, researchers traditionally used estimates like miles driven, number of licensed drivers, registered vehicles, and so forth in the denominator. Past work has utilized annual mileage or miles driven as exposure metrics for estimating crashes but these approaches are debatable. Moreover, usage of annual mileage is questionable in analyses where specific groups of drivers or environmental conditions are of interest. For example, if the relationship between light condition and driver age group is to be examined, the amount of travel done by various driver groups under different light conditions is needed to estimate exposure accurately. If one assumes that disaggregation of annual mileage by age groups of drivers is possible using a national database, the disaggregation of annual mileage by light condition is almost impossible. Even though some researchers have used data from Nationwide Personal Transportation Study in obtaining such exposure estimates, they may not be very accurate since the light condition has been estimated using the time of the trip instead of the actual light condition. This article will discuss an alternative approach that has been used over the past 40 years with good results in addressing this issue.