Motor vehicle accident rates with the unit of risk expressed as mileage are useful for general safety promotional and educational purposes. Their usefulness for specific engineering and enforcement purposes is quite limited, however, as a result of the gross methods whereby mileage data are collected. Their meaningfulness also is reduced as the distance over which the rate is computed becomes smaller. In the limit, at a single point, the mileage-based statistics are completely meaningless. Another criticism of these statistics stems from the fact that the unit of risk is seldom identified with the site of the risk. The single instance where site of risk is identified with unit of risk is in evaluations of the hazard of a particular stretch of road. It is demonstrated that even in a case as specific as this, a mileage-based index reduces in fact to volume-based indexes divided by a distance constant. This constant contributes nothing to the statistical behavior of the random risk factor, and, in fact, may obscure the true hazard of the road. From these arguments it is concluded that motor vehicle accident rates should be expressed simply as volume-based indexes. These measures would relate the number of accidents occurring at a given point during some period of time to the volume of vehicles passing that point during the period. The point could be any stretch of road over which the volume of traffic is essentially constant. To increase the usefulness of the results, the points should be defined in operationally meaningful lengths.
Abstract