To What Extent Is There Bias by Selection?: Selection for Road Safety Treatment in Norway

Abstract

The selection for road safety treatment for national roads in Norway is examined. The main question asked is, To what extent are the sites selected for road safety treatment biased in favor of sites that have recorded an abnormally high number of accidents during the period before treatment? To answer this question, the observed accident rates at treated sites during the before period were compared with normal accident rates for similar sites. Accident rates were adjusted for clear non-linearity in the relationship between traffic exposure and the number of accidents. Sites were classified into those that had higher-than-normal accident rates, lower-than-normal accident rates, and close-to-normal accident rates. For treated intersections, it was found that 47% had a higher-than-normal accident rate, 47% had a lower-than-normal accident rate, and 6% had an accident rate close to the normal rate. For treated road sections, it was found that 39% had a higher-than-normal accident rate, 39% had a lower-than-normal accident rate, and 22% had an accident rate close to the normal rate. Perhaps the most surprising finding of this study was that a sizable proportion of treated sites were safer than normal during the before period. Selection of sites for road safety treatment in Norway is a complex process, influenced by multiple considerations, not just the accident record. Selection for treatment can be made more effective by selecting fewer safer-than-normal sites. However, it is not possible to quantify the extent to which a more effective process of selection for treatment can be implemented.