Crash reconstruction and crash modification factors

This paper addresses the following question: Under what conditions can reconstructed road crashes be used to estimate the effect of a safety-related countermeasure? Results developed by Pearl and his associates are used to draw two main conclusions. First, when one can (1) identify a structural equation describing a type of crash, (2) identify an additional structural equation describing the countermeasure's impact, and (3) estimate the initiating conditions for a set of reconstructed crashes, then a lower bound for a crash modification factor can be estimated by simulating whether or not each of the reconstructed crashes would still have occurred had the countermeasure been present. If the countermeasure's effect is monotonic this bound becomes tight. Second, in situations where it is not possible to reliably identify the structural equations needed for simulation, but where one can (1) identify a set of crash inputs which, when given, make the crash outcome conditionally independent of the countermeasure, and (2) predict how the distribution of these inputs will change in response to the countermeasure, then nonparametric estimation of the countermeasure's crash modification factor is possible. When it is not possible to predict the countermeasure's effect on the conditioning variables it may still be possible to identify constraints or specifications which the countermeasure should satisfy in order to realize a target crash modification.