Fault in Cyclist-Motorist Accidents

Abstract:

The determination of fault, not in the legal sense but rather of the act of committing a violation, in cyclist-motorist collisions may be considered in terms of a “civilian’s perspective” and an “officer’s perspective”. These perspectives may be represented by an ordered probit, and either two independent binary probit models or a single bivariate probit respectively. Simulation enables comparison between the three forms. The ordered probit predicts fault with at least a 62.5% accuracy 99% of the time provided fault must be assigned to either party. Without such provision, accuracy drops to 60.9% for the dataset upon which the other models are simulated. In estimation of the two binary probit models, it was found best to use only observations where a party was solely at fault. The result was a prediction with at least a 54.8% accuracy 99% of the time. For the purposes of statistical efficiency, the models should not be estimated separately. Thus a bivariate probit is used to simultaneously estimate the models. Like the binary probit, it was found best to estimate the model from only sole fault. The result was a prediction with at least a 55.6% accuracy 99% of the time. In comparison of the models the difference in perspectives accounts for 92% of the variation – suggesting the “civilian’s perspective” may better represent the decision process.

Masters Thesis:

• Fault in Cyclist-Motorist Accidents