An attempt was made to predict whether each of 360 truck drivers had at least one accident recorded during the past three years with California's Department of Motor Vehicles. The prediction was based on 27 non-driving variables on which data could have been collected before licensing. The same prediction procedure was also used in an attempt to predict recorded violations for the same time period. Multivariate analysis was employed in the statistical analysis of the data. Factors studied included personal, cognitive, visual, motor /reaction time/, medical, and personality scores from two groups of 360 drivers each. The prediction of recorded accidents, although not high in any absolute sense, was statistically significant. Thus, a justification existed for a discussion of the variables that contributed to that prediction. In a rough sense, accidents are more likely to be associated with the younger driver with glasses who is slow in braking. The prediction of recorded violations was in the right direction, but it failed to reach a statistically significant level. Analysis seemed to indicate that this failure was due to the use of the same procedure for recorded accidents; had a procedure been used that would utilize the essentially continuous characteristic of recorded violations, the prediction of recorded violations would also have been significant. Thus, it was felt justified to discuss also the variables that contributed to that prediction. In a rough sense, violations are more likely to be associated with the younger, taller /or heavier / driver whose knowledge of driving regulations is low, whose side vision /and possibly night vision/ is better than average, and whose systolic blood pressure is lower than average.
Abstract