Using log-linear modeling techniques, the probability of fault among collision-involved drivers in Hawaii is related to three categorical variables: age, sex, and vehicle type. Very young and very old drivers face up to three times the risk of being at fault compared with middle-aged drivers. Substantial gender effects also occur at both ends of the age distribution. Pickup truck drivers have higher odds of being at fault than automobile drivers, while van drivers have lower odds. An "odds multiplier" computation combines the effects of the variables to permit comparison of the relative odds of fault between various categories of drivers. The implications of these findings are reviewed in terms of policy responses toward these high-risk combinations. (A)
Abstract