Rollover frequency in single vehicle crashes is much higher for pickup trucks and utility vehicles (60-80 percent) than it is for cars (30-50 percent). Vehicle parameters affecting stability (and thus rollover) were examined to determine their contribution to the difference in rollover frequency among passenger cars, pickup trucks and utility vehicles. Logistic regression techniques were used to develop parameter estimates for the risk of rollover in single vehicle fatal crashes. Fatal Accident Reporting System (FARS) data for 1981-1 987 were used together with engineering data for 11 models of pickup, 16 models of utility vehicle and 11 models of passenger car. Separate parameter estimates were derived for the three vehicle types to predict risk of rollover in rural and urban areas. Among vehicle design parameters, track width to center of gravity height was the strongest predictor of vehicle rollover for pickup trucks and utility vehicles; for utility vehicles, wheelbase to track ratio was also significant. For passenger cars, wheelbase was the best predictor of rollover risk. A combined model was run to establish whether a single model could be used to predict rollover risk for all three vehicle types. Wheelbase, track width to center of gravity height and driver age were all significant predictors of rollover risk, but were not sufficient to explain all the variation in rollover risk between light trucks and passenger cars. The models showed that reducing the track width to center of gravity height ratio of trucks to that for passenger cars could reduce rollover frequency by as much as 60 percent.