Ellen S

Rollover risk for passenger vehicles is studied using data from six states. The metric that is used is the number of rollovers per single vehicle accident. The analysis is performed with logistic regression and relevant factors such as weather and driver age are controlled for in the model. The use of categorical analysis permits the data to be used on the observation level, preventing the loss of information that occurs when aggregated data, such as an average rollover rate, are used. The static stability factor (SSF) is the ratio of one half the track width to the height of the center of gravity. The crashes from the state data in the analysis are restricted to those involving make/models for which the SSF has been measured, but this is a large share of the fleet. SSF turns out to be an excellent predictor of rollover rate. The scores of some dynamic tests are also considered, along with SSF, and these turn out to have some association with rollover but not as much as SSF. Predicted rollover rates are compared with actual rates. Also, hypothetical rollover rates for an “average” crash are computed as a function of SSF and performance in a dynamic maneuver test. The model that uses static stability factor in the form log(SSF-0.9) gives the best predictions, especially in the low SSF range.