To estimate fatality risks in crashes involving cars, Fatality Analysis Reporting System (FARS) and General Estimate System (GES) files for the years 1991-1999 were combined. For the analyses, crashes were selected by type for a large part of the U.S. Mathematical models expressing the drive fatality risks in cars without air bags as functions of driver age and sex, car weight, and speed limit were developed for single car crashes and collisions between cars. These models were applied to crashes involving cars with air bags, and the difference between the modeled risk and the actual deaths used to estimate air bag effect. It was found that air bags reduced the driver fatality risk by 33% in single-car, non-rollover crashes, and by 42% in collisions between two cars. Reductions were also found in side impacts. These estimates depend critically on the assumption that presence of an air bag does not increase reporting of non-fatal crashes involving cars with them. [electronic version only]
Abstract