The effectiveness of restraint systems in preventing fatalities or reducing injury has been estimated by extrapolation of data from several sources: (1) Sled tests with dummies (2) Analysis of accident case studies (3) Statistical comparison of belted and unbelted persons in crashed cars. (4) Before and after studies (e.g., with respect to belt-usage legislation, or as with the 1974 starter-interlock program)Fatality reduction estimated by the case study method is on the order of 30 percent, but by the statistical comparison method at 50 percent or sometimes as high as 60 percent. Other differences (e.g., driving habits) between belted and unbelted persons explain the disagreement between the two estimates. More complete analysis of available accident data suggests that the higher values were obtained without correction for such factors as crash severity or occupant age. When such adjustments are introduced into the statistical analysis of accident data, the result is closer to that shown by the case study method.