A number of analysts have attempted to determine the effectiveness of passenger vehicle safety belts over the past 25 years. One of the more widely used is the double pair comparison (DPC) method. This note looks at the larger picture of motor vehicle crash data to show the limitations of DPC and to present a more general method. The author organizes crash data according to whether occupants were restrained or not, whether their crashes were potentially survivable only with belt use, and whether the occupants were actually killed. Based on this model, the author develops a virtually exact means of calculating restraint use and effectiveness in preventing fatalities. Calculations using data from the Fatal Accident Reporting System (FARS) show that safety belts may be more effective in preventing fatalities than previously thought. However, part of the reason for the higher calculated effectiveness is that belt use appears to be over reported in FARS. Finally, the author discusses the origin of uncertainties and errors in the results, and shows how the data can be adjusted to give more reasonable results.