This paper compares results from NHTSA's New Car Assessment Program (NCAP) 35 mph barrier crash tests with actual on-highway fatality rates from the Fatal Accident Reporting System (FARS). The NCAP tests have been popularly portrayed as providing guidance regarding car safety to purchasers of new cars. Overall fatality rates are compared to various predictor measures developed from the NCAP test results. In addition, in an attempt to improve the quality of the fit, the fatality data are constrained by such items as belt usage, direction of impact, and car size. In none of the comparisons is there any substantial relationship between the NCAP predictor variables and the actual fatal accident results. Possible reasons for this lack of relationship include crash test variability, inaptness of crash test response as a fatality-related measure, and confounding variables such as driver and highway factors which influence the fatality rate.