In a study of the relationship between Insurance Institute for Highway Safety frontal offset crash test ratings and real-world fatality rates, there was a clear trend for better-rated vehicles to have lower driver fatality risk, although the correlation was not uniform across all vehicle groups or statistically significant in all cases. For all types of crashes combined, fatality rates per registered vehicle were generally lower for vehicles rated good than for vehicles rated poor, but rates for acceptable and marginal vehicles were not always within this range. A more precise examination of fatality risk was accomplished by comparing driver outcomes in fatal two-vehicle crashes. When a rated vehicle collided with a nonrated vehicle, the fatality risk for the rated vehicle driver was highest for poorly rated vehicles, then progressively smaller for vehicles with marginal, acceptable, or good ratings. For two-vehicle crashes of similar vehicles rated good and poor, the odds of driver fatality was 34 percent lower for the good vehicle than for the poor vehicle, but this estimate was not statistically significant. Finally, in head-on crashes of rated vehicles, the estimated odds of driver fatality was approximately 74 percent lower for the good vehicle than for the poor vehicle, with confidence limits ranging from 28 to 91 percent.