OBJECTIVES: To determine whether the association between self-rated health (SRH) and 4-year mortality differs between black and white Americans and whether education affects this relationship. DESIGN: Prospective cohort. SETTING: Communities in the United States. PARTICIPANTS: Sixteen thousand four hundred thirty-two subjects (14,004 white, 2,428 black) enrolled in the 1998 wave of the Health and Retirement Study (HRS), a population-based study of community-dwelling U.S. adults aged 50 and older. MEASUREMENTS: Subjects were asked to self-identify their race and their overall health by answering the question, "Would you say your health is excellent, very good, good, fair, or poor?" Death was determined according to the National Death Index. RESULTS: SRH is a much stronger predictor of mortality in whites than blacks (c-statistic 0.71 vs 0.62). In whites, poor SRH resulted in a markedly higher risk of mortality than excellent SRH (odds ratio (OR)=10.4, 95% confidence interval (CI)=8.0-13.6). In blacks, poor RSH resulted in a much smaller increased risk of mortality (OR=2.9, 95% CI=1.5-5.5). SRH was a stronger predictor of death in white and black subjects with higher levels of education, but differences in education could not account for the observed race differences in the prognostic effect of SRH. CONCLUSION: This population-based study found that the relationship between SRH and mortality is stronger in white Americans and in subjects with higher levels of education. Because the association between SRH and mortality appears weakest in traditionally disadvantaged groups, SRH may not be the best measure to identify vulnerable older subjects. © 2007, The American Geriatrics Society.