An emergent issue in the health inequalities debate is how socioeconomic status (SES) and physical health relate over the life course. Many studies indicate that the SES-health relationship diminishes in later life. The present research tests the hypothesis that this convergence in health inequalities is an artifact of mortality selection, which biases downwards the "true" association between SES and health in later life. By including respondents who had subsequently died or were loss-to-followup into the analysis, I assess the sensitivity of the age-specific association between education and health to sample selection processes. I study U.S. adults followed for approximately ten years using the NHANES I Epidemiologic Followup Study. Results based on the surviving sample are robust to the inclusion of people selected out of the sample due to mortality or attrition. Sample selection biases do not appear to explain the convergence in health inequalities in late life.