Minority ethnic groups have disproportionately high rates of diabetes prevalence and complications. This study examined the degree, nature, and mediation of ethnic differences in changes over time in four health outcomes: functional limitations, emotional symptomatology, psychosomatic symptomatology, and self-reported health status. The theoretical context incorporated tenets from the social characteristics and the minority status perspectives. Data were drawn from two rounds (1992 and 1996) of the Health and Retirement Survey, a national plane survey of midlife Americans. The study focused on a subsample of 744 respondents who had been diagnosed with diabetes or high blood sugar by the time of the first round. Descriptive and lag-time regression analyses were employed. Five models were used to: (1) assess the independent effects of being Black and being Hispanic on changes over time in each of the outcome variables without and with the set of mediating variables; and (2) gauge the specific manner in which mediating variables affected initial statistically significant effects among Blacks and Hispanics. The mediating variables included five social position measures (education, income, net worth, gender, and a role integration index), health insurance coverage, and four health-related lifestyle measures (body mass index, drinking, smoking, and physical inactivity). The results pointed to a complex pattern of effects between Blacks and Hispanics and across outcomes. The social characteristics hypothesis was supported in two instances: increases in functional limitations over time for Blacks and poorer self-reported health over time for Hispanics. The minority status hypothesis was supported only for increases in emotional symptomatology among Hispanics. No support for either hypothesis was observed for psychosomatic symptomatology. Critical mediating variables also differed between Blacks and Hispanics across the outcomes. The two variables mediating the Black effort for functional limitations were body mass index and gender, whereas the two variables mediating the Hispanic effect for self-reported were smoking and education.