STUDY OBJECTIVE: To describe variation in levels and causes of excess mortality and temporal mortality change among young and middle aged adults in a regionally diverse set of poor local populations in the USA. DESIGN: Using standard demographic techniques, death certificate and census data were analysed to make sex specific population level estimates of 1980 and 1990 death rates for residents of selected areas of concentrated poverty. For comparison, data for whites and blacks nationwide were analysed. SETTING: African American communities in Harlem, Central City Detroit, Chicago's south side, the Louisiana Delta, the Black Belt region of Alabama, and Eastern North Carolina. Non-Hispanic white communities in Cleveland, Detroit, Appalachian Kentucky, South Central Louisiana, Northeastern Alabama, and Western North Carolina. PARTICIPANTS: All black residents or all white residents of each specific community and in the nation, 1979-1981 and 1989-1991. MAIN RESULTS: Substantial variability exists in levels, trends, and causes of excess mortality in poor populations across localities. African American residents of urban/northern communities suffer extremely high and growing rates of excess mortality. Rural residents exhibit an important mortality advantage that widens over the decade. Homicide deaths contribute little to the rise in excess mortality, nor do AIDS deaths contribute outside of specific localities. Deaths attributable to circulatory disease are the leading cause of excess mortality in most locations. CONCLUSIONS: Important differences exist among persistently impoverished populations in the degree to which their poverty translates into excess mortality. Social epidemiological inquiry and health promotion initiatives should be attentive to local conditions. The severely disadvantageous mortality profiles experienced by urban African Americans relative to the rural poor and to national averages call for understanding.