Age-adjusted death rates for pancreas cancer during 1950--69 were correlated by sex and race with demographic and industrial data for the 3,056 counties of the contiguous United States. Only a small fraction of the county-to-county variation in mortality was explained by these variables, in contrast to their strong correlation with other common neoplasms. The only geographic cluster occurred in an area encompassing parts of Louisiana and Mississippi. Throughout the country, however, the rates for pancreas cancer were higher in urban areas, especially in males, and in counties with many residents of Scandinavian and East European (particularly Russian) descent. No associations were found with socioeconomic, industrial, or alcohol-consumption indices. The mortality patterns for pancreas and lung cancers were highly correlated in males, suggesting the influence of tobacco consumption on both tumors. In females, pancreas cancer was significantly correlated with diabetes mellitus, consistent with other evidence linking these two diseases.