New Estimates of Census Coverage in the United States, 1850–1930

Abstract

Despite growing reliance on census data for historical research in the United States, there has been little systematic evaluation of census quality. This article relies on back-projection methods, new estimates of nineteenth-century mortality, and the 1850–1940 Integrated Public Use Microdata Series (IPUMS) samples to estimate age- and sex-specific net census underenumeration of the native-born white population in the United States in the 1850–1930 censuses. National and section of birth estimates are constructed. In general, the results suggest slightly higher net undercounts for native-born white males relative to native-born white females, slightly higher net undercounts in the South, and a modest trend toward greater census coverage over time. A few censuses stand out as anomalous. The 1870 census suffered a higher net rate of omission than any other census. The net undercount was especially high in the South, probably reflecting the unsettled conditions in the aftermath of the American Civil War. The net undercount was not nearly as great as nineteenth-century observers speculated and subsequent historians have long believed, however. The 1880 census appears to have achieved the most complete coverage of the native-born white population before 1940.