'I Hate to See the Evening Sun Go down': Recounts and Padding in the 19th Century U.S. Decennial Census of Population

Ronald Goeken, University of Minnesota
Diana Magnuson, University of Minnesota

The plan is to present results from our final version of linking person records from the first and second enumerations of St. Louis in the 1880 decennial U. S. Census. Given that the two enumerations (taken in June and November) mostly contained the same individuals, we have high linkage rates and low false positive rates (in part due to a linkage strategy that utilized household, street address, and district information). The comparison of linked records indicates that the information transcribed from the 19th century censuses contains many flaws (e.g., over fifteen percent of our linked records have an absolute age difference of two or more years). We also find relatively high disagreement rates for other key variables (names and birthplace). We also discuss disagreement rates for variables recently added to the 19th century complete-count databases (school attendance, literacy, and health information). In order to address the representativeness of the 1880 St. Louis inked data, we also discuss our ongoing efforts to link the 1870 and 1880 complete-count decennial databases. We focus specifically on linkages between St. Louis 1870 and both enumerations in 1880 in order to investigate whether the first enumeration in 1880 was in fact a poorly taken census. St. Louis in 1870 is also of interest because of allegations of census fraud related to that year’s decennial enumeration (i.e., the addition of thousands of fictitious names to the returns). We expect the linked data (along with an analysis of anomalous distributions in the 1870 data) will provide evidence that the St. Louis enumeration was flawed. We will tie our quantitative findings to a narrative discussing the various 19th century recounts and the late 19th and early 20th century episodes of census fraud (some of which resulted in convictions and jail sentences).

No extended abstract or paper available

 Presented in Session 203. Linking 19th Century Census Records