Unduplication of Persons and Housing Units in the 2004 Census Test

The capture of respondent names in Census 2000 enabled for the first time a real-time assessment of duplication within the census. After preliminary analysis suggested the housing unit count on the production files was too high and duplication was the likely cause, an ad hoc operation was mounted to research and eliminate certain categories of duplicate housing units from final Census 2000 counts. The approximately 1.4 million housing units deleted as a result of the unduplication operation made this one of the largest coverage improvement activities in the census. Subsequent evaluation revealed the need for a comprehensive research effort to improve the unduplication operation for the 2010 census. This paper discusses the research undertaken in the 2004 census test, as well as plans for future testing, in order to have an integrated unduplication process in place by 2010. This process will consist of automation and followup fieldwork on cases of potential duplication, as identified by respondent data collected in the census