222 – Administrative Record Research for the 2020 Census
Explaining Differences Between Administrative Records Modeling Simulations and 2010 Census Results
Andrew Keller
U.S. Census Bureau
In the 2010 decennial census, about one third of the US population was enumerated by a personal visit during the nonresponse followup (NRFU) operation. A major driver of cost in the 2010 Census was the NRFU operation with a total cost of about $1.6 billion. Consequently, the Census Bureau is researching the possible use of administrative records (AR) to provide an occupancy status and count for some NRFU addresses. Since some of the households included in the NRFU operation were present in administrative records (AR), the Census Bureau is planning to enumerate some nonresponding households with AR. However, occasionally census and AR outcomes differ. This paper seeks to understand inconsistencies between census and AR data. This will help inform when AR should be used to enumerate the household as opposed to enumerating the household with an in-person followup interview. Specifically, the analysis centers on housing units where AR data indicated a vacant unit and the 2010 Census showed an occupied unit or vice versa. Finally, in the scenario that AR and the 2010 Census both indicated an occupied unit, household rosters are compared for overlap of persons within households.