Using 2010 Census Data to Evaluate Imputation Methods to Improve the American Community Survey Estimates of the Group Quarters Population for Small Geographies

• Activity Number: 25
• Type: Topic Contributed
• Date/Time: Sunday, July 29, 2012 : 2:00 PM to 3:50 PM
• Sponsor: Section on Survey Research Methods
• Author(s): Michael A Beaghen*+ and Mark E Asiala
• Companies: U.S. Census Bureau and U.S. Census Bureau
• Keywords: sample design ; small area estimation

Abstract

The Census Bureau has developed a new imputation-based methodology to improve the American Community Survey (ACS) estimates of the group quarters (GQ) population for small areas. The motivation for this work was that there are small geographies which either do not have GQ sample or have GQ sample that is not representative of the area, which could lead to distorted estimates of characteristics and/or total population. The new method imputes whole person records to GQ facilities which appear on the sampling frame but were not selected into sample. Previous evaluations have established the method's feasibility and allowed for its refinement. This evaluation aimed to establish that the new methodology improved the usability of estimates for census tracts. We applied the new methodology to the 2006-2010 ACS 5-year data and compared the imputation-based results with design-based results, using the 2010 Census as a benchmark. We found the imputation-based methods had a better distribution of GQ population by major type of GQ across tracts and a better distribution of demographic totals across tracts, as measured by the mean squared error.