An Analysis of Alternate Variance Estimation Methods for the American Community Survey Group Quarters Sample

• Activity Number: 20
• Type: Topic Contributed
• Date/Time: Sunday, August 1, 2010 : 2:00 PM to 3:50 PM
• Sponsor: Section on Survey Research Methods
• Author(s): Don Keathley*+
• Companies: U.S. Census Bureau
• Keywords: American Community Survey ; group quarters ; variance

Abstract

The American Community Survey (ACS) began producing estimates of the group quarters (GQ) population, for the full implementation of the survey, in 2006. The ACS uses a successive difference replication (SDR) method to compute variances for these estimates. Part of the SDR method involves the assignment of a set of replicate factors to each sample GQ person. In 2006 and 2007, the ACS assigned these sets at the person level. This assignment method resulted in the underestimation of GQ population variances, as they did not take clustering of persons in GQs into account. An alternate assignment method more accurately reflects the clustering effect that is associated with GQ sample persons. Alternative variance estimation methods (jackknife, random group) provide benchmarks with which to compare the a preferred SDR method.