680 – Imputation and Analysis of Missing Survey Data
Simplifying the Noninterview Adjustment Used in Weighting the American Community Survey Housing Unit Sample
Evan B. Gutentag
U.S. Census Bureau
Edward C. Castro Jr.
U.S. Census Bureau
Mark E. Asiala
U.S. Census Bureau
The American Community Survey (ACS) weighting methodology uses three noninterview adjustment factors. The first and second factors are defined by building type, data collection month, and census tract. Two factors are used instead of one to reduce variance due to small cell sizes. Since the ACS is a multimode survey with most noninterviews occurring in the final mode, the third factor adjusts for mode bias that the first two factors may introduce since they are applied to interviews in all modes. The three factors are expected to reduce nonresponse bias without notably increasing variance, thereby reducing mean square error (MSE). The purpose of this research is to simplify the adjustment without impacting ACS estimate quality. MSEs of ACS estimates were computed and compared when formed using 1) current methodology, or 2) the first noninterview adjustment factor only, defined by building type and census tract. The simplified method used to compute estimates did not notably affect the MSEs of those estimates, suggesting that building type and small, local geography sufficiently account for the nonresponse bias of survey estimates without significantly increasing their variances.