Triple-System Modeling of Census, Post-Enumeration Survey, and Administrative-List Data

Abstract

Dual-system measurement of census coverage using a post-enumeration survey has been criticized for correlation bias, resulting when responses to the census and survey are not independent. A third system provides additional information to assess that independence. This study focuses on urban Black male adults, using data from the 1988 Dress Rehearsal Census and its Post-Enumeration Survey and from other government sources. Results using a variety of models confirm that their population is underestimated by dual-system methods. Problems involving classification and matching errors are also discussed. The results suggest that triple-system modeling has great potential for more precise estimation of the hard-to-count population and its census coverage.