A case study in using model-assisted estimation to integrate survey and administrative data

This paper reports research to introduce model-assisted estimation into the American Community Survey (ACS), a large-scale ongoing survey intended to replace the long-form sample in the U.S. decennial censuses. The proposed application integrates information from administrative records into ACS estimation. The approach to model-assisted estimation restricts the use of the administrative records to adjustments to the survey weights, while retaining the data on characteristics reported by respondents in the ACS. Although the ACS is a general-purpose survey not specifically tied to health, this case study may suggest possible methodological applications in areas of health statistics.