The Administrative Records Experiment in 2000 (AREX 2000) was an attempt to simulate an administrative records census, using data compiled and combined from seven federal databases. These files were combined with several field operations to simulate an administrative records census. Several evaluations are planned.
This evaluation is designed to evaluate the demographic properties of households from the AR files by comparing matched addresses. We assess three basic questions: To what extent do administrative records household compositions appear similar to their Census 2000 counterparts? Looking specifically at households in the Nonrespsonse Followup universe, to what extent can administrative records data substitute for Census 2000 NRFU operations? What kinds of modeling tools are appropriate for using administrative records in an imputation framework for Census 2000 households? Results for questions #1 and #2 are primarily descriptive in nature. Results for question #3 are primarily inferential in nature, as we fit regression models to attempt to infer Census 2000 household properties from AR data.
|