An Empirical Evaluation of the Use of Administrative Records to Predict Census Day Residency

• Activity Number: 318
• Type: Topic Contributed
• Date/Time: Wednesday, August 6, 2003 : 8:30 AM to 10:20 AM
• Sponsor: Section on Survey Research Methods
• Author(s): Dean H. Judson*+ and Elizabeth A. Stuart
• Companies: U.S. Census Bureau and Harvard University
• Keywords: Bayesian hierarchical model ; administrative records census ; nonresponse followup ; imputation ; migration ; data quality

Abstract

Administrative records (AR) databases offer the prospect of reduced respondent burden and subsantial cost savings in census taking. However, available AR databases have not yet been shown to be sufficient for strict enumeration purposes, mainly because of the lack of synchronicity between census day and the dates that data enter into the AR files. Nonetheless AR data are promising for identifying people missed in the census. An important unsolved problem in using AR data is determining which correspond to people actually resident on Census Day. We propose a hierarchical model in which one level describes the migration process, and the other describes the probabilities of observation in each of the available records systems. The observation model uses the full information in the records, including the dates associated with the records and available covariate information, and can accommodate a variety of administrative record types. In this application, we use data obtained from the Statistical Administrative Records System, a simulated "administrative records census," to make block-level total population estimates. We illustrate by comparing estimates to Census 2000.