Estimating Erroneous Enumerations in the U.S. Decennial Census Using Four Lists

• Activity Number: 327
• Type: Topic Contributed
• Date/Time: Wednesday, August 11, 2004 : 10:30 AM to 12:20 PM
• Sponsor: Section on Survey Research Methods
• Author(s): G. Gordon Brown*+ and Paul P. Biemer and Dean H. Judson
• Companies: RTI International and RTI International and UNC-CH and U.S. Census Bureau
• Keywords: census undercount ; latent class analysis ; capture-recapture ; erroneous enumerations ; heterogeneity

Abstract

We present a method for adjusting the U.S. decennial census day residency counts using a system of four lists: Census, PES, MER, ARL. Our primary objective is to develop a method that can detect erroneous enumerations that are present in the data since failure to account for erroneous enumerations in the lists will result in overestimation of the census day residents. Indicators of residency on the lists are used to obtain estimates of erroneous enumerations in a manner similar to one used in capture-recapture experiments to detect heterogeneity. The erroneous enumerations are assumed to come from a "nonresident" population that is rostered on the four lists with different probabilities than the actual census day "resident" population. Using the four lists, latent class analysis, and the software package LEM, we were able to fit several reasonable models that reflect likely census populations. We discuss the assumptions required for the inference to be valid and indicate the robustness of these models to the violation of these assumptions. The results from a simulation study are used to show when our method is reliable and situations in which it fails.