|
Abstract:
|
The 2020 Post-Enumeration Survey used dual-system estimation to estimate the net coverage of the 2020 Census in the United States and Puerto Rico. Part of the dual system methodology involves selecting a sample of census enumerations and assessing whether the enumerations are correct or erroneous. A source of erroneous enumerations in the decennial census is duplicate enumerations, in which two or more census records correspond to the same unique person. When one record of a duplicate pair is selected into the Post-Enumeration Survey sample, we expect that we are equally as likely to have sampled the correct record of the duplicate pair as we are to have selected the erroneous record. In this paper, we use data from the 2020 Post-Enumeration Survey to show that the match codes assigned by clerical matchers are biased towards calling the in-sample record the correct enumeration, which results in an upward bias of the dual-system estimates. We explore the bias caused by this misclassification for the 2020 Post-Enumerations Survey and discuss the implications for future iterations of the survey.
|
