|
Abstract:
|
Preparing for the 2020 Census, the Census Bureau is researching the use of administrative records information for enumeration to reduce the number of field visits during the Nonresponse Followup operation. One of the concerns stemming from this research is the possible undercoverage of children in administrative data sources. In recent mid-decade tests, predictive models have been used to identify units with administrative records information of sufficient quality for enumeration. In this paper, we extend the predictive modeling approach by incorporating data from a child-to-parent linking dataset. This unique dataset associates children with their mother and father using data from the Social Security Administration. We explore methods of using these associations to include children on the administrative records roster for households where the children do not otherwise appear in the administrative records sources. We evaluate the impact of these alternative methods by assessing quality metrics using the 2010 Census data.
|
