Are Self-Participation Rates Predictive of Accuracy in the U.S. Census?

Abstract

Metrics related to the U.S. Census have been widely available for several decades but there has been a dearth of studies examining the relationship among key metrics in the Census. This paper provides empirical evidence about the link between self-participation rates and census accuracy using data from the 1990, 2000, and 2010 U.S Censuses. The preponderance of the evidence shows lower self-participation rates are highly correlated with higher net undercounts and omissions rates for key socio-demographic groups and states. Nine out of 11 correlations examined in this paper are statistically significant and in the predicted direction. One key reason self-participation rates are associated with census accuracy is the fact that the population not captured in the self-participation operation goes into the households for the Nonresponse Followup (NRFU) operation. Census Bureau data show data collected in NRFU is not as accurate as that collected in self-response. The larger the share of data collected for a population that is collected in NFRU, the lower the quality of data for that group. The connection between self-participation rates and census accuracy mean the differential self-participation rates seen in the 2020 Census suggest patterns of net Census undercounts seen in the past are likely to be seen in the 2020 Census.