The best data on the quality of 2020 Census count of young children will not be available until the fall of 2022. That is when the Census Bureau is likely to release 2020 Census counts for the population ages 0 to 4 that can be compared to the Census Bureau’s Demographic Analysis estimates to calculate net undercounts or overcounts of young children.[1] However, self-response rates are important predictors of census accuracy and data exists now to look at how young children are distributed across census tracts based on the tract self-response rates. Specifically, low-self response rates are associated with higher net undercount and omissions rates in the census (O’Hare 2020). Thus, groups over-represented in low self-response tracts are likely to be undercounted in the Census.
Young children are the focus of this analysis because young children have had high net undercounts in several recent U.S. Decennial Censuses. In the 2010 Census, there was a net undercount of 4.6 percent for young children compared to a very small net overcount for the total population. Figure 1 shows there has been a relatively high net undercount of young children in each census since 1950. Moreover, the gap between the net undercount rates for young children and for adults (ages 18 and older) as well as the gap between all children (ages 0 to 17) and young children (ages 0 to 4) have been growing since 1980.
This report focuses on the distribution of young children living in two kinds of census tracts which are likely to be problematic. First, young children living in low self-response census tracts are examined. Second, young children living in tracts where the self-response rate decreased by 10 percentage points or more between 2010 and 2020 are analyzed. The data offer empirical evidence about what to expect when the data on young children are released for the 2020 Census.
Tract-level response rates used for the study are the final tract-level self-response rates issued by the Census Bureau in January 2021 (U.S. Census Bureau 2021). Data on race, Hispanic Origin status, and poverty status are taken from the Census Bureau’s 2015-2019 American Community Survey (ACS) tract-level estimates and merged with the file showing self-response rates. For the ACS data at the census tract level, the data combining by race and age are only available in the “race-alone ” configuration so that is what is used here. Rather than repeat “race alone” with every mention of a race group readers should assume figures for races are for “race alone.” There were 1,040 tracts in Update/Leave areas or with household population of 100 or less which were not included in the analysis.
[1] Data from the Post Enumeration Survey is likely to be available sooner, but correlation bias in the PES makes net undercount estimates for young children inaccurate.
