Research Update: New Data Show 2020 Census Undercounted Young Children in Every State and Most Counties

New data published by the U.S. Census Bureau indicate that young children ages 0-4 were undercounted in the 2020 Census in every state and in more than four out of five counties examined. The Census Bureau previously found that young children were undercounted by 5.4% nationwide — a larger undercount than for any other age group and a worse undercount than in any decennial Census going back to 1970.

After each decennial Census, the Census Bureau produces Demographic Analysis (DA) estimates that use administrative data like birth, death, and migration records to produce independent national estimates of the population in order to estimate “net coverage error,” which is a measure of whether there was an undercount or an overcount. If the DA estimates are higher than the Census counts, that means that the Census missed certain people and undercounted; if the DA estimates are lower, there was an overcount. Typically, the DA estimates are only produced at the national level, but the Bureau has now released an experimental set of these data for young children at the state and county levels.

At the state level, the undercount of young children ranged from -0.02% in Vermont to -15.85% in the District of Columbia.  (However, since the DA data use birth records, babies born in DC but living and counted by the Census in Maryland or Virginia could inflate the undercount estimate.) Fourteen states had an undercount larger than the nation overall, two-thirds of which were in the South.

Estimates were also produced for counties with a population of at least 1,000 children. More than four out of five of these counties (84%) had an undercount of young children, and one-third had an undercount larger than the national undercount. Counties along the West Coast, in the Southwest, and in the South had some of the largest undercounts. Explore the map below for a closer look. Counties with larger shares of children in families with incomes below the Census poverty threshold, female-headed households, renter-occupied housing, and adults without a high school degree tended to have larger undercounts.

Young children may have been more likely to be undercounted in the decennial Census because they are more likely to live in complex arrangements like multigenerational households or blended family households. For example, parents and caregivers may not realize they are supposed to include young children in their response. Moreover, in 2020 in particular, places like daycare centers and schools that may have otherwise supported Census outreach were closed during the pandemic. The decennial Census is not the only dataset that undercounts children, but it is one of the most impactful. Research from the Census Bureau also indicates that children are undercounted in the American Community Survey (which CCF relies on for tracking uninsured rates and coverage trends). Because young children in particular are more likely to have Medicaid/CHIP coverage, too few children in the survey sample can contribute to underestimates of Medicaid/CHIP coverage compared to state and federal administrative enrollment data. Understanding where children may be more likely to be missed is a critical step in improving the count in 2030.

Aubrianna Osorio is a Research Manager at the Georgetown University McCourt School of Public Policy’s Center for Children and Families.