Get Full Government Meeting Transcripts, Videos, & Alerts Forever!
Get email alerts on the Census Methodology topic
No spam. Unsubscribe anytime.
Expert: Census evaluation methods explain why young children were most likely missed in 2020
Summary
At a Dec. 8 Massachusetts Senate Committee hearing, demographer Joseph Salvo told senators nationwide methods used to assess the 2020 census (demographic analysis and the post‑enumeration survey) show young children had the largest net undercount and identified causes including COVID timing, proxy responses and data‑quality issues with administrative records.
Joseph Salvo, a senior adviser at the National Conference on Citizenship, told the Massachusetts Senate Committee on the Census on Dec. 8 that two independent methods the Census Bureau uses to judge count completeness — demographic analysis (DA) and the post‑enumeration survey (PES) — point to systematic problems in 2020 that produced uneven undercounts and overcounts.
Salvo said DA, which sums historical birth and death records with estimates of international migration and Medicare‑enrolled populations, provides a national “litmus test” of whether the census undercounted or overcounted the U.S. population. "Demographic analysis is literally an independent way of coming up with the national population," he said. He explained DA is limited to national estimates for most groups because migration and long historical records make substate measures unreliable.
Why that matters: when Salvo compared the DA estimate to the 2020 enumeration, he said the nation’s largest net undercount by age occurred among 0–4 year olds. "0 to 4 year olds have the highest undercount in the decennial census of any age group," he said, noting the PES and DA results place that undercount at the highest rate of any age cohort.
Salvo described PES as an independent, sample‑based re‑enumeration of roughly 10,000 blocks and about 161,000 housing units that the Census Bureau uses to measure misses. He told the committee that the PES was hampered in 2020 by timing and the COVID pandemic: "there's recall bias" and greater distance between the enumeration and the PES follow‑up produced more unmatched or unresolved cases.
He also highlighted several operational drivers of poor quality in 2020: expanded use of administrative records where records are unevenly available; a rise in proxy interviews that produce weaker information about age, race and ethnicity; whole‑household imputations when no reliable data exists; and age heaping, where reported ages cluster at 0 or 5 year stops. "Proxy responses are considered to be of poor quality," he said, adding that administrative records often do not include the characteristic detail needed for undercounted groups.
Salvo pointed to geographic and programmatic levers that jurisdictions control: participation in the Local Update of Census Addresses (LUCA) program and careful local review of group quarters (nursing homes, dorms, congregate facilities) where enumeration problems were concentrated. He said LUCA is “kind of an insurance policy” because a better address list reduces the chance that housing units are missed entirely.
Salvo noted racial and ethnic patterns in 2020: he said PES showed significant undercounts for Black populations, high undercounts on reservations for American Indian communities, and a substantial increase in the Hispanic undercount from 2010 to 2020. He cautioned that some uncertainty remains in the magnitude of migration and other inputs to DA, but that the broad patterns are robust enough to guide outreach planning for 2030.
What’s next: Salvo urged states and localities to prepare for LUCA participation and to help the Census Bureau obtain higher‑quality administrative records and household contacts ahead of 2030 to reduce reliance on proxies and imputations.
