Fact check: What were the key factors in California's 2020 census undercount?

1. What advocates and researchers were claiming in 2020 — “A looming undercount where it hurts most”

Researchers and advocacy groups published warnings in mid‑2020 that people of color, immigrants, and low‑income households faced the highest undercount risk, owing to both historic undercount patterns and new pandemic obstacles. A UCLA forecast and other analyses highlighted that COVID‑19 disruptions, combined with shortened field operations, threatened enumeration among hard‑to‑count communities, potentially costing representation and federal funding ^{[1] [4]}. These 2020 warnings emphasized that reduced field work and health risks to outreach staff would depress response rates in areas with concentrated non‑U.S. citizens and racial minorities, forecasts grounded in demographic patterns and early pandemic conditions ^{[2] [1]}.

2. The “Manhattan effect” and the puzzling wealthy nonresponse story

An early September 2020 analysis identified a counterintuitive factor: wealthier Californians in dense, expensive urban neighborhoods were apparently less likely to respond, a pattern dubbed the “Manhattan effect,” which raised concerns that nonresponse at the top end could cost California a congressional seat and federal dollars ^[5]. This claim reframed undercount risks beyond traditional hard‑to‑count groups, suggesting geography and household turnover among affluent areas also mattered. The source was contemporaneous with pandemic uncertainty, and its emphasis on elite nonresponse points to a broader, cross‑cutting set of drivers rather than a single demographic story ^[5].

3. Digital divides and access barriers that shaped response patterns

Analysts flagged lack of internet access and digital literacy as central impediments to an accurate 2020 count in California, where the Census Bureau moved to an online-first model. The Public Policy Institute of California and related research identified areas with poor connectivity and high immigrant populations as particularly vulnerable to nonresponse, coupling structural access problems with language and trust issues to paint a multi‑factor risk landscape ^[2]. These access barriers were amplified by pandemic restrictions that limited in‑person assistance programs, making remote response infrastructure both pivotal and uneven across communities ^[2].

4. Pandemic effects and operational cutoffs — a compound threat to completeness

Multiple 2020 sources tied the heightened undercount risk directly to COVID‑19’s disruption of outreach and enumeration and to policy choices that shortened field operations. The timing of these limitations mattered: early warnings stressed that reduced in‑person canvassing and early termination of some data collection activities would disproportionately affect low‑response areas, compounding chronic undercount drivers and potentially translating into long‑term resource and representation losses ^{[1] [4]}. These arguments framed the undercount as the product of intersecting public‑health, logistical, and administrative shocks.

5. Post‑census assessments that challenged the worst forecasts

By late 2022, at least one policy brief assessed California’s 2020 census as “highly accurate,” crediting state investments in outreach and targeted efforts for supporting counts among hard‑to‑reach populations ^[3]. This later perspective introduces a competing narrative: that aggressive state and local mobilization, despite pandemic constraints, mitigated predicted shortfalls. The contrast between mid‑2020 risk forecasts and subsequent accuracy claims underscores the importance of timing and measurement: early projections used indicators and simulation models, while later evaluations relied on post‑enumeration analysis and administrative data comparisons ^{[3] [2]}.

6. Conflicting data, possible agendas, and how to weigh sources

The early‑2020 warnings and later accuracy claims reflect different methods, incentives, and publication contexts, so readers should treat all sources as partial. Advocacy groups and academic forecasts sought to mobilize resources and attention before the count, which can emphasize worst‑case risks ^{[1] [4]}. Conversely, retrospective policy briefs that highlight accuracy may reflect state investments and a desire to frame outcomes positively, especially where funding and political reputations are implicated ^[3]. The temporal gap between predictions and evaluations is critical to reconcile apparent contradictions.

7. Bottom line: an undercount risk that was real, uneven, and partially mitigated

The evidence across 2020 warnings and later assessments indicates the 2020 census in California faced legitimate and multi‑dimensional undercount risks—pandemic disruption, shortened operations, digital divides, and surprising wealthy nonresponse—targeting different populations in different ways ^{[5] [2] [1]}. Subsequent state and local efforts likely reduced errors, producing assessments of high accuracy, but those later conclusions do not erase the real vulnerabilities identified during the count, nor the unevenness of impacts across communities ^{[3] [4]}. The policy implication remains: targeted, sustained outreach and robust operational contingency planning are essential to secure accurate counts in crises.