The Report in Context
In August 2024, WIRED published an investigation that pulled back the curtain on a U.S. Census Bureau analysis titled “Noncitizen Voting in the 2020 Election: A Beginning Analysis.” The study claimed to have identified roughly 24,000 noncitizens who voted in the 2020 presidential election—a figure that quickly became a talking point for former President Donald Trump and his allies.
What makes the story newsworthy is not the political spin but the methodological chaos uncovered by data scientists, former Census officials, and independent scholars. The report was approved through an email vote, omitted key statistical metrics, and relied on a patchwork of administrative datasets that are notoriously difficult to match without generating false positives. As a result, leading experts have labeled the findings “statistical nonsense.”
Understanding why this matters requires a deep dive into the data pipelines, the political incentives that drove the analysis, and the broader implications for government‑generated statistics.
Technical Breakdown of the Flaws
1. Incomplete Approval Process
- Email‑only vote: The disclosure review board approved the report on August 7 via a three‑email thread, reviewing only three summary tables. Full methodological appendices were never circulated among reviewers.
- Lack of author attribution: The final PDF omitted the names of the four primary authors, a departure from standard Census practice that raises transparency concerns.
2. Problematic Data Linkage
The analysis stitched together more than a dozen massive administrative sources, including:
🔹 --------
• Year(s): --------
• Primary Use: -------------
🔹 Data Clear 2020 voting file
• Year(s): 2020
• Primary Use: Voter registration
🔹 USCIS immigration records
• Year(s): 2025 extract, Feb 2026
• Primary Use: Citizenship status
🔹 DACA approvals
• Year(s): 2025
• Primary Use: Residency verification
🔹 State Department visas
• Year(s): 2017‑2023
• Primary Use: Travel history
🔹 ICE student/removal records
• Year(s): —
• Primary Use: Enforcement actions
🔹 IRS Form 1040
• Year(s): 2020
• Primary Use: Income reporting
🔹 Social Security Numident
• Year(s): 2019‑2020
• Primary Use: Identity verification
🔹 Passport issuances
• Year(s): 1978‑2025
• Primary Use: Citizenship proof
The core linking mechanism used Protected Identification Keys (PIKs)—hashed identifiers designed to protect privacy while enabling record linkage. While PIKs are useful in controlled environments, they become unreliable when applied to commercial datasets that lack rigorous quality controls. Small variations in name spelling, address formatting, or missing middle initials can produce false matches, inflating the count of “noncitizens” dramatically.
3. Absence of Statistical Confidence
The report repeatedly described its findings as “high confidence” yet never presented confidence intervals, p‑values, or error margins. In a dataset of 128 million voter records, a 0.02 % false‑match rate would still generate thousands of spurious noncitizen matches—exactly the magnitude reported.
4. Incomplete Coverage
Only 96 million of the 128 million voter records were examined; the remaining 32 million were left unanalyzed without justification. This selective sampling introduces bias, especially when the omitted records are disproportionately from states with higher immigrant populations.
5. Dedup
lication Issues
The authors briefly mention that duplicate voter records were “cleaned” using a proprietary algorithm, but they never disclose the criteria for flagging a duplicate or how they handled cases where a single individual appeared in multiple data sources with slightly different identifiers. Without a transparent deduplication protocol, the risk of over‑counting—especially when merging immigration and voting files—remains high. In practice, a single non‑citizen could be counted multiple times if, for example, they appeared in both an ICE removal file and a DACA approval list with minor name variations.
Political Fallout
Trump’s Amplification
Shortly after the report’s release, former President Donald Trump cited the 24,000‑figure in a series of rallies, framing it as evidence of “massive fraud” that allegedly cost him the election. The claim was amplified on his social‑media platform, X, where posts linking to the Census PDF garnered millions of impressions.
Government Responses
- Commerce Secretary Howard Lutnick praised the analysis on X, describing it as “a vital step toward safeguarding election integrity.”
- Utah Lieutenant Governor Deidre Henderson commissioned an independent review of 2.1 million voter records in her state, finding only 27 confirmed non‑citizens—none of whom were linked to a vote. Her office publicly released the methodology, highlighting the stark contrast with the Census findings.
Congressional Scrutiny
Both the House Committee on Oversight and Reform and the Senate Committee on Homeland Security have requested testimony from the report’s authors and the Census Bureau’s Office of Data Quality. Lawmakers are probing whether the expedited approval process violated the Federal Statistical System’s standards for peer review and public comment.
Expert Reactions
🔹 --------
• Affiliation: -------------
• Key Comment: -------------
🔹 *John Abowd*
• Affiliation: Former Chief Scientist, Census Bureau
• Key Comment: “The methodology is riddled with avoidable errors; the headline numbers are statistically meaningless.”
🔹 *Danah Boyd*
• Affiliation: Cornell University
• Key Comment: “These ultimately appear to be people coming from a Republican operation that’s pretending to be 501(c)(3)…”
🔹 *Amy O’Hara*
• Affiliation: Former head of Census administrative data linkage unit
• Key Comment: “Using PIKs across commercial and government datasets without rigorous validation is a recipe for false positives.”
🔹 *Michael J. Barone*
• Affiliation: Senior Fellow, Brookings Institution
• Key Comment: “Even if a handful of non‑citizens did vote, the scale is negligible compared to the total electorate; the report inflates a non‑issue into a political talking point.”
Implications for Future Census Work
1.
Read the full breakdown originally published at https://ltdeveloperblogs.github.io/posts/a-us-census-report-on-noncitizen-voting-used-bad-data-to-reach-faulty-conclusions/
Top comments (0)