Census report claims 500 noncitizen voters in Wisconsin. Researchers say the methodology cannot support it
See more of our coverage in your search results.
The Census Bureau report claiming 24,000 noncitizens voted in 2020 says 500 of them were in Wisconsin. Election researchers in the state say the number can't be taken seriously, and this time they have the report's methodology to point at.
The brief describes its test plainly: if a person appears in the voter database and does not have proof of citizenship attached, they are counted as a noncitizen voter. Missing documentation becomes evidence of a crime.
Philip Rocco, chair of the political science department at Marquette University, called the analysis meaningless. "There are questions about data stewardship and data linkage from the very start, before we even get into more issues like authorship and citation," Rocco said. "This is not a Census Bureau product."
The finding is smaller than the error the method produces
Matching two databases that describe people, without a shared unique identifier, generates false matches. Rocco said researchers normally expect a false positive rate between 1 and 6 percent for this kind of Census data linkage, because records go stale, voter rolls are messy, and identifying details get stripped out to protect privacy.
Now compare that to what the report claims to have found. Across 128 million voter records, it reports 24,000 apparent noncitizen voters, a rate of about 0.019 percent.
| Expected false matches | |
|---|---|
| At a 1% false positive rate | 1,280,000 |
| At a 6% false positive rate | 7,680,000 |
| What the report actually found | 24,000 |
Run this kind of match across 128 million records where nobody was a noncitizen voter, and you would still expect somewhere between 1.3 million and 7.7 million false hits. The report's entire finding is 50 to 300 times smaller than the error the method generates on its own. The signal sits far below the noise.
"This is an outlier in my experience," Rocco said. "I have never seen a Census report that so clearly does not meet the standard of the bureau. The report is indistinguishable from having zero evidence."
Two other details compound it. The bureau used voter data from DataClear, a Virginia-based data firm, and working with a third party meant stripping Social Security numbers, the one identifier that would make matching reliable. The report also doesn't say which voter rolls it used, so they may be outdated.
Rocco offered a concrete version of the problem: two people named John Smith on the same Wisconsin avenue with different birthdays. Without checking the underlying records, there's no way to know whether that's two men or one man and a typo.
Barry Burden, director of the Elections Research Center at the University of Wisconsin-Madison, made a similar point to the Associated Press, noting the report "is unsigned, is described merely as a 'beginning' analysis and relies on data from an unfamiliar private company." Burden added that "necessary methodological details such as exactly how a 'high-quality' data match was defined are missing entirely."
What Wisconsin's own checks have found
Wisconsin already looks for this. The Wisconsin Elections Commission runs audits after every election, and cases are referred to district attorneys for possible prosecution. Voter registration accuracy is checked against the Division of Motor Vehicles, the Department of Corrections, and the Electronic Registration Information Center. A valid driver's license does not make a noncitizen eligible to vote.
Those systems have turned up very little. The commission has been made aware of four alleged election fraud cases tied to citizenship since 2015, and none occurred during the 2020 election. The handful of fraud cases prosecuted in Wisconsin since 2020 were mostly for double voting.
Emilee Miklas, a commission spokesperson, declined to comment on the Census report but said there is "no evidence to indicate that a significant number of noncitizens have voted in previous elections or will do so in upcoming elections."
So Wisconsin's own auditing, with access to the actual state records, has identified four allegations across a decade. The federal report, working from third-party data with the strongest identifier removed, produces 500 for a single election.
The margins
Even taken at face value, 500 votes would not have changed any federal race in Wisconsin in 2020. Joe Biden carried the state by 20,608 votes, about 41 times the claimed figure. The closest federal race that year, Ron Kind's win over Derrick Van Orden, was decided by 10,328 votes, roughly 20 times larger.
President Trump cited the report on August 18, posting "I WON THE ELECTION!" and saying the next installment will show more.
The pattern matches what national experts said when the report first surfaced, which we covered in our earlier piece on the report: no listed author, no published methodology, and a match rate that experts say is consistent with error rather than fraud.
Sources
Quotes from Philip Rocco, Emilee Miklas, and the Wisconsin Elections Commission, along with the report's described methodology and the DataClear detail, are from the Milwaukee Journal Sentinel. Barry Burden's comments were made to the Associated Press. The false-positive comparison applies the 1 to 6 percent range Rocco described to the report's own figure of 24,000 matches out of 128 million records; the multiplication is ours. Wisconsin's 2020 election margins are the certified results.
Add CensusEasy as a preferred source on Google to see our stories first.
Which race is decreasing in the United States?
The non-Hispanic white population is the only major group that has shrunk since 2010, down 4.6 million to 192.2 million. It fell in 34 states and 74% of counties. Here is where, and why.
US trade deficit widens to $88.6 billion in July as computer imports hit a record
The July trade deficit is the largest since the March 2025 tariff rush. Capital goods imports set a record at $140.3 billion, computer imports have doubled this year, and Vietnam and Taiwan have both passed China as the largest bilateral deficits.
The US cities where the most adults hold a graduate degree
About 14% of American adults hold a graduate or professional degree. In Bethesda, Maryland it is 58.1%, with Brookline, Palo Alto, Newton, and Cambridge close behind. Every city on the list has a single institutional explanation.
What methodology did the census noncitizen report use?
According to the brief, if a person appears in the voter database and does not have proof of citizenship attached to their record, they are counted as a noncitizen voter. The bureau used voter data from DataClear, a Virginia-based data firm, which required removing Social Security numbers, the identifier that makes record matching reliable. The report does not specify which voter rolls it used.
Why do experts say the finding is smaller than the error rate?
Marquette professor Philip Rocco says this kind of Census data linkage typically produces false positives at a rate of 1 to 6 percent. Applied to 128 million voter records, that would generate between 1.3 million and 7.7 million false matches even if no noncitizens voted at all. The report found 24,000, roughly 50 to 300 times fewer, which Rocco says makes it indistinguishable from having zero evidence.
How much noncitizen voting has Wisconsin actually found?
The Wisconsin Elections Commission has been made aware of four alleged election fraud cases tied to citizenship since 2015, and none occurred during the 2020 election. The state audits results after every election and cross-checks voter registration against DMV, Department of Corrections, and Electronic Registration Information Center records. Most fraud cases prosecuted since 2020 involved double voting.
