Without much fanfare, the U.S. Census Bureau released what it terms a “data product” bearing the anodyne headline “Noncitizen Voting in the 2020 Election, A Beginning Analysis”. It’s thick as a Chunky Bar with statistical jargon, but the key takeaway is the claim that following a preliminary review of the voter records related to the nearly 160 million votes cast nearly six years ago, “over 24,000 ... represented noncitizens at the time of the election”. And counting, because there are still 32 million voter records to go, but many are skeptical and there’s some clean-up work for the bureau to do before its claims can be accepted at face value.
A Quick Recap of the 2020 Election
Not since 1876 — when then-President Ulysses S. Grant followed Washington’s example and opted not to seek a third term, setting the stage for a hotly contested (and still-debated) matchup between the GOP’s Rutherford B. Hayes and Democrat Samuel Tilden — has there been an election in this Republic like 2020.
That year, the nation was locked-down under (hotly contested and still-debated) Covid-19 pandemic restrictions , and (hotly contested and still-debated) special election rules were implemented in purported response to the coronavirus threat.
As one medical and electoral analysis in the U.S. National Library of Medicine, “The COVID-19 pandemic and the 2020 US presidential election”, explains, the pandemic shifted “the political narrative and Trump’s re-election prospects”.
It continues:
Prior to the pandemic, the US economy was performing well, and Trump, while extremely polarizing, enjoyed strong support among Republican voters. The virus changed the narrative and shaped the political campaign. As of the end of November 2020, the USA had suffered the largest numbers of cases (over 8 million) and deaths (over 220,000) in the world. Trump’s pandemic response, which contrasted with those of many leaders in other Western democracies, was repeatedly criticized by epidemiologists and scientists.
Keep in mind this analysis was published on January 15, 2021, five days before the winner of the 2020 presidential contest, Democrat Joe Biden, was sworn into office, and many of the “epidemiologists and scientists” whose opinions were valued and trusted at the time have experienced a reassessment since.
In any event, Biden received 306 electoral votes to Trump’s 232, which sounds like a wide gap until you consider that (outside of Maine and Nebraska) electoral votes are allocated on a winner-take-all basis and in some states the final vote tally between the two was as tight as a tick.
In Wisconsin, for example (10 electoral votes), fewer than 21,000 votes made the difference in Biden’s favor in a state where more than 3.24 million were cast. In Arizona (11 electoral votes), the margin for the winner was less than 10,500 votes in a state where roughly 3.334 million voters went to the polls. And in Georgia, with 16 electoral votes that Biden received, fewer than 11,800 votes separated the two candidates out of more than 4.935 million cast in the Peach Tree State.
The then-unsuccessful incumbent has never accepted the results, but had he won in 2020 he would not have been eligible to run successfully in 2024 under the 22nd Amendment to the U.S. Constitution.
“Noncitizens”, U.S. Citizens, and U.S. Nationals
Note that, for reasons never explained, the data product uses the term “noncitizens” in lieu of the legal term “alien”, which is defined in section 101(a)(3) of the Immigration and Nationality Act (INA) as “any person not a citizen or national of the United States”.
The “citizen of the United States” part is self-explanatory, but the term “national of the United States” is in turn defined in section 101(a)(22) of the INA as a person who is: “(A) a citizen of the United States, or (B) a person who, though not a citizen of the United States, owes permanent allegiance to the United States”.
In other words, all U.S. citizens are nationals but not all U.S. nationals are citizens, though for most purposes the only U.S. nationals who aren’t citizens are (pursuant to sections 101(a)(29) and 308 of the INA), those born in the U.S. territories of American Samoa and Swains Island to parents who are not themselves U.S. citizens.
As the U.S. Department of the Interior (which administers American Samoa and Swains Island through its Office of Insular Affairs) explains, “neither citizens nor nationals of U.S. territories vote in Federal elections”, but then they aren’t required to pay federal taxes, either, so I guess there are both benefits and detriments that apply to such status.
It’s not clear whether the Census Bureau was using the term “noncitizens” in its data product to mean “aliens” or “aliens plus U.S. nationals who are not citizens”, but it appears from reviewing that document in toto that it intended the latter — but this is just one of many instances in which use of the euphemism is less than helpful.
The Census Bureau “Data Product”
Which brings me to the latest Census Bureau “data product”, which bears the “approval ID” of CBDRB-FY26-0285, suggesting the bureau’s imprimatur despite the fact that Trump-skeptical outlets such as NPR have contended: “Work on the analysis was not conducted by career civil servants at the Census Bureau, and the team behind it included individuals affiliated with the America First Policy Institute.”
That aside, here is the key passage: “The Census Bureau’s beginning analysis readily determined with high confidence that over 128 million of these voter records represented citizens, and that over 24,000 of these voter records represented noncitizens at the time of the election.”
Those “noncitizen voters” were not evenly distributed across the states, as the Census Bureau reports that 4,300 of them were identified as having voted in California, 2,500 in Texas, 1,800 in Florida, and 1,700 in New York, but fewer than 50 were identified in West Virginia, Vermont, and the Dakotas (respectively) and fewer than 15 cast ballots in Wyoming.
The bureau compiled these figures by first “drawing on” the “DataClear 2020 commercial voter file and a set of federal administrative records, all resolved to a single anonymous person identifier, the Census Bureau’s Protected Identification Key (PIK)”.
“DataClear” — or more precisely “Data-Clear” — is a private entity describing itself as “a premier Women-Owned and SBA 8(a) certified firm, providing unique quality reputable differentiators to address complex data challenges across the country”.
As per NPR, the use of the Data-Clear data “was a surprise to Carolyn Carlson, the firm's president”, though the outlet added that the company “provided voter registration records for every state, plus Washington, D.C., to the bureau beginning in the Biden administration as part of a 2024 contract that ended in September 2025”.
Again, in any event, the Census Bureau analysis explains the Data-Clear information was then “checked against an independent official benchmark: the U.S. Election Assistance Commission’s Election Administration and Voting Survey (EAVS) for 2020, the federal government’s compilation of turnout counts certified by state election offices”.
That check revealed that the Data-Clear statistics captured “146,700,000 of the 159,934,200 ballots the states reported to the EAVS across the fifty states and the District of Columbia, or 91.7 percent” of the total, from a high of 97.1 percent of the votes cast in New Jersey to a low of 74.5 percent of the ones in Utah.
At that point, the voter data was compared against various databases of “noncitizens”, including from USCIS, the State Department, ICE, and the Social Security Administration to create what it referred to as the “candidate pool” of voters who appeared in “any federal noncitizen source” and who were thus “treated as [noncitizens] at the time of a vote unless citizenship evidence establishes citizenship at or before that vote”.
The analysis explains:
This is paired with a monotonicity rule: citizenship moves in only one direction, from noncitizen to citizen through naturalization. A naturalization oath dated after a vote therefore confirms the person was a noncitizen when the vote occurred, and only undated citizenship or citizenship dated before a vote can exclude it.
Thus, if an alien received a green card in 2005 but naturalized in 2015, that person would appear “in an immigration record as a noncitizen” and enter the candidate pool, but thereafter be screened out of the pool based on the fact that the person became a U.S. citizen in advance of the 2020 election.
Preliminary, and More Explanation Is Required
The results are preliminary, as the document explains that “32 million voter records” — about 20 percent of the total cast in the 2020 election — “remain to be analyzed”.
That raises the question of why exactly the Census Bureau decided to release its preliminary results now. Plainly, there is an ongoing debate as to whether aliens ever vote in federal elections and if so, to what extent, but plainly any vote cast by an unauthorized voter has the effect of cancelling out one properly cast by an authorized citizen and therefore the issue is a salient one.
On August 18, the president took to Truth Social with his own take on the analysis:
In response, the Washington Post pointed out that, “Even taking Trump’s figure of 24,000 illegal votes at face value — a purported illegal voting share of less than two-hundredths of a percent — it would not establish that he won the 2020 election. Joe Biden won by more than 7 million votes.”
While that may be true (again, the analysis is preliminary), with due respect to the Post it misses a much bigger point: The more assurance voters have that their elections are on the up-and-up, the more likely they are to cast a ballot.
As the bipartisan States United Democracy Center reported in July 2025:
People who expressed higher levels of confidence that their vote would be counted as intended were more likely to vote. This pattern was consistent across partisan groups and most prominent among independents.
If Americans felt more confident about the security of the 2024 election, turnout could have increased by 3.0-3.7 percentage points.
If all Americans felt the highest levels of confidence going into November 2024, as many as 4.7-5.7 million more voters may have cast a ballot.
Back to the main point: more work must be done before this latest analysis will be widely accepted.
If the Census Bureau is going to issue “data products” like this one, it should be clearer in explaining how it arrived at its conclusions — and not use legally meaningless words like “noncitizen”. Because if 24,000-plus aliens did vote in the 2020 election, it’s a problem, even if those votes didn’t change the ultimate outcome.