By Fritz Martin in Longwave — 01 Sep 2026

How Rare is Noncitizen Voting, Really?

Audit after audit, study after study, every political leaning — they all land on the same tiny number.

Tuning In

As we inch closer to the midterm elections, the efforts of the Trump administration to bring elections under federal control have suffered some losses, but notched a few wins. The SAVE Act is one of the foundational blocks in their strategy and is again the focus of this week’s Longwave.

Strong Signal

Tuning in the strong stories with weak reception.

Are there millions of noncitizens voting in U.S. elections? According to proponents of the SAVE Act the answer is yes and is the reason the bill is needed. Today we explore the origins of that claim and look into recent reporting on noncitizen voting.

This is the second installment in a three part series on the SAVE Act. Last week we covered the bill’s proof of citizenship requirements to vote.

First, a little refresher. The Safeguard American Voter Eligibility Act (also called the SAVE Act or SAVE America Act) [SB 1383], is a federal bill presented as a way to keep noncitizens from voting in U.S. elections. The bill is currently in congressional limbo but President Trump continues to push hard for its passage at every opportunity. Opponents of the bill say it makes it too difficult for citizens to vote and the problem it says it is trying to solve doesn’t actually exist. That last point is what we will explore today.

The Problem?

The detailed breakdown of this story can be found in the linked Snopes article above. I am going to give a short synopsis in this first section, but you really need to read the original to get the full scope.

One Study to Spawn Them All

The genesis of the claim about millions of noncitizens being registered to vote can be traced back to one study from 2014.

The paper in question was by Old Dominion University political scientist Jesse Richman (with David Earnest and Gulshan Chatta) published in the journal ‘Electoral Studies’ using data from the Cooperative Election Survey (CES, then called CCES), a large biennial online survey run by a Harvard-led academic consortium.

Contained in that survey were a small number of people who checked a box saying they were noncitizens: 339 people out of 32,800 respondents in 2008; 489 out of 55,400 in 2010. From that small subsample, Richman’s paper estimated 6.4% of noncitizens voted in 2008 and 2.2% in 2010 — and presented the idea that noncitizen votes could plausibly have been large enough to swing close elections.

Not So Fast My Friend

In 2015 several of the principal investigators from CES (Stephen Ansolabehere at Harvard, Brian Schaffner at Tufts, and colleagues) published a response showing most of the “noncitizen” respondents were almost certainly actual citizens who misclicked a checkbox — a known survey artifact called measurement error. They re-interviewed 19,000 prior respondents: of 121 people who’d said “noncitizen” in 2010, 36 switched their answer to “citizen” in 2012 (and 20 people flipped the other direction), showing the citizenship question itself was unreliable at this tiny scale.

In 2017, more than 200 political scientists and statisticians signed an open letter stating the Richman findings “should not be cited or used in any debate over fraudulent voting.” In 2018, Ansolabehere testified under oath in a Kansas voter ID trial that Richman’s noncitizen voting estimate was “not statistically different than zero.”

Richman himself walked the claim back in 2023 as an expert witness in an Arizona voter ID case. In that case, using actual state data, he found 1,934 Arizona voters (0.43% of the state’s voting age noncitizen population) had records suggesting noncitizen status at some point. Separately, using improved 2022 CES data with a better citizenship question and larger sample, he found less than 1% of noncitizens indicated they were registered — a small fraction of his 2014 numbers.

Richman told Snopes directly that the lower estimates “lent some credence” to his original critics, and that the earlier headline-grabbing figures need heavy caveats. Additionally, Schaffner told Snopes the problem with using the CES to extrapolate noncitizen voter registration is that is was not designed to sample noncitizen adults.

TL;DR

A 2014 academic research paper that used online surveys—which were later found to be flawed—proposed the idea that noncitizen voting could be large enough to sway close elections. The lead author of the paper has since walked back that idea, and using improved data and methods has since estimated the number of noncitizens registered to vote to be a small fraction of his original estimate. In 2017 Over 200 political scientists and statisticians signed an open letter rejecting the findings from the original 2014 academic paper that started it all.

Don’t Let Facts Get in the Way of a Viral Story

It was too late to get the genie back in the bottle. In 2017 an independent researcher named James Agresti took the same flawed data from the 2008/2010 surveys and ran his own analysis. Unlike Richman, Agresti isn’t a political scientist — he’s an engineer by training who runs Just Facts, a self-published advocacy site, not a peer-reviewed institution.

Agresti eventually claimed 10% to 27% of noncitizens are illegally registered to vote — even higher than Richman’s original numbers. Richman himself has said he disagrees with Agresti’s numbers.

The claim went viral in May 2024, after a campaign by conservative influencers which was then amplified by conservative think tank Heartland Institute and then picked up by conservative news and conspiracy theory sites. Elon Musk poured rocket fuel onto this burgeoning online fire when he posted a headline from one of the online conservative outlets that stated “up to a third of illegal immigrants are registered to vote” and asked if it was true.

Chain of Events

Flawed Academic Paper 📝 > New Claims, Same Flawed Data 📉 > Conservative Influencers 🎙️ > Conservative Think Tank 💭 > Conservative News Amplification 📡 > Elon Musk 🚀

Suddenly we went from a niche academic research paper with flawed conclusions that nobody had ever heard of to millions of people believing its flawed analysis as undeniable fact. Welcome to the Matrix.

Does it or Doesn’t it?

A small number of researchers still argue you can’t completely rule out noncitizen voting using a survey where people report their own citizenship status — since anyone actually voting illegally would have reason to lie, not just misclick. It’s a fair point. But it’s a different, much smaller claim than “millions,” and it doesn’t rescue the 2014 study’s numbers — which even its own author no longer stands behind.

Ruling out noncitizen voting was never really the point, though. There will always be dishonest people who do dishonest things, no matter the domain being discussed. If you have one robbery in a neighborhood does that mean it’s a “bad” neighborhood? If the complete absence of crime is the only proof a system works, then the world has yet to devise a working system for anything.

The real question is, does noncitizen voting happen at anything close to a meaningful number? Any other framing is a straw man.

“There is no proof that there is this overwhelming number of noncitizens on the rolls,”
—Gabriel Sterling, the chief operating officer in Republican Georgia Secretary of State Brad Raffensperger’s office / MS Now / October 23, 2024

Well, is it a Meaningful Number?

To answer that question, we need to look at multiple datapoints. Luckily, this is a well researched topic with a mountain of data backing it up.

Below are studies and audits built on real records — voter rolls, DMV data, prosecutions — not the kind of self-report survey that produced the Richman number.

National Academic/Nonpartisan Studies

Brennan Center (2017), “The Missing Millions”: Interviewed 44 election administrators in 42 jurisdictions across 12 states about the 2016 election. This review included 8 of the 10 counties with the largest noncitizen populations in the country. Across 23.5 million votes, officials referred an estimated 30 incidents of suspected noncitizen voting for investigation — 0.0001%. Zero incidents were reported in 40 of the 42 jurisdictions.
Rutgers political scientist Lorraine Minnite studied the first three years of a DOJ voter-fraud initiative that ended in 2005 and found 14 total convictions of noncitizens nationwide.
In a separate multi-year study of California and Oregon complaints, Minnite found 28 complaints in California and 5 in Oregon, resulting in 4 convictions combined. “The Myth of Voter Fraud” (Cornell University Press, 2010).
News21 (Arizona State University investigative project): reviewed all reported U.S. voter fraud 2000–2012, found 56 alleged noncitizen-voting cases nationwide over 12 years.
Cato Institute: reviewed state investigation data and concluded noncitizen voting is “virtually nonexistent” (2026) and separately rebuked the Richman methodology directly in 2020.

When States Actually Checked

Georgia, 2024⁴ — 20 confirmed noncitizens (9 had voted, mostly pre-2012). 1 in ~400,000 of 8.2M registered voters.
Nevada, 2016¹ — 3 confirmed noncitizen votes. 0.0003% of 1.1M ballots.
North Carolina, 2016³ — 41 noncitizen votes.
Ohio, 2015¹ — 44 since 2000.
Ohio, 2017¹ — 126 since 2013 (82 in 2017 alone). 0.0015% of ~8M who vote each cycle.
Ohio, 2020¹ — 13 alleged. 0.00016% of that cycle's ballots.
Iowa, 2012–2014¹ — 10 charged after $250K investigation. Started from 3,000 flagged.
New Mexico, 2011¹ — 9 referred out of 1.2M voters reviewed.
Colorado, 2012¹ — Started as an 11,805 claim, found 35.
Utah, recent² — 1 confirmed noncitizen. Never voted; 2.1M voters reviewed.
Idaho, 2024² — 36 "very likely" noncitizens out of ~1M voters.

¹ Compiled from research by the Brennan Center for Justice, which sources each figure to the original state or news report. ² Utah and Idaho figures from the Cato Institute. ³ North Carolina’s figure appears in Brennan Center’s compilation, but Brennan doesn’t link to a source for it — this link to the original state audit was located separately. ⁴ Georgia’s figure was updated independently, replacing Brennan Center’s original (2022, preliminary/flagged) figure with the state’s own more rigorous 2024 audit.

Other Surveys, post-2016

New York Times surveyed officials in 49 states + D.C.: found 2 possible instances out of 137.7 million votes cast.
Washington Post reviewed the Nexis news database nationwide: found 4 total fraud cases, zero noncitizen-specific.
The Heritage Foundation — the conservative think tank that spearheaded Project 2025 — keeps its own database of election fraud cases. Different reviewers have counted it at anywhere from 23 to 99 noncitizen-voting cases, depending on when they looked. The exact number moves, but it’s always been closer to 100 than a million. Three of the studies using the database:

But New Jersey!

There will always be criminals, cheats, and mistakes — no system is perfect. The question is scale. A recent story about a DMV software glitch in New Jersey that let 6,600 noncitizens register between 2023 and 2024 is making the rounds in discussions about noncitizen voting. The salient point from this incident, however, is the fact it was a software bug, not a scheme. Applicants checked “no, not a citizen,” and the system registered them anyway. New Jersey Governor Mikie Sherrill said fewer than 400 cast a vote. Sherrill went on to say the voters were scattered around the state and registered as Republicans, Democrats, and unaffiliated voters.

That’s 0.009% of votes cast, by people registered across party lines. Across party lines is an important thing to keep in mind since the narrative often presented by conservatives is democrats want noncitizens to vote because they all vote for democrats.

Was this incident bad? Sure. Enough to steal an election? Not even close.

Survey Says!

None of this data agrees on much, except the one thing that actually matters: nobody’s found a meaningful number. From liberal and conservative think tanks to journalists and state auditors, the results are always the same: smaller numbers than your average influencer’s follower list.

When that many people of all stripes keep landing on the same answer, it stops being a coincidence and starts being the answer. In the end—as of September 1, 2026—there is no credible research that shows noncitizen registration, much less voting, is an actual problem in the U.S.

Trump’s Stolen Election Claims

Now that we know where the assertion that ‘millions of noncitizens are voting’ originated and how often credible studies conclude it actually happens, we need to understand why it has permeated conservative media and political talking points recently. To do that, we need to go back to President Trump’s unsubstantiated claims that the 2020 presidential election was stolen from him—the same claims that led to the Capitol riot on January 6th, 2021. Despite all the evidence to the contrary, even from his own party, President Trump has continued these false claims into his second term in office and has used them as the basis for his attempt to gain federal control of elections.

It is important to view every statement from the Trump administration about elections through this lens.

Numberwang

The idea behind the Department of Homeland Security’s current attempt to find noncitizens on active voter roles works something like this: take publicly available voter information for a state, run it against federal immigration records, flag anyone who matches. But a ‘match’ isn't a confirmed noncitizen — it’s just two records an algorithm judged close enough to raise a flag.

🔍

Quick clarification: the tool DHS is actually using here is called the SAVE (Systematic Alien Verification for Entitlements) System. It’s a real-time tool that cross-references several separate federal databases and reports back an answer used to verify immigration status. The SAVE Act is the bill this piece is about. Two different things that happen to share a name.

At this point no human has verified the match. Data-matching at this scale is never perfect — researchers consider error rates of up to 2.5% normal for this kind of comparison. On a dataset the size of a state’s voter rolls, that alone would produce thousands of false flags before a single real noncitizen is counted. It’s why North Carolina’s own audit found that DMV-flagged ‘noncitizens’ turned out to be actual citizens 97.6% of the time once someone actually checked.

The SAVE System was originally built to check immigration status for benefits eligibility, not voting. Its data comes from USCIS records — essentially a snapshot of someone’s immigration status as of their last recorded entry into the system. That’s a problem. Becoming a naturalized citizen is a status change, and that update doesn’t reliably or quickly flow back into the data the SAVE System is checking against. DHS rushed an expanded version of the tool into use before it could reliably tell the difference between someone who’s still a noncitizen and someone who used to be one.

That’s also exactly why naturalized citizens get hit by this so disproportionately — it’s not a random error spread evenly across everyone, it’s a specific blind spot that targets the one group of people who, by definition, all have a status-change event sitting somewhere in their file that the system might not have caught up with yet.

More than 70 Missouri county election clerks — Republicans and Democrats — wrote a joint letter to the state’s House speaker warning: “These lists are deeply flawed: they are outdated, inaccurate, and include individuals we know to be U.S. citizens.”

There’s Already a Tool for This

If keeping voter rolls accurate is the actual goal, there’s a system built for exactly that — and it’s been running quietly for over a decade. The Electronic Registration Information Center (ERIC) is a nonprofit that 25 states and D.C. belong to. Member states share their voter and DMV data every 60 days, and ERIC cross-checks it against death records and other states’ data to flag people who’ve moved, died, or registered twice.

Three of the four states DHS has been most vocal about — Nevada, New Jersey, and Pennsylvania — are already ERIC members. Only California isn’t.

A handful of states have left ERIC in recent years, but not because it stopped working. Georgia’s own Secretary of State, Brad Raffensperger — a Republican who kept his state in ERIC — put it this way: states that left were “reacting to disinformation” and “hurt their own state and others” by walking away from “the best & only group capable of detecting double voting across state lines.”

Match Game

Below is a timeline of the most recent ‘match’ claims from the Department of Homeland Security.

On July 17th Department of Homeland Security Secretary Markwayne Mullin announced DHS had found “over 250,000 potential non-citizens” registered across CA, NV, NJ, and PA, timed the night after Trump’s primetime election-fraud address. More than a month after the initial claim DHS has not disclosed its methodology to any of the four states.

According to ABC News Mullin sent letters to the top election officials in the four states claiming that social security numbers were used to match potential noncitizens, even though Social Security numbers are not part of the publicly available voter data.

On July 27th an article in The Hill reported an internal DHS correspondence said a “preliminary review” reportedly found 118,000 matches, less than half the original 250,000.

On August 13th

“…in a private virtual conference with Nevada officials, DHS officials disclosed that they had identified only 185 “potential” noncitizens in Nevada — far from the nearly 16,000 the department initially alleged to have found through an analysis of the state’s public voter file earlier this year, according to recordings obtained by Democracy Docket.”
—from ‘Bulls**t’: Nevada election chief blasts Trump’s noncitizen voter roll claims after DHS walks them back / Democracy Docket / August 21, 2026

By DHS’s own admissions the 250,000 number is incorrect, but even if we take it at full face value, CBS News calculated that’s roughly 0.1% of registered voters nationwide. This is the unverified upper claim, not a confirmed number.

On top of the fact that the numbers change day to day, DHS has steadfastly ignored requests to release the methodology used to derive the numbers and made verification of the information shared with the states “nearly impossible”.

“But the list the Homeland Security Department shared with Nevada is incomplete — it doesn’t even include names — and is nearly impossible to verify, according to files released by the secretary of state’s office. Nevada officials have requested more information, but the department has not yet offered it.”
—from ‘Homeland Security Made a Big Claim About Noncitizen Voting. Now It Can’t Prove It.’ / The New York Times / August 20, 2026

This entire moving target mystery numbers game reminds me of a recurring sketch on the underrated gem from the early-mid 2000’s ‘That Mitchell and Webb Look’. It’s all Numberwang.

The Why

All the reputable data and research points to the same conclusion: noncitizen voting is a minuscule occurrence in U.S. federal elections and has no effect on election outcomes. The question then becomes, with such overwhelming data proving noncitizen voting isn’t a problem, why is the Trump administration pushing so hard to pass the SAVE Act?

That is the question we delve into next week in the final installment of the SAVE Act series.

Signing Off

This edition required extensive research time and is such an important topic I didn’t want it sharing the spotlight. I will get back to multiple topics in the coming weeks. Also, be sure to catch up on last week’s podcast if you missed it. As always, thanks for tuning in.

Fritz