A Data-Driven Analysis Of Voter Fraud Claims
A data-driven analysis of voter fraud claims, their real-world impact, and the misleading narratives that fuel them.
Shri Khalpada
Voter fraud is a contentious issue in American politics. It's been used to justify restrictive voting laws, undermine confidence in elections, and incite violence. The data shows that the discourse around voter fraud has been alarmingly disconnected from reality.
Let's dive into common arguments about voter fraud and see how they hold up.
"Voter fraud changed the outcome of the 2020 election"
The 2020 election saw a record number of votes cast, with over 158 million Americans participating. Despite the high turnout, The Cybersecurity and Infrastructure Security Agency (CISA) called it "the most secure election in American history."
The visualizations below use data from The Associated Press to illustrate the scale of voter fraud in the 2020 election in key battleground states. The AP Review, described as "a process that took months and encompassed more than 300 local election office", covers multiple types of investigated voter fraud, including double voting, dead voters, and non-citizen voting.
On the left, we see the number of votes that were flagged as suspicious by the AP review. On the right, we see the margin of victory for Joe Biden in each state, or the number of votes that would have needed to be fraudulent to tip the results of the election.
To illustrate just how rare voter fraud was in the 2020 election, we can compare it to other improbable events. Across the six key swing states shown above, there were 473 flagged votes out of 25.5 million votes cast—a rate of just 0.0019%, or roughly 1 in 52,000 votes.
The AP Review puts the odds of a single vote being suspicious (which doesn't necessarily mean fraudulent) as less likely than:
- Finding a four-leaf clover (1 in 5,000)
- Being struck by lightning in your lifetime (1 in 15,300)
- Flipping a coin and getting heads 15 times in a row (1 in 32,768)
- Randomly selecting the exact minute of Lionel Messi’s first La Liga goal from an uninterrupted 29-day-long video of every minute he played over 17 years (1 in 42,168)
For another perspective, the Heritage Foundation—the famously conservative think tank who authored Project 2025—has documented just under 1,600 cases of voter fraud since 1982 out of billions of votes cast. That is a minuscule rate, which undermines their own claims of widespread voter fraud.
In other words, the idea that voter fraud happened on a scale large enough to sway an election isn’t just unlikely, it’s astronomically improbable based on the best available data.
"Why is any amount of voter fraud acceptable?"
It's not. But a scale of hundreds of votes in an election of hundreds of millions of people is minuscule. Of the votes flagged as suspicious, only a small fraction were prosecuted. And of those, both candidates were affected.
Two things are true at once: voter fraud is unacceptable, and the scale of voter fraud in the 2020 election was negligible for affecting the outcome. Claiming that the 2020 election was stolen because any voter fraud occurred is a baseless argument. It's like saying that because there are counterfeit bills in circulation, the entire U.S. economy is fraudulent (in fact, the estimated rate of counterfeit bills is 1 in 40,000, which is similar to the rate of voter fraud calculated above).
The cure is worse than the disease, especially when used to justify restrictive voting laws that disproportionately affect marginalized communities.
"What about reports of thousands of cases of voter fraud?"
We should be skeptical of extraordinary claims, especially when they come from sources with a vested interest in the outcome. That doesn't mean that we should dismiss them outright, but we should demand strong evidence to back them up. The burden of proof is on the side making the claim.
Take Arizona’s 2020 election audit, for example. The Cyber Ninjas audit, a partisan review of the state’s election results, initially claimed to have found tens of thousands of cases of potential voter fraud. These staggering numbers were spread widely by conservative media and politicians.
Liz Harrington, Trump's spokesperson from 2021 to 2024, tweets about 275,000 potential cases of voter fraud based on the Cyber Ninjas audit. It's unclear where this figure came from, as the audit itself never made such a claim.The problem is that the numbers were essentially unusable. According to the Brennan Center for Justice, Cyber Ninjas’ methodology for identifying fraud was riddled with basic statistical errors.
Perhaps the most basic error was that Cyber Ninjas flagged voters who shared the same name and birth year as potentially fraudulent (not date and year, just year), warranting a "full review." This is a classic example of the Birthday Paradox, where our intuition about probability often fails us.
Let's see the principle in action.
The interactive below simulates how often people with the same name randomly share a birth year. In a state like Arizona, where about 500 people are named Michael Smith (according to Google), Cyber Ninjas’ flawed approach would have flagged many legitimate voters as suspicious—simply because they happened to be born in the same year.
That was just one of many issues with the audit. For a more comprehensive breakdown of the audit's flaws, I'd once again recommend the Brennan Center's analysis.
Cyber Ninjas had no prior experience in election audits, and their analysis was discredited by multiple reviews. Republican officials in Maricopa County, where the audit took place, called the audit "littered with errors & faulty conclusions". Doug Logan, the CEO of Cyber Ninjas and a Trump supporter, texts that the numbers were "screwy".
"Voter fraud is one thing — what about election fraud?"
While voter fraud refers to individuals casting illegal votes, election fraud refers to systemic issues that affect the integrity of the election itself. This can include ballot tampering, ballot stuffing, gerrymandering, and voter suppression.
In the case of 2020, there were no credible reports of widespread election fraud. Georgia, for example, conducted three separate recounts and an audit, all of which confirmed the original results. Over 60 lawsuits challenging the election results were dismissed for lack of evidence by judges appointed by both Republican and Democratic presidents. William Barr, Trump's Attorney General, stated that the Department of Justice found no evidence of widespread fraud.
While there was no evidence of widespread election fraud, there remain real threats to election integrity. Between gerrymandering, voter suppression, and the fake electors plot, in which Rudy Giuliani and others coordinated with state legislators to send fake electoral votes to Congress, the 2020 election was far from perfect. By all accounts, these threats pose a far greater risk to democracy than the minuscule rate of voter fraud.
Parting Thoughts
This is a common pattern with misinformation: alarming statistics get sensationalized, misrepresented, or outright invented, only to unravel when examined. Rhetoric, unfortunately, moves faster than fact-checking.
More than ever, this is a time to be vigilant about the information we consume and share. Extraordinary claims should require extraordinary evidence to be taken seriously. Extraordinary claims without evidence can perpetuate real damage.
We're just getting started.
Subscribe for more thoughtful, data-driven explorations.
