Election fraud is vanishingly rare in all U.S. states. The evidence in support of this assertion is overwhelming.
The conservative Heritage Foundation identifies 1,340 instances of election fraud in the entire United States over forty years, at multiple levels of government (from 1982 to 2021), in a country where the popular vote in the most recent Presidential election in the year 2020 was 81,268,924 v. 74,216,154 (for a total of 155,485,078 votes cast in one election of many in the election cycles its statistic covers). It documents a fraud rate of less than one in hundred million votes cast in the United States.
The liberal Brennan Foundation identifies dozens of linked independent sources that corroborate the conclusion that voting fraud is negligible in the United States. Some of the court case decisions it notes are particularly relevant:
Even the Supreme Court, in its opinion in Crawford upholding
Indiana’s voter ID law, noted that the record in the case “contains
no evidence of any [in-person voter impersonation] fraud actually
occurring in Indiana at any time in its history.”
The conservative leaning Fifth Circuit Court of Appeals, in an opinion finding that Texas’s strict photo ID law is racially discriminatory, noted that there were “only two convictions for in-person voter impersonation fraud out of 20 million votes cast in the decade” before Texas passed its law.
Similarly, it noted that:
A comprehensive 2014 study published in The Washington Post found 31
credible instances of impersonation fraud from 2000 to 2014, out of
more than 1 billion ballots cast. Even this tiny number is likely
inflated, as the study’s author counted not just prosecutions or
convictions, but any and all credible claims.
This is on the same order of magnitude as the Heritage Foundation count.
There were just four cases of documented voter fraud in the 2016 election in the entire United States.
In terms of materiality, the question is whether there was enough fraud in terms of falsely cast ballots or miscounting of ballots to change an election outcome in modern times (that was not prevented by being caught before the election results took effect). There are no examples of this happening in modern times in any U.S. elections.
An Associated Press review of the 2020 Presidential election found far fewer allegations of vote fraud (mostly claims that isolated individuals voting more than once) than needed to impact the outcome.
Claims of mass voter fraud rarely hold up to further investigation:
The review by Republican lawmakers in Michigan that found no systemic
fraud cited various claims they had investigated. For example,
senators were provided with a list of over 200 voters in Wayne County
who were believed to be dead. Of these, the report noted, only two
instances involved actual dead voters. The first was due to a clerical
error in which a son had been confused with his dead father and the
second involved a 92-year-old woman who had died four days before the
Similarly, in Pima County, Arizona, the local prosecutor received 151 complaints of voter fraud in the 2020 election, not one of which was substantiated and resulted in a criminal prosecution as the prosecutor closed the books on the last of those claims on January 14, 2021.
The kind of fraud that voter ID laws are targeted at is particularly rare.
Another 2020 academic study found that:
We estimate the change in the reported number of voter fraud cases
when states switch to conducting elections by mail. We consider two
types of states: states where a large number of voters receive their
ballots by mail (receive-by-mail states, RBM) and a subset of these
states where all registered voters are automatically sent ballots by
mail (vote-by-mail states, VBM). We then compare the number of voter
fraud cases in RBM (VBM) states to the number of cases in non-RBM
(non-VBM) states, using two approaches standard in the social
sciences. We find no evidence that voting by mail increases the risk
of voter fraud overall. Since 2016, RBM (VBM) states have reported
similar fraud rates to non-RBM (non-VBM) states. Moreover, we estimate
Washington would have reported eighty more cases of fraud had it not
introduced its VBM law in 2011. While our analysis of the data
considers only two of many possible approaches, we argue our findings
are unlikely were fraud more common when elections are held by mail.
Far more distortion of electoral outcomes is due to people accidentally not marking a vote in a race where they intended to do so due to poor ballot design, accidentally marking more than two candidates in a single candidate race, due to erroneous failure to count valid provisional ballots, and due to people mistakenly believing that they were not allowed to vote when in fact they were allowed to vote.
So, did any trend become apparent during those recounts?
Perhaps the most notable instance of election fraud in recent years was in the 2018 Congressional midterm election in North Carolina where allies of a Republican candidate, Mark Harris (who did not run in the following special election) engaged in election fraud and were caught after an initial count was made, resulting in a special election conducted after the original election was not certified (this is a state with strong voter ID requirements and anti-fraud oriented laws).
Anecdotally, there has been a trend towards voter fraud being conducted mostly by older Republican loyalists who convince themselves contrary to all evidence that voter fraud by Democrats is widespread and want to "level the playing field" for their candidates. But, these cases are so rare, it is hard to say that any of the data is really statistically significant in establishing trends one way or the other.
For example, four voters in a Florida retirement village that was a Republican stronghold (two registered as Republicans and two unaffiliated) were charged with voter fraud in late 2021 and early 2022.
Similarly, in another case from the 2020 election:
Hartle was married to Las Vegas businessman Donald Kirk Hartle, a
registered Republican. In November 2020, Hartle told Las Vegas
television station 8 News Now (KLAS-TV) that he felt "disbelief" when
he found out that a mail-in ballot was submitted in his late wife's
name. It was "pretty sickening," he said at the time, adding that he
didn't know how it could've happened. But Hartle had actually cast the
phony ballot himself.
Here are a couple of other representative cases:
In southeast Pennsylvania, 72-year-old Ralph Thurman, a registered
Republican, was sentenced to three years’ probation after
pleading guilty to one count of repeat voting.
Donald Holz is among the five people in Wisconsin who face voter fraud
charges. He said all he wanted to do was vote for Trump. But because
he was still on parole after being convicted of felony drunken
driving, the 63-year-old retiree was not eligible to do so.
Then there was this case out of Colorado arising from the 2016 election:
Colorado prosecutors can charge people with misdemeanor mail ballot
offenses and felony forgery at the same time, the Colorado Court of
Appeals ruled on Thursday.
In a voter fraud case involving a former Colorado Republican Party
chairman and conservative radio host, a three-judge panel set
precedent in ruling that prosecutors were not wrong to charge Steven
Curtis with violating laws outside of the state’s mail ballot statute
for voting twice in the same election. . . .
Curtis, 61, was sentenced to four years probation and 300 hours of
community service in 2018 for voting twice in the 2016 general
That happened when Curtis filled out, signed and mailed his former
wife’s mail-in ballot after doing his own, forging his wife’s
signature in the process. Prosecutors became aware of that when his
former wife, Kelly Curtis, who had moved to South Carolina a year
earlier, contacted the Weld County Clerk’s Office to get her mail-in
ballot, and was told that she had already voted.
DNA evidence showed that Curtis had sealed the ballot return envelope,
and handwriting samples showed he filled out the return address and
signed his ex-wife’s name, the ruling said.
Another Colorado case suggests a psychology behind a recent surge of Republican election fraud (although still tiny and immaterial relative to the number of votes cast):
Suzanne Morphew hasn't been seen in a year, and last week, her
husband, Barry Morphew, was arrested and charged with first-degree
murder. That's obviously tragic enough on its own, but what followed
were additional charges for casting an illegal vote.
According to the local reporting, last October, Barry Morphew also
allegedly submitted a ballot on behalf of the wife he allegedly
killed. Her ballot was submitted without a signature -- a requirement
in Colorado's vote-by-mail system -- but it did include his name as a
According to the arrest affidavit, asked why he submitted the ballot
for his missing-and-presumed-dead wife, Barry Morphew said, "Just
because I wanted Trump to win.... I just thought give him [Trump]
He added, according to the affidavit, that he believed "all these
other guys are cheating."
The Republican Clerk and Recorder of Mesa County, Colorado (a conservative county in Colorado where that post is the top county election official in ordinary times), Tina Peters, was suspended from her elect administration duties by a court and is currently facing a grand jury investigation for election administration misconduct (not necessarily changing any votes).
Do states with stricter ID laws significantly limit fraud?
Do states without those strict laws suffer much higher fraud?
No. For example, an analysis of the Heritage Foundation voter fraud database found that there were extremely low rates of voter fraud in the five states with universal mail in voting (less than one case per million votes cast by mail).
Do recounts not provide that information either way?
State recounts provide some information but not a whole lot. One study reviewed all
5,778 statewide general elections from 2000 to 2019, and found that
recounts occurred 31 times (0.54 percent). . . . The average shift
across all recounts was 430 votes, which accounted for 0.024 percent
of the statewide vote in those races.
(This is an average shift in 1/4667th of the statewide vote).
Three of the 31 recounts resulted in a reversal of the original election result. In each case, a Democratic candidate overcame an initial vote deficit to defeat the Republican candidate. The three recounts averaged a vote swing of 393 votes.
U.S. Senate in Minnesota, 2008 Al Franken (D) Initial count -215, final count +225, vote swing: 440. (Details here).
Vermont State Auditor, 2006 Thomas Salmon (D) initial count -102, final count +239, vote swing 349 (Details at the New York Times).
Governor of Washington, 2004 Christine Gregoire (D) Initial count -261, final count +129, vote swing 390 (Details here).
Often, a major source of shifted vote counts in a recount involves provisional ballots that were disallowed in the original count that are allowed after further examination determines that the provisional ballots were validly cast in a recount, or undisputed ballots that weren't included in a vote count by accident.
Do any other opportunities to compare ballot integrity state-to-state,
besides recounts, allow objective conclusions to be drawn about how
the volume of electoral fraud is influenced by those laws?
There is no statistically significant connection between the existence of anti-fraud laws and the volume of election fraud by voters or purported voters (or in vote counting) in particular states.
This was not always true in the United States.
Election fraud was commonplace in the U.S. in the era of "political party machines" in the late 1800s and early 1900s, ending about a century ago with reforms to the system such as the demise of patronage jobs in favor of civil service systems, with further progress made with the end of prohibition in 1933 that also reduced government corruption more generally in the United States.