Voter Fraud — The Documented Record

When a politician says "there are 1,620 proven cases of voter fraud," that number sounds significant. Whether it is significant depends entirely on the denominator — the total number of votes against which those cases are measured. Every fraud rate in this article is calculated per relevant ballots cast, not as a raw count. The raw count without the denominator is meaningless.

The United States has conducted federal elections continuously since 1788. Between 1982 and 2025 — the primary span of the Heritage Foundation's database — an estimated 2.5 billion ballots were cast in federal elections (presidential + congressional combined). Presidential elections alone from 2000 through 2024 account for over 938 million ballots.

The Heritage Foundation's Election Fraud Database (electionfraud.heritage.org) is the most comprehensive conservative-curated collection of documented fraud cases in the United States. As of December 12, 2025, it contains 1,620 proven instances of election fraud — criminal convictions, civil penalties, judicial findings, and overturned elections — spanning primarily 1982 through 2025.

Heritage explicitly states on its own about page: "This database is not an exhaustive or comprehensive list of all election fraud in the states. It does not capture all cases and certainly does not capture reported instances or allegations of fraud, some of which may be meritorious, some not, that are not investigated or prosecuted." Heritage describes this as a "sampling" intended to illustrate vulnerabilities, not measure total incidence.

The denominator is not a guess. Federal election turnout is documented by the Election Assistance Commission and the U.S. Census Bureau. Presidential and midterm elections combined from 2000 through 2024 alone account for over 1.5 billion ballots. Extending back to 1982 (the Heritage database start) reaches approximately 2.5 billion.

This table includes only federal elections. When state and local elections are added, the total ballots cast in the Heritage database window likely exceed 4 billion. The 2.5 billion figure is conservative.

The political claim is not just that fraud exists — it is that fraud has altered or could alter election outcomes. To evaluate that claim, it's necessary to calculate what outcome-determinative fraud would require in a presidential election.

In 2020, Joe Biden won the four closest swing states — Georgia, Arizona, Wisconsin, and Pennsylvania — by the following margins:

This math does not prove no fraud occurred. It establishes the scale at which fraud would need to operate to change the outcome — and compares it to the entirety of what has ever been documented and proven in the United States.

Different forms of election fraud are not equally common, are not equally detectable, and are not addressed by the same policy responses. Voter ID laws target in-person impersonation — the rarest type. Mail ballot restrictions address absentee fraud — somewhat more common, still vanishingly rare. Understanding which type you're measuring is necessary before any policy argument can be made.

In-person voter impersonation — where someone shows up at a polling place and votes in another person's name — is the form of fraud most commonly cited as justification for voter ID laws. It is also the form with the lowest documented incidence by a significant margin.

The Brennan Center for Justice's comprehensive analysis of the period 2000–2014, covering over one billion ballots cast, identified 31 credible cases of in-person impersonation fraud. The Brennan Center calculated an incident rate of 0.0003% to 0.0025% at most. A George W. Bush administration DOJ unit specifically tasked with finding election fraud examined the 2002 and 2004 elections and was unable to identify a single case of in-person voter impersonation fraud over a five-year period. The unit proved fraud in just 0.00000013% of ballots — a figure Heritage itself does not dispute.

Mail ballot fraud — forgeries, illegal collection, or submission of ballots without the voter's knowledge — represents the largest category in the Heritage database. A Brookings Institution analysis of Heritage's data found approximately 0.00006% of mail ballots cast over 20 years were fraudulent. This makes it roughly 20 times more common than in-person impersonation, but still statistically negligible at the level where it could affect statewide or national outcomes.

The highest-profile real case of absentee fraud in recent history is the 2018 North Carolina 9th Congressional District scheme — documented in full in Part 3.

Claims of mass noncitizen voting — that millions of undocumented immigrants or foreign nationals are voting in federal elections — are among the most frequently made and least evidentially supported in election fraud discourse.

Fraudulent voter registrations — submitting forms with fake names or ineligible individuals — are frequently conflated with fraudulent votes. They are not the same. A fraudulent registration does not produce a fraudulent vote unless someone shows up and votes using that registration, which requires passing through an additional layer of scrutiny at the polls.

The ACORN controversy of 2008 is the canonical example. ACORN voter registration workers submitted an estimated 400,000 potentially fraudulent or duplicate registration forms. The subsequent federal and state investigations produced zero prosecutions for fraudulent votes cast from those registrations. The fraud was in the forms — not in the ballot box. Heritage's own database confirms this distinction: its "False Registration" category lists hundreds of cases, most of which did not result in a single fraudulent vote being cast.

The Department of Justice's Public Integrity Section (PIN), established in 1976 with an Election Crimes Branch added in 1980, handles federal election offense prosecutions. PIN does not publish consolidated annual voter fraud statistics. The available independent data:

Kris Kobach ran for Kansas Attorney General in 2022 specifically on a platform of prosecuting voter fraud. He won. He was granted unprecedented state authority to investigate and prosecute election crimes — authority that Kansas AGs had not previously held. His office actively pursued fraud cases.

Result: 6 prosecutions. 1 conviction. That conviction was later overturned.

This is the most useful real-world data point in the voter fraud debate: a politician who genuinely believed fraud was widespread, who had the legal authority and political mandate to find it, who actively looked — and found effectively nothing that held up in court. The Kobach test does not prove fraud is absent; it measures what aggressive investigation actually produces.

Following the 2020 presidential election, supporters of Donald Trump filed approximately 60–65 lawsuits across 12 battleground states alleging a range of misconduct — from illegal ballot counting procedures to manipulated voting machines. A critical distinction often obscured in political discussion: many cases were dismissed on procedural grounds (lack of standing, mootness, timeliness), but a significant number were also reviewed on the evidentiary merits — and rejected.

Of approximately 60+ lawsuits filed, 47 were dismissed by federal judges — at least 8 of whom were Trump appointees. Zero succeeded in overturning or invalidating a single electoral vote.

On January 28, 2026, FBI agents executed a search warrant on the Fulton County Election Hub in Union City, Georgia, seizing approximately 700 boxes of physical materials from the 2020 election cycle — including original ballots, tabulator tapes, electronic ballot images, and state voter registration rolls.

The affidavit used to obtain the warrant was authored by Kurt Olsen — an attorney who had been a legal advisor to Trump on strategies to contest the 2020 election results and who was sanctioned in federal court for statements made while challenging the 2022 Arizona gubernatorial election loss of Kris Lake. The affidavit's claims had been examined and rejected in previous investigations and lawsuits. Fulton County officials challenged the warrant in court, describing the affidavit as relying on "unsubstantiated hypotheticals" rather than verified evidence. The raid did not produce new evidence of fraud.

The investigation was led by Thomas Albus, a St. Louis-based U.S. Attorney given a special nationwide election fraud appointment by AG Pam Bondi under 28 U.S.C. § 515. Director of National Intelligence Tulsi Gabbard was physically present during the raid, conducting a parallel intelligence review of voting machines — prompting formal inquiries from U.S. Senators Mark Warner and Alex Padilla.

The Justice Department, responding to White House pressure to pursue noncitizen voter fraud cases, has faced a consistent challenge: the cases are not materializing at the scale claimed. In Nevada, law enforcement officials closed their investigation into potential noncitizen voting in the 2020 election in early 2026, having identified fewer than 40 potential instances — and the statute of limitations for any prosecution from that election had expired.

AG Bondi appointed Thomas Albus as a special prosecutor with nationwide jurisdiction for election integrity cases. The DOJ also established a "Weaponization Working Group" and directed resources toward investigating local election administrators based in part on allegations generated in online forums, rather than traditional law enforcement referrals.

In contrast to the contested fraud narratives, Congress did pass one significant piece of bipartisan election security legislation: the Electoral Count Reform Act (ECRA) of 2022. The ECRA addressed real vulnerabilities in the certification process that the 2020 post-election period exposed — specifically the ambiguity about the Vice President's role in certification and the threshold for congressional objections to state electoral slates. It clarified that the VP's role is strictly ceremonial and raised the bar for objections. This is what documented electoral vulnerability remediation looks like when it is based on actual documented problems rather than unproven fraud claims.

The absence of documented fraud does not prove the absolute absence of all fraud. Undocumented cases may exist. Specifically:

Detection gaps are real. Fraud that is never investigated never enters the record. Jurisdictions with less rigorous auditing may have more undetected fraud. This is acknowledged by Heritage, Brennan Center, and virtually every researcher in this field.

The small-sample problem. When 40 years of national data produces 1,620 documented cases, the confidence intervals on any single type are wide. A few additional undetected cases in any category would shift rates slightly but would not change the qualitative conclusion.

What researchers say. The academic consensus — including from Heritage-adjacent researchers — is that undetected fraud almost certainly exists at some level. The question is whether it exists at the scale required to alter outcomes. Studies consistently conclude that large-scale undetected fraud would produce detectable statistical anomalies in audits, recounts, and turnout patterns. No such anomalies have been found in the elections most scrutinized.

Multiple states have conducted post-2020 audits and investigations specifically designed to find widespread fraud. The results are consistent with the Heritage database rates:

When politicians with the authority, funding, and political motivation to find widespread fraud actively look for it, the result is consistent with the baseline rate: individual cases exist, coordinated schemes at scale do not.