United States · click a state
Hover a region for details
Case-count shading:
0
1
2–3
4–6
7+
Every documented, court-verified voter fraud case tied to a US federal election cycle from 2000 to 2024, mapped to its state and county, tagged with the party of the perpetrator and the number of ballots actually affected. Plus a side-by-side Claims vs Reality panel comparing the loudest fraud claims against the documented record.
A widely-repeated political narrative claims that voter fraud is widespread enough to swing US federal elections. This map shows every documented, court-verified case the team could find that occurred during a US federal-election cycle from 2000 through 2024, drawing primarily on US Department of Justice prosecutions, state Attorney General announcements, and contemporaneous reporting from major news outlets.
The data shows a clear uptick from the 2016 cycle onward, peaking at 20 cases during the 2020 cycle. This almost certainly reflects more aggressive detection and prosecution rather than more underlying fraud:
The crucial point: even at the 2020 peak, with 20 documented cases against 158,429,631 ballots cast, the case rate is 0.0000126%, or 1 case per 7.9 million ballots. That is the highest detection rate in the entire 24-year window — and it still does not approach the scale required to alter a federal-election outcome.
The party tally is not symmetric, but it is not one-sided either. Republican perpetrators dominate the post-2016 dataset, primarily because individual Trump voters who attempted to vote twice (in their own name plus a deceased spouse or in two different states) drove a wave of single-ballot prosecutions in 2016, 2020, and 2022. Democratic perpetrators include former US Congressman Michael "Ozzie" Myers of Philadelphia, Bridgeport city councilman Michael DeFilippo, and Maryland congressional candidate Wendy Rosen. Election officials who committed fraud — nine separate cases — are tagged as a distinct bucket since their personal party is irrelevant to the crime. Use the party filter above to inspect the distribution yourself.
Why the Republican tally dominates — without bias in the dataset. The 29-vs-9 Republican-to-Democrat split is honest, not partisan, and it has a specific explanation. Most of the Republican cases on this map are individual voters — not operatives — who tried to cast a second ballot in 2016, 2020, or 2022 (their own ballot plus a deceased relative's, or their ballot in two states). These cases surged in three cycles when high-profile claims of a "rigged" election were most loudly broadcast; voters who believed the system was being stolen from their preferred candidate were disproportionately the ones who tried to "balance" what they thought was already happening — and they got caught. The Democratic cases on the map are smaller in number but include far more institutional defendants: former US Congressman Michael "Ozzie" Myers (Philadelphia), Bridgeport city councilman Michael DeFilippo, Maryland congressional candidate Wendy Rosen, and Edinburg, Texas mayor Richard Molina. The party tally counts heads; it does not measure scheme severity or institutional reach. If you want to weight by ballots affected or by election-overturning impact, the picture changes considerably.
The most widely-repeated specific voter-fraud claims of the last decade, set side-by-side with the documented record. Every "Reality" entry is sourced to a primary or counter-partisan source: a court ruling, an audit report, a Republican-led legislative investigation, the Brennan Center, the Bipartisan Policy Center, the state Secretary of State, or peer-reviewed academic work.
A case is included only if all four conditions are met:
For each defendant we sought a public record of party affiliation: voter registration (where reported in news coverage), the party of the campaign or office the fraud was committed in support of, or explicit reporting on which candidate the defendant voted for fraudulently. Where the public record does not establish party we tag the case "Unknown" rather than guess. Election officials acting in their official role are tagged "Election-official" because their personal party is incidental to the crime — they violated their oath of office regardless of party.
This is the strongest possible counterargument to the map, and we want to address it head-on rather than hope nobody asks. The objection runs: "Of course documented fraud is rare — you're only counting the cases that got caught. Most fraud is, by definition, undetected."
The reply is the detection-incentive argument, and it carries a lot of weight in this specific topic. Over the last decade, the political and financial reward for proving widespread federal-election fraud has been enormous. If meaningful fraud existed at scale, the people most motivated to find it have had every tool and every dollar required to do so, and the absence of results from those investigations is itself evidence. Specifically:
Put more plainly: if widespread federal-election fraud existed, the people most motivated to prove it — with subpoena power, millions of dollars, friendly courts, and a decade of opportunity — would have found it by now. They have not. The combination of zero-result well-funded investigations and the modern statistical audit toolkit (which mathematically bounds the probability of an undetected wrong outcome) places a hard ceiling on how much fraud could be going undetected. The ceiling is small enough that even if every undetected case were three or four times the documented count, the rate would still be a tiny fraction of a percent of the vote.
For most cases this is 1 (one person voted twice, or cast one improper ballot). For larger conspiracies, this is the number widely reported by prosecutors or election boards as part of that operation. The Bladen County NC-9 absentee-ballot scheme is the only case with a large figure (~250 ballots reported by the North Carolina Board of Elections across the operation), and the four sub-operatives are tagged 1 each to avoid double-counting. Transparency check: if the Bladen operation is excluded entirely from the dataset, the remaining 59 cases account for 455 fraudulent ballots, or roughly 1 in 3.3 million ballots cast — still a vanishingly small share of the vote. Removing the single largest case does not change the conclusion.
The Heritage Foundation maintains a widely cited Election Fraud Database. Heritage is a conservative think tank with explicit policy-advocacy commitments (most prominently as the lead organizer of Project 2025), and citing it as a primary source on a fact-checking page would be reasonable to challenge on independence grounds. This map deliberately does not use Heritage as a primary source for any case.
The Heritage database was used only as a research-discovery aid. Every case in this final dataset is sourced to an underlying independent primary source: a Justice Department press release, a state Attorney General announcement, a court filing, or contemporaneous reporting from a major news organization. Any candidate case that could not be independently re-sourced outside Heritage was excluded.
This map argues from court records, prosecutions, audits, and primary sources. None of that guarantees it will change a mind that is already settled. A few honest observations on that, since the topic is hard:
A hard truth is hard to swallow. When the facts cut against a story we believe deeply, the easier reflex is to attack the facts than to update the story. That is human, not partisan. It happens on every side of every contested topic.
Belief can be manufactured. If enough money, attention, and repetition are aimed at a claim — any claim — a meaningful share of people will come to hold it as obvious common sense, regardless of whether the underlying record supports it. This is not a comment about voter fraud specifically; it is a comment about how the information environment of the last decade works in general, and about why a page like this one is needed at all.
Everyone has a right to their own opinion. No one has a right to their own facts. Every opinion is valid as an opinion; not every opinion turns out to be true. The point of this page is not to tell anyone what to think. It is to put the documented record — sourced, dated, and citation-linked — in front of any reader who wants to look at it before they decide.
— Patrick Neil Bradley
All 60 cases, all fields, all source links — for independent audit.
Where each of the 60 cases in this map gets its primary source:
| Source domain | Cases | Type |
|---|
Each individual case in the map links to its primary source. Lightning strike odds from the US National Weather Service.