The mistake that destroys credibility
The single most common error in election coverage is collapsing very different things into one word: rigged. A troll farm buying Facebook ads, a court annulling a vote over paperwork, a foreign spy agency stealing emails, and an official fabricating a results protocol are not the same event. Treat them as one and you can no longer be trusted to describe any of them.
So the first thing this database does is refuse to use one bucket. Every finding is sorted into exactly one of five categories.
The five categories
-
Result falsification. The recorded numbers do not reflect the votes cast — ballot stuffing, fabricated protocols, count manipulation. This is the strongest possible charge, so we only record it when a court, an official annulment, or an authoritative forensic finding establishes it.
-
Process violations. The count may be honest, but the contest was not fair: unequal media access, repression of the opposition, illegal financing, vote buying, misuse of state resources.
-
Foreign interference. A foreign state or its proxies acting against the election: cyberattacks, hack-and-leak, covert financing, impersonating media or institutions.
-
Information operations. Coordinated inauthentic activity to manipulate what people see — troll farms, fake local-media networks, covert political ads. This one can be foreign or domestic.
-
Statistical anomalies. Patterns in the data that look unusual. These can justify a closer look, but on their own they are not evidence of fraud, and we always publish them with their limitations.
The canonical example: the United States, 2016
Hold two sentences in your head at once:
- A Russian influence campaign, social-media operations, and hack-and-leak targeting the 2016 U.S. election were documented — by the U.S. Intelligence Community, the Senate Intelligence Committee, and federal indictments.
- That vote-counting systems altered the recorded result was not established. The Department of Homeland Security assessed that the systems Russian actors targeted were not involved in tallying votes.
Both statements are true at full strength. Interference in the environment around an election is a real and serious thing. It is also a different question from whether the count was changed. A record you can trust keeps the two apart — and says so plainly when the second one is unproven.
Why the categories carry confidence tiers
Sorting a finding into the right box is only half the job. The other half is saying how strongly the evidence supports it. That is why every finding also carries a confidence tier — from confirmed (a court ruling or official finding) down to debunked (authoritatively refuted). A famous claim that turns out to be false is not deleted; it is recorded as debunked, so the refutation is part of the permanent record.
When you read a page here, look at both signals: the category tells you what kind of thing is being claimed, and the tier tells you how sure we are. Neither alone is enough.