Political violence

They stay in every count on this page, as on the main site, and are shown as their own category. The coders’ definition says the category may include violence that is not political, so read totals with them in mind; the filter at the top of the page can leave them out.

Sex and race describe the perpetrator or, for groups, the group (“men and women together” is the main site’s “multiple perpetrators”). The number of perpetrators is recorded as −99 when unknown, and in a few rows as 0; both read as not recorded. 1 event records a transgender perpetrator.

Online-only events are left out, and so are the few events CCC could not place in a county. The numbered rules mark changes in how CCC records events, so a step across one may partly reflect the record. For 2017 to 2020 CCC coded protests against Trump as left wing and for him as right wing. The record depends on what reached the news and social media, and the latest months may still grow as CCC adds events it finds late.

Crowd size is reported for 37% of protests, and larger, better-covered events are likelier to have one, so the page does not sum turnout. The county file has CCC’s low and high estimates.

The FBI counts an incident once, and an incident with several motivations in each of their categories. In 2025: race, ethnicity or ancestry 6,163; religion 2,549; sexual orientation 1,649; gender identity 523; disability 160; gender 124. Disability has been collected since 1997, and gender and gender identity since 2013, when the FBI also widened its religious and racial categories.

Reporting is voluntary and the FBI estimates nothing for agencies that do not take part. From 2021 it took data only through its incident-based system (NIBRS), and many agencies had not yet switched. The coverage figures come from each year’s published report (2008 on); the incident file also holds reports agencies sent later, so for some years, 2021 most of all, it counts more incidents than the report did.

These are the prosecutions DOJ chose to announce, not every threat. Most threats are never charged, many charged cases get no press release, and state prosecutions are not here at all. A case is counted in the month of DOJ’s first release about it; its stage is as of DOJ’s latest release and may be out of date, since an acquittal or dismissal DOJ never announced does not appear.

Releases come from justice.gov’s press-release archive, read every night. The archive holds the U.S. Attorneys’ offices’ releases from Jan 2013; before that it has only the Department’s national offices, so the figures start there (2 earlier cases are in the file). Every release that mentions a threat or a threat statute is read by a language model, which decides whether it announces a prosecution for threatening a public official and records the target, the defendants and the stage; the lab checked its answers by hand on a sample (Methods and sources). Releases about the same defendant, office and charges are grouped into one case.

The series covers elected and appointed officials, judges and prosecutors. Cases in which the only people threatened were police officers, sheriffs’ deputies, corrections officers, federal agents or members of the armed forces are left out of every count and of the file (290 such cases); a case that threatened an official as well as the police is counted.

339 of 812 cases name the official threatened; the others withhold the name (“a Member of Congress”), so the series rests on the office, not the person.