Crime & protection history, month by month.

The point is context: the crime rate most people carry in their head is decades out of date, and headlines make unrest feel constant. This is the long view, from the FBI’s own data and the standard academic protest count.

How to use it

• Toggle offenses on the left, zoom to an era with the slider or a preset, and hover any chart to read the exact rate at that month.

What you get

• Violent and property crime as a monthly rate per 100,000 people, back to 1985, plus each individual offense — so you can see both the headline and what’s driving it.

The groups

Crime (FBI Crime Data Explorer):

• Headline rates — all violent offenses combined, and all property offenses combined. Both are far below their early-1990s peak.
• Violent offenses — homicide, robbery, rape and aggravated assault, each on its own.
• Property offenses — burglary, larceny-theft, motor-vehicle theft and arson.
• Reporting — the share of the U.S. population whose agencies reported that month. Context, not crime.

Firearm background checks (FBI NICS, back to 1999):

• NICS checks — monthly background-check volume, the standard proxy for gun-buying and a clean read on public unease. It spikes at every crisis; the all-time record is March 2021 (~4.7M). A check is not a gun sale — one check can cover several firearms, and some states run permit rechecks — so read the trend, not the level.

Civil unrest (Crowd Counting Consortium, back to 2017):

• Civil unrest — protest events, arrests, and events with property damage. Property damage is the line to watch for genuine unrest rather than ordinary, peaceful protest.
• Civil unrest — harm — injuries (crowd + police) and deaths (deaths recorded only from 2021 on).
• Civil unrest — turnout — people counted, and the share of events with a crowd-size estimate.

Reading these honestly

Crime rates are coverage-adjusted — offenses divided by the population whose agencies actually reported — so a change is a change in crime, not in who is counting. Watch the Reporting line anyway: the dip to about 80% in 2021–2022 is the switch to NIBRS-only reporting, when many agencies (including some large cities) briefly stopped reporting. A fall there means fewer agencies counting, not less crime.

Two things to know about the unrest lines: the long ramp is partly the counter, not the country — CCC logged ~900 events a month in 2017 and ~3,300 in 2025, much of it the project’s own coverage widening, so compare within the steady post-2021 window. And turnout is a floor, not a count — only about a third of recent events carry a crowd-size estimate, so “People counted” drifts down for record-keeping reasons; the coverage line shows how much is actually measured.

Two definition notes: rape was broadened to a new definition in 2013, so that line steps up then by definition rather than by a real jump — compare within an era, not across it. Homicide is the most reliable line, because a body is hard to miss.

Good to know

The data lags about a year, and the most recent months fill in as agencies report. The clearest recent story is homicide: a sharp rise in 2020, and a steady fall since. This is a context chart, not a predictor.