Methodology and sources

1. Download new incidents. Each morning the pipeline asks every portal for incidents newer than the last one it saw, plus the previous month again, so late reports are picked up.
2. Sort each incident into an offense type. Departments use their own category names. We map them to seven types: homicide; assault and battery; theft, burglary and robbery; vehicle theft; property damage and arson; weapons offenses; drug offenses. An incident that fits none of them is left out rather than guessed.
3. Count by month. Incidents are counted by the month they occurred, per offense type. Individual records are not kept.
4. Turn counts into rates. Rate per 100,000 per year = incidents in the last 12 months ÷ population × 100,000, using the U.S. Census population. Where a city has fewer than 12 months of usable data the rate is annualized and labelled as such. This is what makes a city of 8 million comparable with one of 2 million, and with the FBI’s figures.
5. Neighborhoods. Where the source records the neighborhood of each incident (Chicago’s 77 community areas), the same counts are kept per area and rated against the area’s 2020 Census population.
6. Publish only complete months. A month is shown once the source’s reporting lag has passed. The current month never appears, and a source that stops updating is labelled historical instead of being shown as current.

Violent crime on this site is homicide plus assault and battery. Property crime is theft, burglary and robbery together with vehicle theft and property damage. Weapons and drug offenses are counted in the totals but in neither family.

Two honest caveats. Robbery is a violent crime in the FBI’s definition, but several source datasets label it alongside theft, so for now it sits with property crime rather than being split by guesswork. And “assault and battery” includes simple assaults, which the FBI’s violent index excludes, so our violent rates run higher than FBI figures for the same city. Comparisons between cities on this site use the same rules everywhere, so they remain fair.

• Crime that was never reported to police. Surveys consistently find that a large share of incidents, especially thefts and assaults, are not reported.
• Differences in how departments record. Two cities can classify the same event differently, and a department can change its practice from one year to the next.
• Where in the city it happened. Citywide rates average very different neighborhoods together. Neighborhood-level figures are planned where the source data supports them.
• Population change within a year. Rates use the latest Census estimate for the whole period shown.

Because of all this, treat a difference of a few percent, or a single month’s jump, as noise. The comparison against the same month a year earlier is the most reliable signal on the site.

No names, addresses, photos, arrest records or individual incidents, and no pages about specific people. Only monthly counts, which anyone can download from the data status page and reuse with attribution.