How to read US crime numbers

Every number starts at a police department. Most departments send monthly counts to the FBI, which uses one set of definitions for the whole country. That is the only figure that lets you compare one city with another, and it is what our rankings use.

Some cities also publish every incident on their own open-data portal. Those counts are bigger and fresher, and they say where each incident happened, which is what makes neighborhood pages possible. But each department records things its own way, so they are not comparable across cities.

Chicago shows the gap. Over the 12 months to Sep 2026, the Chicago Police Department's own data counts 2,494 violent crimes per 100,000 residents; what it reported to the FBI works out to 392. The city counts every assault and battery, the FBI only aggravated ones, and the department sends the FBI about half of what it records. Both numbers are on the Chicago page, labelled.

Violent crime is homicide, rape, robbery and aggravated assault in the FBI's figures; city portals usually add simple assault. Property crime is burglary, theft and vehicle theft. On this site theft is split four ways where the data allows:

• Robbery: taking something from a person by force or threat, carjacking included. It is violent crime.
• Burglary: entering a home or business to steal.
• Theft from vehicles: items or parts taken from a car, break-ins included.
• Theft: everything else taken without force or entry, from shoplifting to a bike off a rack.

Fraud, identity theft, receiving stolen property and theft of services are not counted as theft, even when a department files them that way. More detail in how we count.

Police figures only include what someone reported. The federal government measures the rest with the National Crime Victimization Survey, which asks a large sample of households what happened to them whether or not they called the police. In 2024 the share reported was:

Source: Bureau of Justice Statistics, Criminal Victimization, 2024, table 4 (September 2025).

So a stolen car almost always shows up in the data and a stolen bike usually doesn't. When a city's theft count rises, it can mean more theft, or easier reporting (an online form, an insurance requirement). A change in what gets reported can look like a change in crime.

A rate divides the incidents of a year by the population and multiplies by 100,000, so a city of 8 million and one of 200,000 can sit side by side. In a small place, though, a handful of incidents moves the rate a lot. That is why we use 12 months, label any rate built from fewer, and give neighborhoods under 2,500 residents counts only. When the numbers are small, read the count.

Rates use the people who live in a place. In a downtown, most people present during the day live somewhere else: they commute in, shop, go out. Every incident counts, but the denominator only counts residents, so the rate comes out far higher than the risk for any one person there.

Chicago's Loop is the clearest case. It has 42,550 residents and about 486,232 jobs (Census, 2023), about 11 for every resident, plus shoppers and visitors. Its violent crime rate of 4,183 per 100,000 is among the highest in the city for that reason. The Loop page shows the count next to the rate.

We mark every neighborhood where Census job counts are at least 1.5 times the residents and above 20,000, using the LEHD employment data. Airports, business districts and big campuses show up the same way.

• Compare it with its own city first. Neighborhood figures come from the city's data, so they line up with the city's own total, not with the FBI's.
• Look at the counts and at the 12-month trend, not one month. A few incidents more or less is normal.
• Check what kind of crime it is. A high property rate driven by shoplifting near a mall says little about a quiet street a mile away.
• Use the map. Incidents cluster on a few blocks, often commercial ones, and the rest of the area can look very different.
• Remember what is missing: unreported crime, and the people who were there without living there.

No single number tells you whether a place is right for you. That is why this site gives figures with their date and source, and no scores or grades.

• Late reports. Departments keep adding incidents for weeks. We only show a month once it has passed the source's reporting lag and holds what a normal month holds. The current month never appears.
• New reporting systems. When a department moves to the FBI's NIBRS standard, it can stop reporting for months or count differently. Los Angeles took most of 2024 to switch, so its own series on this site starts in 2025.
• Figures that can't be complete. Some agencies send the FBI only part of their cases. Savannah, GA, for example: in the 12 months to Jul 2026, 23 homicides but only 131 violent crimes in all, so homicides would be 18% of violent crime, when across the US they are about 1 to 3%. We give those places a page without a rate instead of a misleading one. The full list is on the data status page.