How to Check Crime Before Buying a House

Every buyer asks about crime, and almost every crime number they find is softer than it looks. Listing sites show letter grades and "safety scores" built from modeled estimates, while the underlying official record — the offenses actually reported to police — sits in a free federal database most buyers never open. Here is how to check crime the way the record actually works: what exists, what doesn't, and how to read the numbers without fooling yourself in either direction.

There is no official crime count for a single address. The FBI publishes offenses reported to law enforcement, aggregated to the county — 3,133 of 3,143 U.S. counties for 2024. Check the county rate per 1,000 residents, check which agencies reported and for how many months, and treat any site offering a precise score for one house as a modeled estimate.

What crime data actually exists (and what doesn't)

The FBI's Uniform Crime Reporting program collects offenses known to law enforcement from local agencies. It is the closest thing the country has to an official crime record, and its honest grain is the county — the FBI collects by agency, not by street address or ZIP code. For 2024, the dataset we build on covers 3,133 of 3,143 counties and just under 7 million reported offenses (6,966,667 in the county file).

That has one consequence buyers have to accept up front: no source can truthfully give you a crime count for 123 Main Street. Some cities publish their own incident maps at block or address level, and those are worth checking where they exist. Everything else that looks address-precise — a grade, a score, a "1 in X" chance for a specific home — is a modeled estimate built by a private company. Models can be useful. They are not counts, and a site that won't tell you which one it's showing you is telling you something.

The five-minute county check

• Open the FBI Crime Data Explorer (cde.ucr.cjis.gov) or your county's page in our county records guides and find the home's county. Note the data year — published FBI county data lags, and 2024 is the current full year.
• Read the rate, not the raw count. A county of a million people will always post bigger counts than a county of 20,000. The comparable figure is reported offenses per 1,000 residents. Across the 3,133 covered counties in 2024, the median county posted 9.75 reported offenses per 1,000 residents.
• Check the composition. What kind of crime drives the number matters more than the number. Nationally in 2024, larceny-theft — packages, bikes, things from cars — was 59.8% of all reported offenses, and outnumbered robbery roughly 20 to 1. Aggravated assault was about 71% of violent offenses. A county whose total is mostly theft from vehicles is a different place to live than one driven by assaults, even at the same rate.
• Read the reporting note before you believe a low number. In 1,661 of the 3,133 covered counties, at least one agency assigned to the county did not report all 12 months of 2024. Where reporting is partial, the published count is a floor, not a total — a "quiet" county may just be an under-reporting one. The FBI file says which agencies reported and for how long; our reports print that note verbatim.
• Then zoom in with local sources. The county figure frames the area; the city police department's own incident map, calls-for-service data, or annual report is where street-level texture lives. If the city publishes nothing, that absence is itself information about how much verification is possible.

Sources: FBI Crime Data Explorer, UCR master files (Return A and Arson), data year 2024, downloaded November 4, 2026; county aggregation and rates computed from that file across 3,133 counties.

A worked example: Palm Beach County, Florida

Palm Beach County posted 26,016 reported offenses in 2024 — 21,641 property offenses and 4,375 violent offenses — a rate of 16.44 per 1,000 residents, above the national county median of 9.75. Read that the honest way: the county contains over 1.5 million people, several cities with very different profiles, and beachfront tourist corridors that pump up theft counts. The county rate tells you the ocean you're swimming in; it does not grade any individual street in Wellington or West Palm Beach, and any tool claiming it does is modeling, not reporting.

The composition check matters here too: 17,956 of the county's reports were larceny-theft. The dominant reported crime problem in Palm Beach County, as in most of America, is theft — not violence. Buyers who read only the total miss the story the total is made of.

Sources: FBI Crime Data Explorer, UCR master files, data year 2024 (Palm Beach County, FL record, retrieved November 2026).

Reported, modeled, and invented: how to tell

Three kinds of crime numbers circulate on listing and research sites:

• Reported offenses (official). Counts of offenses known to law enforcement, from the FBI file or a police department's own publications. Always labeled with a geography (county, city, agency) and a year. This is the only kind that is a count.
• Modeled estimates (private). Companies blend reported data with demographic and geographic modeling to estimate risk at ZIP or block level. Our own sample reports show both grains side by side, labeled: for one West Palm Beach ZIP (33415), a modeled estimate put total crime at 17.18 per 1,000 residents — close to, but not the same thing as, Palm Beach County's reported 16.44 per 1,000. When a number has no agency, no year, and no "reported" label, assume it's modeled and ask what it's made of.
• Invented (marketing). A single "safety score" with no source, no year, and no method. There is no official safety score for any address in America. If the methodology isn't published, the number is decoration.

What a crime check can't tell you

Reported crime is not all crime — many offenses, especially minor theft and some assaults, are never reported to police, and reporting habits vary by place. County data can't see the difference between one end of a county and the other. And last year's data describes last year. What the check can do is replace a vibe with a verified baseline: the real composition of reported crime where you're about to buy, how completely it was reported, and how the area compares with the rest of the country on the same ruler. That is the same homework a registry check does with a different record — and it belongs on the same pre-offer list as flood, tax, and title records in our guide to the 7 public records to check before you offer.

Every CurbFacts report runs this check the way this article describes it: FBI reported offenses at county grain with the reporting note printed, modeled ZIP estimates only where published and labeled as modeled, and the sources shown. You can see the format in the sample report, or run your own address through the free preview.

FAQ

Can I get an official crime count for one specific address?

No. The FBI collects offenses by law-enforcement agency and publishes them at agency and county grain; there is no federal address-level crime dataset. Some individual cities publish incident maps at block or address level — check the local police department. Any nationwide site showing a precise figure for a single house is showing a modeled estimate.

What does "reported offenses" mean?

Offenses that became known to law enforcement and were recorded by agencies participating in the FBI's reporting programs. It excludes crimes never reported to police, and in counties where some agencies reported only part of the year, the count understates the full year. Treat reported figures as a verified floor, not a census of everything that happened.

Why do two websites give me different crime numbers for the same town?

Usually three reasons: different geography (county vs. city vs. ZIP), different years, and different kinds of numbers — one site showing reported offenses, another showing modeled estimates without labeling them. Match the geography, year, and type before comparing any two figures.

Should a high county crime rate kill a purchase?

Not by itself. A county rate averages every community in the county, including areas miles from the home. It is a screening signal: if the county runs hot, do the local work — the city's own incident data, the specific composition of offenses, and a drive through the neighborhood at different hours. Kill a deal on verified specifics, not on a county average.

Is the FBI crime data free?

Yes. The FBI Crime Data Explorer publishes the underlying files at no cost, and state and local agencies publish their own reports free as well. You are never paying for the data itself — only, if you choose, for having it gathered, checked, and read for you.