Crime Rate by Address: Where the Numbers Come From and How to Read Them

"How safe is this neighborhood?" is the second question buyers ask after the price, and the internet answers a search for crime statistics by address with confident letter grades for any house. The grades are not measured at the door: they are estimates derived from statistics that police agencies report for whole cities and counties, stretched over ZIP codes and census tracts by models that rarely publish their error. The underlying data is still the best evidence there is, if you know what it counts, what it misses and how to compare it. This article is the crime chapter of our pre-offer checklist on what to check before buying a house.

Where official crime data comes from

Three official systems produce almost every crime number you will see, and each answers a different question.

The FBI's UCR Program collects crime statistics from more than 19,000 federal, state, county, tribal and university agencies, which submit voluntarily through a state program or directly to the FBI, and publishes them on the Crime Data Explorer; since 2021 the FBI has been phasing out the older summary counts in favor of the incident-based NIBRS³. The 2024 release, published in August 2025, drew on more than 16,000 agencies covering 95.6 percent of the U.S. population and detailed more than 14 million offenses, submitted through NIBRS and the legacy Summary Reporting System¹. Because participation is voluntary, coverage is high but not complete.

The National Crime Victimization Survey is the other half of the picture. Each year the Bureau of Justice Statistics interviews a nationally representative sample of about 240,000 persons in about 150,000 households about crimes they experienced, whether or not they told the police⁴. Its annual report, Criminal Victimization, the 52nd in a series begun in 1973, is the official source for how much crime never enters police statistics². It cannot tell you about your city, but it calibrates everything else.

Local open data is where the detail lives. The City of Los Angeles publishes every LAPD-recorded incident from 2020 to 2024 with the date of occurrence, the LAPD area, the crime description and coordinates⁵. Many large cities run a similar portal; where none exists, the agency-level FBI figure is all you have.

Crime rate by address: why it really means by agency, city or county

Nobody counts crime by address. The counting unit is the agency: a city police department, a county sheriff, a campus or transit police force. The FBI publishes each agency's figures along with estimates for the states and the nation³¹; a county figure, wherever you see one, is a sum over the agencies inside the county with an allowance for those that did not report. When a website shows a score for a ZIP code or a census tract, it has spread those agency totals (or the incident-level open data, where it exists) over smaller areas with a model. A crime rate by ZIP code is always derived, never counted, and the resolution you see is finer than the resolution that was measured.

Three gaps sit underneath every map:

• Reporting to police. For 2024 the NCVS counted 23.3 violent victimizations for every 1,000 persons age 12 or older; 11.2 per 1,000 had been reported to police². Roughly half of violent crime is invisible to any police-based map, and the reported share varies by crime type and community.
• Agency participation and system changes. Coverage of 95.6 percent is not 100 percent¹. Agencies also change record systems: LAPD moved to a NIBRS-compliant system on March 7, 2024, froze its legacy 2020 to 2024 dataset and has published NIBRS datasets since October 2024⁵. A chart spanning that break compares two counting methods.
• Location precision. Incident-level data is published with approximate locations to protect privacy; LAPD gives addresses only to the nearest hundred block, so even the finest official data describes a block, not a house⁵.

Read any address-level grade as a modeled estimate for the surrounding area, and go one level up to the figure that was actually counted.

Reading rates per 100,000

A rate is offenses divided by resident population, multiplied by 100,000. The population is the Census Bureau's annual resident population estimate for the county or city⁶; the offenses come from the agency. Four habits keep the arithmetic honest.

1. Separate violent from property crime. The FBI's violent-crime estimate comprises murder, rape, robbery and aggravated assault¹; property crime (burglary, larceny-theft and motor vehicle theft) is published separately. They have different causes, reporting rates and meaning for a homebuyer. Never add them.
2. Compare like with like. A police-recorded rate can only be compared with another police-recorded rate for the same year and category. The NCVS violent rate of 23.3 per 1,000 includes simple assault and unreported incidents²; beside an FBI figure it looks like a different country; it is a different instrument.
3. Prefer trends to snapshots. Nationally, the FBI estimated that violent crime fell 4.5 percent from 2023 to 2024, murder 14.9 percent, robbery 8.9 percent, aggravated assault 3.0 percent and rape 5.2 percent¹. A county whose rate is flat while the nation falls is losing ground in relative terms. For small areas use three-year averages; one year in a small town can swing on a handful of incidents.
4. Check the denominator. Rates per resident assume the people at risk are the residents; where they are not, the rate says less than it seems.

Free tools and paid scores

The official tools are free, and everything commercial starts from them. A paid score can save time by assembling several agencies and years in one view; it cannot know more than the agencies reported. If a site does not say which agency data it uses and how it maps it to your ZIP code, treat its grade as a color, not a measurement.

What the numbers leave out

The dark figure. That roughly half of violent victimizations never reach the police² is a national average; the share varies by offense, and the reasons victims give for not reporting are part of the NCVS⁴. A low police-recorded rate can mean little crime or little reporting.

Denominators that do not match the people present. Rates are per resident, using the Census Bureau's population estimates⁶. A downtown, a beach town in August or a neighborhood with a stadium hosts many times its resident population; every incident involving a visitor counts in the numerator while the visitors are absent from the denominator, so such areas look worse per resident than the risk to any one person. The reverse holds for bedroom suburbs.

Category mix. Two counties with the same property-crime rate can differ completely: shoplifting concentrated around a mall versus residential burglary spread across streets. Only incident-level data shows which, and only for agencies that publish it⁵.

What is not in any dataset. Crime statistics do not measure how a street feels at 10 p.m., whether the neighbors know each other, or how fast the police respond. Those you learn by going there.

Fair Housing: statistics about places, never about people

The Fair Housing Act of 1968, 42 U.S.C. 3601 et seq., prohibits discrimination by housing providers, lenders, insurers and municipalities, and the Department of Justice names steering home seekers to certain areas among the concealed practices its testing program uncovers⁷. The protected classes are race, color, national origin, religion, sex, familial status and disability⁸.

For crime statistics this sets a simple discipline:

• Statistics describe an area and a period. They say nothing about any resident, household or group, and must never be broken down by who lives there.
• A rate is a count divided by a population, not a judgment about a neighborhood's people.
• Agents should point buyers to the official sources rather than characterize an area's safety; buyers should ask for the data, not for an opinion about the people.

If a crime map offers demographic overlays next to its grades, close it. That is the mechanism the Fair Housing Act was written to stop.

Worked example for a Los Angeles address

Example with illustrative figures: the method, not a statistic for any real county. You are considering a house in Los Angeles County.

Step 1, the counted figure. In the Crime Data Explorer you pull the violent offenses for the latest full year from the agencies that cover the county, the sheriff's department and the city police departments, and the county's resident population from the Census Bureau's estimates³⁶. Suppose, for the arithmetic, 4,500 violent offenses and 1,000,000 residents: 4,500 ÷ 1,000,000 × 100,000 = 450 per 100,000. Read the national rate for the same year and category from the same tool, and note whether the county sits above or below it.

Step 2, the trend. Repeat for the two previous years. If the county's rate fell each year while the nation fell about 4.5 percent in 2024¹, the county is at least keeping pace. If it rose while the nation fell, that is the finding, not the level.

Step 3, the area around the house. In the LAPD dataset you filter by the LAPD area covering the address and by year of occurrence, and count incidents by category; for 2024 and later you switch to the NIBRS datasets, because the legacy series stopped with the March 2024 system change⁵. You are looking for the mix: residential burglary and vehicle theft matter to a homeowner in a way that shoplifting at a mall does not.

Step 4, the sanity check. Your county figure of 450 per 100,000 is not comparable with the NCVS figure of 23.3 per 1,000 (2,330 per 100,000 persons age 12 or older), because the survey counts simple assaults and unreported incidents that no police series contains². Use the survey for reporting gaps and the police series for comparing places.

Half an hour of free work gives you a level, a trend, a category mix and an understanding of what each number omits: more than any letter grade tells you.

What CheckValue shows and where the score stops

CheckValue's valuation report names the official source for each of its figures, and its safety section follows the discipline above. For addresses in the United States, Germany, Austria, Switzerland, France, Spain and Italy the report shows a crime score from 1 to 10 taken from official police-recorded statistics: in the United States at county level from FBI data, in the other countries at the level the national statistics use: district, federal state, canton, département, province or municipality. The score is a rank: 10 means the safest tenth of the country's districts by recorded offenses per 100,000 inhabitants, 1 the tenth with the most offenses, drawn as a ten-segment bar against the national average next to the flood, noise and energy maps and the valuation itself.

What it is not: a statement about a street, a house or the people who live there. Two addresses in the same county get the same score however different their blocks feel, and the report says so. Outside those seven countries the report shows no crime score rather than a modeled substitute. The score is part of the paid report, priced from $3.99 in the app and €4.99 for a web purchase; the free preview shows the report layout for your address first, and the help pages explain the sources behind each section. For the other location layers, see the guides to the flood zone by address and to noise maps, energy class and other neighborhood factors.

A buyer's routine

Put the pieces in this order and the whole check takes an evening.

1. The counted figure and its trend: the agency or county rate per 100,000 for violent and property crime, three years, against the national figure for the same years³¹.
2. The incident mix for the police area, from the city's open data where it exists⁵: burglary, vehicle break-ins and yard theft are the categories that touch an owner.
3. A walk at three times of day: a weekday morning, the evening rush and after dark on a weekend. Look at lighting, foot traffic and whether front yards are used.
4. Local news for the past twelve months, for the street name and the nearest cross streets. One incident is an event; a pattern is information.
5. Conversations about things, not people. Ask neighbors and the seller's disclosure about break-ins, package theft, car damage. Never about who lives where.
6. Weight, don't decide. A recorded rate at or below the national level with a stable trend is context, not a reason to bid more; a rising trend is a reason to look harder, not a verdict. Does crime affect property value? Over years and through buyer perception, yes, which is why crime statistics enter the offer as one input in the pre-offer checklist, and as one of the factors that move home values over years rather than months.

Statistics about places, read correctly, will not tell you whether to buy a particular house; they tell you which questions to ask when you stand in front of it. This is general information, not legal advice.

This article is general information, not legal, tax or investment advice. Figures and rules carry the year they were published; check the cited source for the current version.

Christian Eckmair

Co-founder, CheckValue · More than 20 years in real estate in Tenerife and Austria · Reviewed by Josef Eckmair MBA

Christian Eckmair is co-founder of CheckValue and of CBDNOL GmbH (Ansfelden, Austria). He has worked in real estate for more than 20 years in Tenerife and Austria – buying, renovating, letting and selling residential property – and writes about home values, buying checks and ownership records.

Articles by Christian Eckmair →