The United States recorded its lowest crime rates since national record-keeping began in 1976. That is a true sentence and a nearly useless one for anyone deciding where to live, lend, insure, or staff a site. Beneath the single national figure sits a spread of more than seven to one between the safest and least-safe states — and beneath the state figure sits a spread larger still.
“Crime is down” has become the headline of 2026, and by the aggregate numbers it is defensible. But an aggregate is an average, and an average is a summary that discards exactly the information most people are querying a crime dataset to find: where. This post works one level down from the national number — to the states — using the FBI's 2024 data as compiled by USAFacts and the near-real-time picture from the Real-Time Crime Index. The goal is not to rank states for a listicle. It is to show what resolution the data supports, what it does not, and why even the state map is too coarse for most of the decisions people try to make with it.
The national number, and what it averages over
Start with the top line. Per the USAFacts crime-rate explorer (drawn from FBI Uniform Crime Reporting data), the 2024 US violent crime rate was 359.1 per 100,000 and the property crime rate was 1,760.1 per 100,000. Year over year, violent crime fell about 5.4% and property crime about 9%. Measured against 2001, overall crime is down roughly 49%.
The Real-Time Crime Index, which samples reporting agencies on a rolling basis rather than waiting for the FBI's annual compilation, tells a directionally consistent story. Its 2024-vs-2023 comparison — spanning 19,328 agencies covering 340.1 million people — shows violent crime down 4.4%, murder down 14.9%, and property crime down 8.1%. Its most recent window, January through April 2026 against the same months a year earlier, shows steeper declines: violent crime down 6.4%, murder down 18.7%, and property crime down 11.4%. That newer figure rests on a much thinner sample — 566 agencies covering 118.6 million people — and should be read as a preliminary signal, not a settled count.
Two datasets, two jobs. The UCR is comprehensive but lags 12 to 18 months; the RTCI is fast but is a moving sample. They will not agree to the decimal, and they are not supposed to. We covered the distinction in detail in Real-Time Crime Index vs. FBI UCR. For this post, the relevant point is that both agree the decline is real and broad — and both report a single national number that conceals enormous internal variation.
The state spread: seven to one, and it is not close
When USAFacts breaks the 2024 figures out by state, the range is wide enough that the national average describes almost no one well. The violent crime rate runs from Alaska's 724.1 per 100,000 down to Maine's 100.1 — a ratio of roughly 7.2 to 1. Property crime runs from New Mexico's 2,751.1 down to Idaho's 736.3, a ratio of about 3.7 to 1.
| Violent crime, 2024 (per 100k) | Property crime, 2024 (per 100k) |
|---|---|
| Highest Alaska — 724.1 New Mexico — 717.1 Tennessee — 592.3 Arkansas — 579.4 Louisiana — 519.8 | Highest New Mexico — 2,751.1 Colorado — 2,592.8 Washington — 2,466.5 Oregon — 2,388.0 Louisiana — 2,296.4 |
| Lowest Maine — 100.1 New Hampshire — 110.1 Connecticut — 136.0 Rhode Island — 153.6 Wyoming — 203.4 | Lowest Idaho — 736.3 New Hampshire — 918.0 Rhode Island — 1,032.4 Massachusetts — 1,112.1 Maine — 1,142.1 |
Source: USAFacts, “Which states have the least and most crime?” using FBI data. US averages: 359.1 (violent), 1,760.1 (property).
A resident of Maine and a resident of Alaska live in the same country and the same UCR national average, and their exposure to reported violent crime differs by a factor of seven. No policy conclusion, no underwriting model, and no relocation decision that starts from “the US rate is 359” can be trusted, because 359 is a population-weighted midpoint of a distribution this wide.
Violent and property maps do not overlap
A second, easily missed point: the geography of violence and the geography of property crime are not the same geography. New Mexico and Louisiana rank near the top on both lists, but they are the exception. Colorado, Washington, and Oregon post some of the highest property crime rates in the country while sitting nowhere near the top for violence. Conversely, Tennessee and Arkansas rank high for violent crime without appearing among the property-crime leaders.
This is a reminder that “crime” is not one quantity. A state where motor-vehicle theft and larceny run high may reflect metropolitan density, organized retail activity, or shifts in prosecutorial charging — mechanisms with little to do with the interpersonal violence that drives the aggravated-assault and homicide counts. Collapsing both into a single “safety” impression is a category error. The one useful regularity in the table is the Northeast: Maine, New Hampshire, Rhode Island, Connecticut, and Massachusetts appear repeatedly on the low end of both lists.
The Hawaii caveat: how to read a one-year surge
In a year when nearly every state moved down, Hawaii moved sharply up. USAFacts reports the nation's largest year-over-year increase there — violent crime up 16.4% and property crime up 16.6%. It is the kind of number that generates a headline and, sometimes, a causal story to go with it.
Restraint is warranted, for three reasons. First, this is a single year-over-year change, and single-year movements revert more often than they persist; one data point is not a trend. Second, Hawaii starts from a comparatively low base, and percentage changes on small bases exaggerate: a double-digit percentage increase can correspond to a modest absolute change in incident counts. Third, a jump of this size in one jurisdiction should always prompt a reporting-artifact check before a behavioral one — a change in how a large agency classifies or submits records can move a state's rate without any change on the ground. The honest read is that Hawaii is worth watching and not yet worth explaining.
The data-quality asterisk on every state comparison
State-to-state comparisons carry a caveat that rarely makes the headline: some of the variation between states reflects how completely each state reports, not how much crime occurred. The FBI data behind these figures spans two systems — the older Summary Reporting System through 2020 and the incident-based NIBRS through 2024 — and the transition between them was not clean.
When the FBI retired SRS in favor of NIBRS-only reporting for 2021, agencies covering roughly 37% of the US population did not submit usable data that year. Coverage has since recovered, but it recovered unevenly across states. A state whose large agencies were slow to migrate can look artificially low in a given year for a reason that has nothing to do with safety. This is why we treat rankings as approximate and why methodological transparency matters more than a tidy leaderboard. We walked through the mechanics in the incident-taxonomy and normalization problem.
None of this overturns the direction of travel — the decline is large enough and broad enough to survive the reporting noise. But it does mean the precise ordering of states near the middle of the distribution should be read with a wide error bar, and that any product presenting a state rate should be candid about the vintage and completeness of the underlying feed rather than implying a precision the source data does not have.
States are still the wrong unit
Here is the point that matters most for anyone building on this data: even a correct, well-caveated state map is too coarse for nearly every operational decision. The variation within a state dwarfs the variation between states. A low-rate state contains blocks with concentrated, persistent crime; a high-rate state is, block for block, overwhelmingly quiet. Reasoning from a state rate to a specific address is the ecological fallacy in its purest form — attributing to a location the average property of the aggregate that contains it.
This is not an abstract concern. It is the exact mistake that ZIP-code and county crime averages bake into real estate search, insurance underwriting, and site-selection models. A trade area that a county average calls “moderate” can contain a corridor an order of magnitude worse than the county mean and residential blocks an order of magnitude better. We made the underwriting version of this argument in The Business Case for Crime Data in Commercial Real Estate, and the geography version of it in Crime Hotspot Mapping for Developers.
How to use the state map correctly
The state layer is not worthless. It is the right resolution for a narrow set of questions and the wrong resolution for most of the rest. Used well, it does two jobs:
- Calibrating expectations. Knowing that New Mexico and Louisiana run high on both violent and property crime, or that the Northeast runs low on both, is useful prior context — a base rate against which a specific address either confirms or departs.
- Tracking direction. Year-over-year state movement, read against the national trend, flags places worth a closer look — Hawaii being the current example — without licensing a conclusion on its own.
What the state map cannot do is answer the question people actually ask it: is this block, this address, this corridor safe, and how has it moved lately? That question requires data at the resolution of the decision — incident-level records, geocoded to the block, normalized across jurisdictions, and fresh enough to reflect the last month rather than the last federal reporting cycle. The national record low of 2024 is a genuine and underappreciated achievement. It is also, for the person standing on a specific corner, the answer to a question they did not ask.
Access Address-Level Crime Data
Real-time incidents · SpotScore™ safety ratings · 36-month trends · 22,000+ US cities. Normalized and verified — because raw data isn't enough.