How car theft statistics are counted: FBI, NICB and insurer data

The three big sources of car theft numbers count three different things. The FBI counts offenses that police record. The National Insurance Crime Bureau (NICB) counts vehicles reported stolen to the national crime database and ranks them by make and model. The Highway Loss Data Institute (HLDI) counts insurance claims and divides them by the number of cars insured. A popular model can lead one list and sit near the bottom of another without any of them being wrong.

Knowing which number answers which question is the difference between "which cars do thieves take most often?" and "which car is most likely to be stolen from me?"

Source one: the FBI and police-reported crime

The FBI's crime statistics begin with a police report. Local departments record a motor vehicle theft, defined by the FBI as the theft or attempted theft of a motor vehicle, and send their data to the bureau. National totals are then published each year, usually in late summer for the previous year.

That sounds simple, but the way the data are collected changed in 2021, and the change is the biggest source of confusion in year-to-year comparisons.

The move from summary reporting to NIBRS

For decades agencies reported monthly totals through the Summary Reporting System (SRS). The FBI retired SRS on January 1, 2021, and now collects data only through the National Incident-Based Reporting System (NIBRS), according to a Congressional Research Service briefing (May 2022). NIBRS records each incident, with details such as the time of day and up to 10 offenses per incident, instead of one summary line per month. SRS used a "hierarchy rule" that counted only the most serious offense in an incident; NIBRS dropped it.

The problem was participation. In the first NIBRS-only year, 2021, the FBI received data from 9,881 of 18,818 agencies (53%), below its own 60% threshold, and did not release the usual quarterly figures for regions or the nation. The Bureau of Justice Statistics put 2021 coverage at 67.9% of agencies and 65.7% of the U.S. population (BJS NIBRS estimation project). Coverage has since improved: a Congressional Research Service report (August 2025) said about 76% of agencies, covering about 87% of the population, reported through NIBRS as of the end of 2024. Even then, the five most populous states (California, Texas, Florida, New York and Pennsylvania) had only 67% population coverage, which matters because California and Texas lead the nation in thefts.

Does the change distort theft counts?

Less than you might fear. The CRS report on NIBRS notes that motor vehicle theft was already counted in its own category under SRS, so the hierarchy rule changed it less than other crimes. Comparison studies found motor vehicle theft rates about 4.5% higher under NIBRS in a 2000 BJS study and about 2.7% higher in a 2014 FBI study. The larger issue is coverage: when only part of the country reports, the national figure is an estimate, built by weighting the agencies that did report. The estimate for 2021 is not comparable to other years, a point USAFacts repeats in its property crime charts, which use FBI data.

On that basis, the FBI's national rate for 2024 was 259 motor vehicle thefts per 100,000 people, down 19.5% from 321 in 2023, per the USAFacts charts. A rate like this is more useful than a raw count when population changes, which leads to the next point.

Source two: the Bureau of Justice Statistics survey

Police data only include thefts that were reported. A second federal source, the National Crime Victimization Survey (NCVS), asks a sample of households about crimes they experienced whether or not they called the police. Stolen cars are among the best-reported property crimes, since an insurance claim normally requires a police report, but the two systems still do not match.

A November 2025 BJS report comparing the two shows how different the units are:

These cannot be subtracted from each other, because one is per person and one is per household, and the survey is a sample with a margin of error. The lesson for a reader: even two federal sources disagree on levels, and neither is "the" number.

Source three: NICB and the national crime database

NICB is a nonprofit funded by insurers; our guide to the National Insurance Crime Bureau explains its role. For theft, it analyzes vehicle theft records that law enforcement enter into the FBI's National Crime Information Center (NCIC). In a 2021 release about its annual "Hot Wheels" report, NICB said it examines theft data submitted to NCIC and identifies the makes, models and model years most reported stolen, as our The most stolen cars in America guide notes.

Its latest published results, in the 2025 Vehicle Theft Report released in March 2026, counted 659,880 thefts nationwide in 2025, down 23% from 2024 (Insurance Journal). It reported California (136,988), Texas (75,269), Illinois (28,327), Florida (27,142) and New York (24,206) as the states with the most thefts, and said more than a third of thefts occurred in the 10 metro areas with the highest volumes, led by Los Angeles-Long Beach-Anaheim at 53,911. Hyundai and Kia made up 14% of thefts in 2025, down from 16% in 2024 and 21% in 2023.

Why NICB and FBI totals do not match

Both ultimately depend on police reports, but they take different routes:

• Data pipeline. The FBI publishes estimates from NIBRS incident reports, weighted for agencies that did not report. NICB works from vehicle records entered in NCIC, which is a database for locating stolen property, not a crime-statistics system.
• Definitions. What counts as a "vehicle" (cars only, or motorcycles and trailers too) and whether attempted thefts are included can differ, and each organization chooses its own scope.
• Timing. NICB published full-year 2025 numbers in March 2026. The FBI's 2024 data came out in August 2025. When two numbers are published months apart, one may have been revised.
• Unit. NICB's lists are counts. The FBI headline is often a rate per 100,000 people.

For a shopper, the practical rule is to use one source consistently when comparing years. A decline from NICB's own series (850,708 thefts in 2024 to 659,880 in 2025) is a fair comparison. A comparison between NICB's count in one year and the FBI's estimate in another is not.

Source four: insurers and HLDI

Insurers care about claims, not police counts. The Highway Loss Data Institute, the research arm that works with the Insurance Institute for Highway Safety, measures theft by what insurers pay. Our guide to IIHS and HLDI insurance loss data: what it shows about cars covers the program; the theft method is the part that matters here.

HLDI does not have a flag in its data for a whole-vehicle theft. In its August 2025 analysis it compared the dollar amount of each theft claim with what insurers pay for the same vehicle when it is totaled in a crash, because most theft claims do not say whether the entire vehicle or only parts or contents were stolen. It reports claim frequency per 1,000 insured vehicle years, where one insured vehicle year is one vehicle insured for one year (two vehicles insured for six months each equal one).

In that analysis of 2022-24 models, the Chevrolet Camaro ZL1 had a relative theft claim frequency of 3,949, where 100 is the average for all vehicles, or about 39 times average, and the standard Camaro was about 13 times average. The 20 lowest included mostly electric or plug-in hybrid vehicles, with the Tesla Model 3 four-wheel-drive at 1 (IIHS, August 2025). The Camaro ZL1 is not on NICB's top ten by count, because relatively few exist.

HLDI's approach has its own limits. It covers only vehicles with comprehensive coverage, since that is the coverage that pays for theft. Its ranking is limited to a few model years. And partial thefts (catalytic converters, wheels, contents) are in the broader theft frequency but excluded from whole-vehicle rankings.

Counts versus rates: a worked example

Here is a hypothetical example, with made-up numbers, to show why a raw count misleads. Say two models exist:

Model A tops a count-based list, with 22 times the thefts. But any single Model B is more than three times as likely to be stolen. Which fact matters depends on your question. An insurer pricing Model B per vehicle cares about the rate. A police department deciding where to put patrols cares about the count. NICB's list is built for the second question; HLDI's for the first.

Why a popular model tops raw lists

Three forces push common cars to the top of count-based lists:

1. Volume. The more of a model there are, the more chances a thief has to find one.
2. Age. Older cars lack modern immobilizers and are easier to steal. In NICB's list of 2020 thefts, the most commonly stolen Honda Civic model year was 2000, according to our guide to the most-stolen cars.
3. Demand for parts and resale. Models with large parts markets are stolen to be dismantled, as described in Where stolen cars go: export, chop shops and VIN cloning.

Hyundai and Kia show a fourth: a specific weakness in certain models that spread quickly once thieves learned of it, and the share of thefts fell after software updates. HLDI found that, for eligible Hyundai and Kia vehicles with the update, all theft claims were 46% lower and whole-vehicle theft claims 52% lower (August 2025 analysis). Our guide to the Kia and Hyundai theft wave covers the insurance effects.

Geography, reporting lag and recovery

Where. Theft counts follow population and vehicle density, so California and Texas lead any count. Rates per 1,000 residents or vehicles can tell a different story, and small states can swing sharply. Territory is a rating factor for exactly this reason; see territory rating. If you live in California or Texas, the headline state count tells you little about your own neighborhood.

When. A theft is counted when it is reported, not when it happens or when the car is found. A report filed in December may land in January's data. Annual reports are also backward-looking by months, so a list published in March 2026 describes 2025, not today.

Recovered or not. Statistics generally count a theft even if the car is found the next day. A theft total tells you how often thieves tried or succeeded, not how many owners lost a car permanently. Recovery rates are tracked separately and vary by vehicle type, and some of the published recovery figures conflict between sources, so treat any single percentage with caution.

How to read a ranking for your own risk

Use this checklist before drawing a conclusion from any "most stolen" list:

1. Who published it, and what is the data source? Police reports (FBI, NICB) or insurance claims (HLDI)?
2. Is it a count or a rate? Counts favor common models; rates favor rare ones.
3. Which model years? A 2000 Civic and a 2024 Civic are different risks.
4. Whole vehicles or any theft claim? Parts theft inflates some lists.
5. Which place? A national list may not describe your ZIP code.
6. What changed since? Software updates, new security features or a social-media trend can shift risk within a year.
7. Does your own coverage line up? Check that you carry comprehensive coverage, since that is what pays for theft, and what the deductible is.

What insurers do with the data

Insurers pull from both worlds. They use claims data, much closer to HLDI's method than to a police count, to set a vehicle symbol for each model, and territory data to price location. Then your own claims history and other rating factors adjust the result. That is why a model on a popular list may not cost much more to insure, while a model on nobody's list might. For how rate studies are built and misread more generally, see how to read car insurance rate studies and average-cost claims and our related article How to read a Highway Loss Data Institute loss bulletin.

Common questions

Why did thefts fall so much in 2024 and 2025?

NICB reported thefts fell 17% in 2024 and 23% in 2025, and the FBI's rate fell about 19.5% in 2024. Explanations offered by NICB include the Hyundai and Kia software updates and other prevention partnerships. Cause is hard to prove from counts alone; a decline can reflect fewer attempts, better security, a change in reporting, or all three.

Which statistic should I trust for my city?

Use the police department's own published data or a state agency, and compare the same measure over time. National lists are useful for context. For insurance, the better question is what your insurer charges for comprehensive coverage on your model in your ZIP code, which you can see by getting quotes.

Is the most-stolen car also the most expensive to insure?

Not necessarily. A common sedan can top a count-based list while its per-vehicle risk is modest, so its comprehensive premium may be ordinary. Insurers price on rates by model and territory, not on rankings.

Do these statistics include attempted thefts or stolen parts?

It depends on the source. The FBI definition includes attempted vehicle theft. NICB's lists are built from vehicle records. HLDI's whole-vehicle rankings exclude partial thefts, while its broader theft claim frequencies include them. Always check the methodology note before comparing two numbers.

Key takeaways

• The FBI counts police-recorded offenses, NICB counts vehicles reported stolen and ranks models, and HLDI counts insurance claims per 1,000 insured vehicle years. Their totals and rankings should not be mixed.
• The FBI's 2021 switch to NIBRS cut participation (65.7% of the population that year, about 87% by the end of 2024), so compare years within one source and treat 2021 with care.
• A raw count favors common models; a rate favors rare, high-risk ones. Decide which question you are asking before reading a list.
• Theft is counted when reported, not when the car is found, so totals say little about permanent loss.
• For your own risk, look at the model year, your location, whether parts thefts are included, and your comprehensive coverage and deductible.