Crash statistics and premiums: reading NHTSA FARS and IIHS data

Not directly. Headline traffic-death numbers describe the rarest and most severe kind of crash, while your premium is built from insurance claims of every size: dented bumpers, whiplash, stolen wheels, and now and then a death. Fatality data from the federal Fatality Analysis Reporting System (FARS) reach your rate only indirectly, through the cost of serious injury claims, the behaviors insurers surcharge, and the vehicle safety research that uses crash data alongside claims. In 2024 the country recorded 39,254 traffic deaths, but the tens of thousands of fatal crashes were less than 1% of the roughly 6.18 million crashes police reported.

This article explains what FARS counts and leaves out, why a "rate per 100 million miles" is a different statement from a raw count, what the data say about speeding, alcohol, distraction and seat belts, how IIHS reuses the data, and how to read a state ranking without drawing the wrong conclusion about premiums.

What FARS counts, and what it leaves out

FARS is the federal census of deaths on U.S. roads. According to the National Highway Traffic Safety Administration (NHTSA), as described in its April 2026 overview of 2024 crashes, FARS holds data on every fatal motor vehicle traffic crash in the 50 states, the District of Columbia and Puerto Rico. To be included, a crash must involve a motor vehicle traveling on a trafficway customarily open to the public and must result in the death of a vehicle occupant or a nonoccupant within 30 days of the crash.

That definition has consequences:

• It counts deaths, not injuries or damage. A crash with nine serious injuries and no death is invisible to FARS. NHTSA estimates non-fatal crashes separately from a sample of police reports.
• It leaves out most private property. NHTSA's 2023 state fact sheet says crashes on private property not regularly used by the public, including some parts of parking lots and driveways, are excluded. A collision in a private lot can still produce a collision or liability claim, but FARS would not count it.
• It does not capture insurance. FARS has no premium, no policy, no insurer and no claim payment. Nothing in it can say what a crash cost an insurer.
• Its numbers get revised. The most recent year is an Annual Report File, which can change when the Final File is released the next year. NHTSA's overview notes that the final count for 2023 was 41,025, updated from 40,901 in the 2023 file. If you see two slightly different totals for the same year, one is probably preliminary.
• Definitions shift. Since 2022, people on motorized bicycles are counted as pedalcyclists, and crashes involving only motorized bicycles no longer appear. NHTSA also changed how it classifies vehicle types in 2020, so vehicle-type comparisons with earlier years are not reliable.

Fatal crashes against the claims that set your premium

Insurance premiums are priced on the frequency and severity of claims. For a deeper look, see claim frequency vs severity: which one actually moves premiums?. Set the crash census next to the claims data, and the proportions are striking.

The NHTSA figures come from its 2024 overview; the claim frequencies are ISO (a Verisk business) numbers published by the Insurance Information Institute. The ISO bodily injury figures exclude Massachusetts and most no-fault states, and the claims data are a sample of insured vehicles, not a count of every crash, so the two sets are not directly comparable. The point is proportion: about six of every thousand police-reported crashes involved a death, and the ordinary claim that sets most premiums is a repair bill.

Fatal crashes are rare per driver but enormous per event, and insurance treats that asymmetry in a specific way.

A death is a bodily injury claim capped by a limit. If you cause a fatal crash, your insurer's payment is generally limited by your policy limit. A $25,000 per-person limit pays $25,000 whether the injury is a broken wrist or a death. So for a minimum-limits driver a death typically adds the limit, plus defense costs, to the insurer's bill rather than the full value of a life. Higher limits mean a higher potential cost per fatal crash, which is part of how liability pricing climbs with the limit; see fatal crashes and liability limits: wrongful death claims explained.

The average bodily injury claim is far smaller than a fatal one. In the ISO series, the average bodily injury liability claim cost $28,278 in 2024, up from $26,178 in 2023. That is an average over every injury claim, from sprains to catastrophic harm. Rising medical prices and legal costs push that average up; see how medical costs and injury claims shape what you pay for liability.

Deaths fell in 2024 while injury claim costs rose. NHTSA recorded a 4.3% decline in deaths from 2023 to 2024. In the same year the ISO average bodily injury severity rose by about 8%. The two series moved in opposite directions because they measure different things: one counts how many people died, the other what insurers paid per injury claim.

Raw counts versus rates per 100 million miles

A raw count answers "how many." It does not say "how risky," because a large, busy state will have more deaths than a small one. Safety analysts therefore use rates. NHTSA's headline measure is deaths per 100 million vehicle miles traveled (VMT). In 2024 the national rate was 1.19, down 5.6% from 1.26 in 2023, as vehicle miles rose about 1.5% to 3,294 billion while deaths fell.

Other denominators tell different stories. NHTSA's state tables for 2023 give four rates for each state: per 100 million miles, per 100,000 residents, per 100,000 licensed drivers and per 100,000 registered vehicles. Mississippi had the highest rate on each of the four in 2023 (1.79 per 100 million miles and 24.90 per 100,000 population), and Massachusetts the lowest (0.56 and 4.90). In 2023 the two extremes were the same on all four measures. IIHS's Fatality Facts for 2024, posted in June 2026, show Massachusetts lowest per mile (0.59) but Rhode Island lowest per resident (4.7), with Mississippi highest on both (1.81 and 25.6).

A worked example shows why the denominator matters. These figures are made up for illustration:

• State A: 1,000 deaths on 80 billion miles of travel is 1.25 per 100 million miles.
• State B: 600 deaths on 30 billion miles is 2.0 per 100 million miles.

State A has more deaths and looks worse in a headline, yet driving in State B is riskier per mile. A state where people drive long distances (large rural states) may also have a higher rate per resident than per mile, and a dense city-state may look very safe per resident but not per mile. When you read a ranking, check the denominator in the footnote before you compare any two places.

What the data say about speeding, alcohol, distraction and seat belts

The federal data break deaths into behaviors, which is where the connection to insurance rules is clearest. All figures below are from NHTSA's 2024 overview.

Four cautions come with those numbers.

1. The categories overlap. A crash can be both speeding-related and alcohol-involved, so the rows cannot be added. NHTSA reports that 1,565 passenger vehicle drivers in fatal crashes showed all three of speeding, alcohol impairment and no restraint.
2. Most drivers in fatal crashes had none of the three. Of 41,912 passenger vehicle drivers in fatal crashes in 2024, 43% had at least one of the three behaviors and 57% had none. The behaviors are major contributors, not the whole story.
3. The definitions are specific. A crash is "speeding-related" if any driver was charged with a speeding offense or an officer indicated that racing, driving too fast for conditions or exceeding the limit was a factor. Alcohol-impaired counts use measured blood alcohol levels when reported, and NHTSA estimates the level when tests are missing. Restraint use was unknown for 11% of passenger vehicle occupant deaths. Distraction counts only crashes in which at least one driver was recorded as distracted.
4. Seat belts show up as a survival gap. Among passenger vehicle occupants who survived fatal crashes, 13% were unrestrained, compared with 48% of those who died. NHTSA's 2024 survey put front-seat belt use among adults at 91.2%. The data show belts change who survives a severe crash; they do not tell you what any insurer charges.

These factors connect to premiums through the rating system rather than through the totals. A conviction for speeding or drunk driving is a rating event: it usually brings a surcharge, and a DUI can bring years of higher rates or, in many states, a required SR-22 filing, as laid out in how a speeding ticket affects your car insurance. The insurer prices that behavior using its own claims history for drivers with such records, not by consulting FARS. The federal data matter because they inform the laws, enforcement priorities and safety standards that determine what counts as a violation and how likely it is to be caught.

What the IIHS version adds

The Insurance Institute for Highway Safety (IIHS) and its sister organization, the Highway Loss Data Institute (HLDI), analyze FARS in their Fatality Facts series. For 2024 IIHS reports the same 39,254 deaths and breaks them out: 58% were passenger vehicle occupants (22,713), 18% pedestrians (7,080) and 16% motorcyclists (6,228). It puts speeding in 29% of deaths, a little more common on minor roads (33%) than on interstates and freeways (29%). IIHS notes the national death rate per 100,000 people was 11.5 in 2024, 44% lower than in 1975, with the rate per mile down 64%, and deaths have fallen since the 2021 peak of 43,230.

IIHS adds three things that NHTSA's tables do not emphasize:

• A state-by-state comparison on its Fatality Facts pages with both per-resident and per-mile rates.
• Breakdowns by road user and crash type, which help show why some places with many pedestrians or motorcyclists have different risk profiles.
• A tie to insurance claims data. HLDI collects claims from participating insurers and compares vehicles by their claims, which crash counts alone cannot do. How to read a Highway Loss Data Institute loss bulletin shows how to read those tables, and IIHS and HLDI insurance loss data: what it shows about cars explains the two organizations' roles.

Both organizations draw on the same FARS file, so their totals match for a given year. Differences between a news story and an official table usually come from using a preliminary count, a different year or a different denominator, not from competing data.

How insurers use crash data indirectly

An insurer sets rates from the claims it pays and, for smaller companies, the loss costs a rating bureau such as ISO compiles; see Verisk (ISO) advisory loss costs: the data behind many insurers' rates. Crash data enter this process at the edges:

1. Rating factors. Territory, age, vehicle and driving record are chosen because claims data show they predict loss. Fatal crash patterns can motivate research into those factors, but the factors themselves are validated on claims. The rating factors glossary entry and territory rating explain the mechanics.
2. Severity trends. Fatal and serious injury crashes drive the largest bodily injury claims. Actuaries watch medical costs and settlement patterns because they show up in severity.
3. Vehicle design. HLDI's work links crash-avoidance features to claim frequency. Fewer crashes of any size should eventually lower losses, but only where the technology is on enough cars.
4. Law and enforcement. Belt laws, blood alcohol limits and speed enforcement follow safety research. Changes to them alter which events result in violations that bring surcharges.
5. Regulatory context. Federal analyses of the economic cost of crashes, such as the one covered in What NHTSA counts in the economic cost of motor vehicle crashes, frame the stakes but do not set any premium.

A look at states: fatality rates and premiums

If death rates drove premiums, the ranking of states by death rate would roughly match the ranking by cost. It does not. This table sets NHTSA's 2023 fatality rate per 100 million miles beside the NAIC's 2023 average auto insurance expenditure, as reported in the NAIC's Auto Insurance Database Report, released in February 2026.

New York and New Jersey have fatality rates well below the national rate and some of the highest premiums; Idaho's rate is above the national figure and its premium is among the lowest. Premiums in such states typically reflect claim frequency, repair and medical costs, litigation and, in no-fault states, personal injury protection benefits, none of which show up in a death count. For a fuller treatment of state price differences, see the most and least expensive states for car insurance and average expenditure vs average premium vs quoted rate.

A checklist for reading a crash-statistics headline

Before you let a headline shape a decision about insurance, run through these steps.

1. Find the year and file. Is the number preliminary (an early estimate or an Annual Report File) or final? Early estimates can differ from final counts.
2. Ask whether it is a count or a rate. If it is a rate, check the denominator: miles, residents, drivers or vehicles.
3. Check who is counted. All deaths, or only vehicle occupants? Passenger vehicles only, or motorcycles and pedestrians too?
4. Read the definition. "Speeding-related," "alcohol-impaired" and "distraction-affected" are defined terms, not everyday meanings.
5. Look for the overlap. Behavior categories overlap, so percentages that sum to over 100% are normal.
6. Ask what insurance data would be needed. To say anything about premiums you would need claims, exposure and rates, and FARS has none.
7. Check the date on any premium. Pair crash statistics with premium figures from the same year, as in the table above, and say which source each came from. fact-checking car insurance headlines about percent increases covers the same discipline for percent-change claims.

Common questions

If traffic deaths fell, why didn't my premium fall?

Premiums follow claim costs, and the main cost drivers in recent years have been repair prices, used-car values and medical and legal costs. Fewer deaths in a year does not mean cheaper repairs. In 2024 deaths fell 4.3% while average bodily injury and property damage claim costs, in the ISO data, kept rising.

Do insurers use FARS to price my policy?

Generally no. Rates are filed with state regulators and rest on the company's claims, expenses and projected trends, sometimes supplemented by ISO loss costs. FARS is research and policy data. Ask your insurer or read its rate filing if you want to know what data it relies on.

Does a high fatality rate in my state mean I will pay more?

Not reliably. As the table shows, some low-fatality states are expensive and some higher-fatality states are cheap. Your price is more strongly linked to claim frequency, repair and medical costs, litigation and state insurance rules than to the death rate.

Where can I look up crash data for my state?

NHTSA's CrashStats site publishes state fact sheets and the FARS data tables, and its overview lists tools such as the Fatality and Injury Reporting System Tool and State Traffic Safety Information pages. IIHS publishes state-by-state Fatality Facts. Your state's department of transportation or highway safety office may publish more detailed local data.

Key takeaways

• FARS counts deaths within 30 days on public trafficways. It contains no insurance information and misses non-fatal crashes and most private-property incidents.
• Fatal crashes were about 36,000 of roughly 6.18 million police-reported crashes in 2024. Premiums are set mainly by the much more frequent injury and damage claims.
• Compare rates per 100 million miles, not raw counts, and check the denominator: Massachusetts is lowest per mile while Rhode Island was lowest per resident in IIHS's 2024 data.
• Speeding, alcohol and non-use of belts are major factors in deaths, but most drivers in fatal crashes had none of the three, and the behaviors reach your premium through violations and surcharges rather than through FARS totals.
• State fatality rankings do not track state premium rankings, so do not read a death rate as a price forecast.