Miles driven, gas prices and claim frequency: why premiums follow

More driving means more chances for a crash, so claim counts tend to rise and fall with miles. But the link is loose, and the effect on your premium arrives late. In 2020 U.S. driving fell about 11%, non-fatal crashes fell 22%, and the government counted more traffic deaths than the year before. Insurers saw fewer claims, then watched the cost of each claim climb, and rates followed the costs with a delay of a year or more.

This article explains the exposure logic insurers use, what the 2020 natural experiment showed, how much gas prices really change driving, and why the price you pay today reflects accidents from years ago.

Frequency is an exposure story

An insurer prices risk per unit of exposure. For private passenger auto, the standard unit is the car-year: one vehicle insured for 365 days. The NAIC's Auto Insurance Database Report (2022/2023 edition, adopted December 2025) defines exposure that way and measures claim frequency as claims per 100 earned car-years.

Notice what the unit leaves out: how far each car travels. Miles enter pricing indirectly, through annual mileage classes and commute-or-pleasure use, which are among the rating factors on most policies. Two cars insured for a year are the same exposure on paper even if one is driven 4,000 miles and the other 20,000. That is the opening for pay-per-mile pricing, covered below.

Several state-level facts show why miles matter. The same NAIC report lists "miles driven per number of highway miles" among three variables correlated with state auto insurance premiums, along with urban population and disposable income, and notes that high-premium states tend to be highly urban with greater traffic density. Its traffic table, using Federal Highway Administration data for 2022, puts the nation at about 761,000 vehicle miles per mile of road. Iowa was near 285,000, Montana near 181,000, and the District of Columbia, Hawaii and New Jersey were at roughly 2.25 million, 2.27 million and 1.94 million. More traffic packed onto each mile of road is more chances for contact between vehicles. It is one reason the same driver can cost more to insure in a dense state, and why claim frequency vs severity: which one actually moves premiums? separates how often claims happen from how large they are.

What 2020 showed

The pandemic produced a test insurers never planned. According to the National Highway Traffic Safety Administration's Overview of Motor Vehicle Crashes in 2020 (March 2022), annual vehicle miles traveled fell 11%, from 3,262 billion in 2019 to 2,904 billion in 2020, the largest year-to-year decline since 1942 to 1943. The table pulls together NHTSA's figures with industry claim data from Triple-I (the Insurance Information Institute), which reports ISO claim frequency per 100 earned car-years in its auto insurance facts page.

The percentage changes for ISO frequencies are my calculation from Triple-I's published figures. Three patterns stand out.

Claims fell more than miles did. Miles dropped 11%; collision claims dropped about a fifth, and property-damage-only crashes fell 25% according to NHTSA. Crashes per mile fell too: NHTSA reports the property-damage-only crash rate per 100 million miles fell 15% and the injury crash rate fell 6.8%. Fewer cars on the road at once is one plausible reason, but the data do not say why.

Comprehensive barely moved. Many comprehensive claims, such as theft, hail, flood and vandalism, happen to parked cars, which fits that line staying near 3 claims per 100 car-years.

Insurers acted on it. According to Insurance Journal, Triple-I put customer refunds at between $8 billion and $10.5 billion as of mid-April 2020, and an S&P Global Ratings analyst said discounts ranged from 15% to 20% and typically ran about two months, subject to state regulator approval. The analyst described miles driven as "the rating variable most correlated to loss amounts."

Why deaths per mile rose when miles fell

The most counterintuitive number in the table is the fatality rate. The 21% jump to 1.34 deaths per 100 million miles was the largest percentage increase on record, according to NHTSA. The agency's report points to behavior. Compared with 2019, speeding-related deaths rose 17%, alcohol-impaired-driving deaths rose 14%, and deaths of unrestrained passenger vehicle occupants rose 14%. Nighttime deaths rose 12%, and daytime deaths rose only 1.4%. Meanwhile injuries fell 17% and property-damage-only crashes fell 25%.

For insurers the shape matters as much as the totals. A crash mix tilted toward the severe end means fewer claims but a larger average claim, especially on the bodily injury side. The same mechanism explains why 2020's claim relief did not carry into later years.

Later NHTSA estimates, all labeled early estimates subject to revision, show the rate drifting back down as driving recovered:

In its April 2026 estimate NHTSA said 2025 deaths were back to the pre-pandemic levels of 2019 and the rate was the second lowest on record, behind 1.08 in 2014, while 2025 travel rose about 0.9%. See Crash statistics and premiums: reading NHTSA FARS and IIHS data for how insurers and analysts read the Fatality Analysis Reporting System data.

How much do gas prices change driving?

When gas gets expensive, it is natural to expect fewer miles and fewer claims. Research suggests the short-run response is small. In a 2008 paper in The Energy Journal, Hughes, Knittel and Sperling estimated the short-run price elasticity of gasoline demand at between -0.21 and -0.34 for 1975 to 1980 and between -0.034 and -0.077 for 2001 to 2006. An elasticity of -0.05 means a 10% price rise cuts gasoline use by about half a percent. The paper also reports an estimate from earlier work by Small and Van Dender that the short-run elasticity of miles driven with respect to fuel price fell from -0.045 (1966 to 2001 average) to -0.022 (1997 to 2001).

Run those numbers on a hypothetical price shock. If pump prices jumped 20%, a miles elasticity of -0.022 implies about 0.4% fewer miles, and -0.045 implies about 0.9%. The 2020 plunge was 11%. Those elasticity estimates describe the early 2000s and earlier, so treat them as a rough guide, not a forecast. What they show is that gasoline cost alone seldom moves total driving much in the short run, while shutdowns, remote work and recessions can.

For pricing, the practical consequence is that the price of gas is not a rating factor. Rates are built from claim counts, claim costs and projected trends, and miles driven shows up only to the extent it affects those. A large, durable change in commuting, such as widespread remote work, would reach rates through the claims data and through mileage classes on policies, not through a gas-price formula.

Why the premium response comes years later

A rate is a forecast, and it is built from old data. The Casualty Actuarial Society's text Basic Ratemaking (fifth edition, 2016) spells out the timeline. Historical losses are trended from the average accident date of the experience period to the average accident date of the period in which the new rates will be in effect. Because policies written over a year run for another year, that second period is centered a full year after the effective date.

The CAS text gives a worked illustration, not market data: losses from accident year 2011, rates effective January 1, 2015, annual policies. The average accident date in the data is July 1, 2011 and the average accident date of the future period is December 31, 2015, so losses are trended 4.5 years. Actual filings differ in how old their data are, but the structure is the same.

Here is how a change in claims makes its way into your bill:

1. Claims happen. An accident year closes.
2. Data matures. Not all claims are reported or settled right away, so insurers estimate final costs.
3. An indication is calculated. Actuaries trend the losses forward and compare projected costs with current rates.
4. A filing goes to the regulator. Some states require approval before use and others review after, as explained in how car insurance rate approval works. See from filing to your bill: the timeline of an auto rate increase for the sequence.
5. The new rate takes effect. It applies to new policies and to renewals as they come up.

That chain is why premiums can keep rising after claim frequency has stopped rising. Triple-I's ISO figures show it: collision claim frequency was 4.16 per 100 car-years in 2024, below 5.00 in 2019, yet average collision claim severity was $5,489 in 2024 against $3,828 in 2019 (about 43% higher). Bodily injury severity went from $19,151 to $28,278 over the same period (about 48%). Comprehensive frequency rose from 2.99 to 3.95. Fewer claims and larger ones can still add up to higher premiums, and the filing lags both. See why car insurance got more expensive, and where prices stand in 2026 for the cost drivers behind it.

A worked example: counting claims per car and per mile

These numbers are invented to show the arithmetic. Say a small insurer has 100,000 car-years in 2019 and its cars average 12,000 miles a year.

• 2019: 100,000 car-years, 5,000 collision claims (5.0 per 100 car-years), 1,200 million miles driven. That is about 4.2 claims per million miles.
• 2020: miles per car fall 11% to 10,680, so total miles are 1,068 million. Collision claims fall 21% to 3,950 (3.95 per 100 car-years). That is about 3.7 claims per million miles.

Claims per car fell 21%, claims per mile fell about 11%, and miles fell 11%. The company sees a lower-cost year, which it can return as refunds or credits. But what rate should it file for 2022? If it simply assumed 2020 would repeat, the rate would be too low once driving rebounded. If it ignored 2020, it would overprice a year in which many customers drove little. The actuary has to choose how much weight an unusual year gets and document that choice in the filing, along with a trend for rising repair and medical costs. There is no single correct answer, which is why different insurers' filings for the same period can look different.

How pay-per-mile pricing responds

Pay-per-mile insurance changes the exposure base from the car-year toward the mile. Nationwide's SmartMiles is a published example: according to the company's SmartMiles page (accessed November 2026), the premium has two parts, a base premium and a per-mile charge, and the monthly payment is the base premium plus the cost per mile times the miles driven in the prior billing cycle. Only the first 250 miles on any single day count toward the charge, and the program is not available in Alaska, Hawaii, Louisiana, Michigan, North Carolina, New York or Oklahoma. See pay-per-mile insurance and how pay-per-mile insurance verifies your mileage for how mileage is measured.

In California, mileage is the state's second mandatory rating factor. The Department of Insurance's regulation, 10 CCR 2632.5, provides that insurers use estimated annual mileage but may use verified actual mileage under rules that include prior notice to the policyholder. See California for the state's rating rules.

A hypothetical comparison, not Nationwide's actual prices: if a policy has a $60 monthly base premium and a 6-cent charge per mile, then 600 miles costs $96 and 1,200 miles costs $132. Against a flat $150 monthly premium, the break-even is (150 - 60) / 0.06 = 1,500 miles. Below that you save; above it you lose. The pay-per-mile break-even calculator lets you run your own numbers.

What drivers can do with this

• Update your mileage when it changes. If you started working from home or retired, tell your insurer. Mileage class is a rating factor, and some insurers verify what you report.
• Ask whether a low-mileage rating or pay-per-mile product is offered where you live. Availability varies by state.
• Do not expect a quick price drop from lower gas prices or a driving dip. Rates follow claims data with a lag of a year or more.
• Check your state's rules. Whether and how an insurer may use mileage, and how rate changes are approved, vary by state. Your state insurance department's website explains the local process.
• Compare quotes after any major life change. A new commute length or a second car in the driveway changes your exposure. See low-mileage discounts: how insurers reward driving less for how discounts work.

Common questions

Did car insurance get cheaper after the pandemic cut driving?

For some customers, briefly. Insurers returned refunds or credits in spring 2020, according to Triple-I and Insurance Journal reports from that time. The savings were temporary because driving recovered and the cost of each claim rose sharply afterward.

If I drive less, do I always pay less?

Not automatically. In many states mileage is a rating factor, but if your policy was written on an estimate, it does not change until you report a new figure or the insurer verifies it. Pay-per-mile policies adjust monthly.

Do high gas prices lower my premium?

No. Gas prices are not in the rating formula. They can affect premiums only indirectly, through changes in total driving and the resulting claims data, and the evidence is that the short-run effect on miles is small.

Why are traffic deaths per mile still a concern if deaths fell in 2025?

The rate in NHTSA's April 2026 estimate, 1.10 per 100 million miles, was still above the 1.08 low recorded in 2014. And fewer deaths do not by themselves lower the cost of property damage and injury claims, which depend on repair prices, medical costs and legal settlement patterns.

Key takeaways

• Insurers price per car-year, with miles entering through mileage classes, so a drop in driving reaches your premium only when it changes claims or the policy's mileage information.
• In 2020 miles fell 11%, claims fell more, and the fatality rate still rose 21% because the crashes that did happen were more severe.
• Short-run gas price changes move driving only slightly, so they are not a lever for premiums.
• Rates are built from accident years that closed long before the effective date, so premiums can rise while claim counts are flat or falling.
• Pay-per-mile pricing is the one structure where less driving lowers your next bill immediately; use a break-even calculation to check whether it suits your miles.