Insurance Information Institute statistics and how to use them

The Insurance Information Institute, now branded Triple-I and still found at iii.org, is an information and research group paid for by its member insurers. It does not collect most of the auto insurance numbers it publishes. Its fact sheets mostly reprint data from the NAIC, Verisk's ISO, the Highway Loss Data Institute, the Insurance Research Council and the Bureau of Labor Statistics, so the sound way to use a "III number" is to treat the III page as a finding aid, find the original source, and cite both.

That approach also protects you from the few places where the fact sheets are confusing, which are covered below.

What the III is, and who pays for it

Triple-I describes itself on its about page as having "more than 60 insurance company members," ranging from regional carriers to global ones, and says it is affiliated with The Institutes, the insurance education nonprofit. It also says: "We neither lobby nor sell insurance," and that its work is "objective, fact-based information about insurance."

Two things follow from that description. First, the organization is a membership body of insurers, so its funding is tied to the industry it explains. The about page does not itemize a budget; it links an annual report if you want to look. Second, "does not lobby" is a statement about the organization's activities, not about its viewpoint. A group whose members are insurers will tend to choose topics, framings and spokespeople that make sense from the insurer side of the table.

That is not a reason to dismiss the data. It is a reason to know where the III sits compared with other voices:

The Triple-I's role in public disputes is visible in news coverage. When ProPublica and Consumer Reports published their 2017 analysis of auto premiums in minority neighborhoods, the III's chief actuary was the person quoted in rebuttal, saying insurers "do not collect any information regarding the race or ethnicity of the people they sell policies to." That is a representative use of the III: a credible, data-literate voice for the industry position. Our article on ZIP-code pricing and racial disparities: what investigations found covers what those investigations did and did not show, and reading auto insurance studies by who paid for them lays out a general method for reading studies by sponsor.

What the auto fact sheets contain

The main auto page, Facts + Statistics: Auto insurance, is a stack of tables. As viewed in November 2026, it carries:

• Average expenditure by year and by state. The countrywide figure was $1,282 in 2023, up 14.0% from $1,127 in 2022, according to the NAIC. Florida, Louisiana and New York were the highest in 2023.
• A separate AAA figure. The page also quotes AAA's Your Driving Costs study, which put the average full-coverage insurance cost for all vehicles at $2,098 in 2026. The page itself warns that this is not comparable with the NAIC expenditure figures.
• An affordability table. The Insurance Research Council's 2022 study, using 2021 data, put average auto insurance spending at 1.50% of median household income, ranging from 0.96% in Idaho to 2.62% in Louisiana.
• Industry premium and profitability tables. Net premiums written and combined ratios for 2016 through 2025, with 2025 liability premiums of $204.5 billion and a liability combined ratio of 100.2, sourced to NAIC data from S&P Global Market Intelligence.
• Top writers. Direct premiums written for 2025, led by State Farm at $69.16 billion (18.7%) and Progressive at $69.01 billion (18.6%).
• Claim frequency and severity from ISO, a Verisk business, for 2015 through 2024.
• Incurred losses, collision losses by model year (from the Highway Loss Data Institute), and Consumer Price Index rows from the Bureau of Labor Statistics.

Notice what is missing from that list: the III's own measurements. Almost every table ends with a line like "Source: NAIC" or "Source: ISO." That is by design, and it is good practice, because it tells you where to go next.

How to trace a III number to its source

Every fact sheet table has a source line directly under it. Use it as the starting point for a three-step trace.

1. Read the source line and the footnotes. The footnotes carry the definitions that change what the number means, for example whether a table includes state funds, excludes certain states, or measures "per 100 earned car years."
2. Find the original publication. For NAIC expenditure and premium data, that is the NAIC's auto insurance database report. The 2022/2023 edition, adopted in December 2025, states that the insurance data came from statistical agents: the American Association of Insurance Services, ISO, the National Independent Statistical Service, Independent Statistical Service, Massachusetts Commonwealth Automobile Reinsurers and the Maryland Auto Insurance Fund, plus data from California and Texas. Its introduction notes that data in the report "may differ from data contained in reports from previous years" because statistical agents update their information. You can read it on the NAIC's site: 2022/2023 Auto Insurance Database Report. See also What NAIC auto insurance data measures.
3. Check the formula. The NAIC defines average expenditure as the sum of liability, collision and comprehensive written premiums divided by liability written exposures. That is why the figure assumes every insured car has liability and not necessarily the other two coverages. The difference between that and a quoted price is explained in average expenditure vs average premium vs quoted rate.

For uninsured motorist numbers the trace ends at the Insurance Research Council. The III's uninsured motorists page says 15.4% of motorists were uninsured in 2023, citing a 2025 IRC study, and explains that the IRC measures the share using the ratio of claims made by people hurt by uninsured drivers to claims made by people hurt by insured drivers. That is an estimate from claim patterns, not a head count of cars, which is the subject of How uninsured driver rates are estimated: IRC vs state counts.

For price inflation, the trace ends at the Bureau of Labor Statistics; see The CPI for motor vehicle insurance: how price change is measured. For claim frequency and severity, it ends at ISO; see Verisk (ISO) advisory loss costs: the data behind many insurers' rates.

Strengths of the III as a source

• Consolidation. Ten years of NAIC premium data, ISO claim data and BLS price data are laid out on one page with archived tables a click away. Finding the same series in the originals takes much longer.
• Plain explanations. The text defines terms such as claim frequency ("claims per 100 earned car years") and explains that a drop in the combined ratio is an improvement.
• Clear attribution. Tables name their sources, and the NAIC tables carry the NAIC's reprint notice.
• Actuarial literacy. Its staff understand rating, loss development and reserving, which makes its explainers more precise than most press coverage.

Caution points

The following are visible on the auto fact sheet as of November 2026. None is a sign of bad faith; they are the normal frictions of a page that stitches together tables from different publishers and vintages.

Tables of different ages on the same page

The expenditure paragraph says 2023 is "the latest data available," while the premium and profitability tables run through 2025 and the claims tables through 2024. NAIC expenditure data lags because it is built from statistical-agent filings; S&P premium data arrives sooner. Compare like with like by year.

A state table whose label and description disagree

The state ranking table is headed "2023" and lists Louisiana as the most expensive state at $1,494.70. But the lead-in text above it says the state tables cover 2018 to 2022, and the paragraph at the top of the page, which cites the NAIC for 2023, says Florida was highest at $1,865, followed by Louisiana at $1,754 and New York at $1,753. Those cannot all be 2023 numbers from the same series. Our own October 2026 verification notes reached the same conclusion: the page's "2023" state tables do not match the NAIC's 2023 state figures and appear to be older data. If you need a state figure, take it from the NAIC report itself.

Frequencies described two ways

In the claims section the text says "0.80 percent of people with liability insurance had a bodily injury liability claim," while the table footnote says frequency is "claims per 100 earned car years." Those are different things. A car year is 365 days of coverage for one vehicle, and one person can file more than one claim in a year. Use the footnote's definition when you cite it.

Exclusions buried in footnotes

The bodily injury frequency and severity series excludes Massachusetts and most no-fault states, and the property damage series excludes Massachusetts, Michigan and New Jersey. A national "average bodily injury claim" is therefore not a national average of every state's system.

Editions of the same estimate

A number can change between editions of the same study. A 2023 NHTSA report on crash costs, which we cover in What NHTSA counts in the economic cost of motor vehicle crashes, cited the III for an estimated 12.6% of drivers uninsured in 2018. The current III table, built on the IRC's 2025 study, lists 12.1% for 2018. I did not trace why the two differ. Estimates are revised between editions and the tables are built from different study years, so the lesson is to cite the edition and the year.

Reprint limits

NAIC tables on the page carry the line "Further reprint or distribution strictly prohibited without written permission of NAIC." If you republish, link to the page and the original rather than copying the tables.

A worked example: turning a III figure into a sentence

Say a news story states, "Americans pay $1,282 a year for car insurance." Here is how a careful reader tests it.

1. Find the table. The number appears in the "Average Expenditures for Auto Insurance, 2014-2023" table, source NAIC. The year is 2023.
2. Read the definition. It is an average per insured vehicle, built with liability premiums and exposures in the denominator. It is not a typical household bill, which would be per policy with several cars, and it is not a quote for any driver.
3. Compare with another series carefully. AAA's $2,098 for full coverage in 2026 comes from a different method: a driver under 65 with more than six years of experience, no accidents, a suburban or urban address and full coverage. The III page itself says the AAA and NAIC figures are not comparable.
4. Restate it. "The NAIC reported that the average annual expenditure per insured vehicle in the United States was about $1,282 in 2023 (via Triple-I)." That sentence is accurate, dated and tells the reader how to find the original.

Now a second example on frequency, using made-up numbers to show the arithmetic. Suppose an insurer has 10,000 car years of liability exposure in a year. At the national bodily injury frequency the III reports for 2024, 0.80 claims per 100 car years, you would expect about 80 bodily injury claims. At the severity it reports, $28,278, those claims would cost about $2.26 million, or roughly $226 per car year. That product, frequency times severity, is the "pure premium" idea behind rates, and it is why a small shift in either number matters. The NAIC's own formulas give the same relationships: frequency is incurred claims times 100 divided by earned exposures, and severity is incurred losses divided by incurred claims. Those are not III figures from your state or your insurer; they show how the pieces fit. The full logic of rate building is in how car insurance rates are calculated.

How to cite a III statistic

A good citation answers four questions: what number, whose measurement, which year, and where did you read it. A workable pattern is:

The countrywide average auto insurance expenditure was $1,282 in 2023, according to the National Association of Insurance Commissioners, as reported by the Insurance Information Institute (Triple-I), Facts + Statistics: Auto insurance, viewed November 2026.

Practical rules:

• Name the original source first, then the III as the place you saw it.
• Give the data year, not only the year you read the page. The III's pages carry no "last updated" date on the page body as viewed in November 2026, so your viewing date is part of the citation.
• Link both the III page and the original publication where you can.
• Do not mix editions. If you quote the IRC's uninsured rate, quote the whole series from one edition, as our verification notes recommend.
• Say "estimated" for anything the source calls an estimate.

Putting III numbers beside consumer-group views

When the III and a consumer group disagree, the numbers are rarely the problem; the question is. The III tends to ask what the industry's costs, profits and risk-pooling look like in total. Consumer groups tend to ask who pays the most and why. Both can be right about different things.

A short routine for a dispute:

1. Write each side's claim as a measurable sentence. "Premiums are rising because claim costs are rising" is measurable with ISO severity and BLS repair indexes. "Insurers charge good drivers in some neighborhoods more" is measurable with quote data and loss data.
2. Check the unit. Per car year, per policy, per household, per dollar of premium and per claim are different bases.
3. Check the period. An industry that was unprofitable in 2022 and profitable in 2025 can be described truthfully both ways. The III's own table shows the liability combined ratio at 111.3 in 2022 and 100.2 in 2025, and collision and comprehensive at 113.4 and 81.0.
4. Find a neutral third source. A state insurance department or the NAIC often publishes the numbers both camps are arguing over.
5. Read the other side's reply. A consumer group's report will not usually include the insurer rebuttal, and an industry fact sheet will not usually include the advocate's.

If the industry fact sheet and the consumer report cite the same underlying series, as many do, the dispute is about interpretation and your own homework is easy. If they cite different sources, the work is figuring out why.

Common questions

Is the III a government agency or a regulator?

No. It is a membership organization of insurance companies. Insurance in the United States is regulated by the states, and the NAIC is the association of state regulators. The III publishes information; it does not set rules or approve rates.

If the III is industry-funded, can I trust its numbers?

Trust them to the extent you can trace them. A table sourced to the NAIC or ISO can be checked against that publisher. The more a statement is an interpretation, for example why premiums rose or what a proposed law would do, the more it reflects the sponsor's vantage point, and the more you should compare it with a regulator's or consumer group's view.

Why do III numbers differ from the quote I got?

An average expenditure is the average over all insured cars in a state, with the coverage mix people actually buy. A quote is the price for your driver, car, address, limits and deductibles on a given day. They answer different questions. Our the average cost of car insurance, and what the numbers mean guide explains the difference.

Which III page should I use for a state-specific question?

Use the III for national context and use the state insurance department and the NAIC for your state. Rules, minimum limits and averages vary by state; the state car insurance requirements lookup tool and your state regulator's website are better places to start.

Key takeaways

• The III is a membership group of insurers. It says it does not lobby or sell insurance, and its data pages are useful, but its interpretations come from the industry side.
• Most III auto tables are reprints. Follow the source line to the NAIC, ISO, HLDI, IRC or BLS publication and cite both.
• Read footnotes: they define "frequency," list excluded states and show whether a table includes state funds.
• Watch the data year. The same page mixes 2023, 2024 and 2025 series, and a state table's label may not match its description.
• When the III and a consumer group disagree, translate each claim into a measurable statement and compare units, years and sources before deciding who is right.