What Actually Moves Your Car Insurance Rate

What affects car insurance rates is a shorter and stranger list than most drivers assume: where the car is garaged, how far it is driven, the make and model's own loss history, the driver's claim and violation record, the coverages and deductibles selected, and — in most states, but not all — a credit-based insurance score.

Two things make that list hard to reason about. The permitted factors are set by state law, so the same driver in the same car is priced off different inputs in Honolulu and in Houston. And several of the factors drivers argue hardest about turn out not to be measured by anybody. What follows separates rating law and loss data from folklore, and names the source for each; it assumes you know how a car insurance policy is assembled.

How does state law decide which rating factors are allowed?

State law sets the list, and California sets it by statute. Insurance Code section 1861.02(a) requires auto rates to be determined by applying, "in decreasing order of importance," the insured's driving safety record, the number of miles driven annually, the number of years of driving experience, and then "those other factors that the commissioner may adopt by regulation and that have a substantial relationship to the risk of loss."

The phrase doing the work is "decreasing order of importance": everything else must be weighted below those three. The regulation filling the fourth bucket, 10 CCR 2632.5, enumerates fifteen optional factors, from type of vehicle and performance capabilities through academic standing and marital status, ending with relative claims frequency and severity, which "shall reflect where the insured vehicle is garaged."

That is a closed list. Most states run the opposite way: the statutes name what is forbidden, and the department reviews whatever class plan an insurer files. The Texas Department of Insurance's page on how costs are calculated names driving record and claims history, location and mileage, age, gender, marital status, occupation, replacement cost, and credit score — a list California would reject in part.

Why does where the car is garaged move the price?

Garaging location is a proxy for how often cars in an area are hit, stolen and repaired, and how expensively. Washington's Office of the Insurance Commissioner, on how insurers set auto premiums, names "where you live and keep your car, your driving record, and how many miles you drive each year."

The strongest evidence that location is a real signal comes from people trying to remove it. The Highway Loss Data Institute's April 2023 collision report standardises vehicle losses by "calendar year; model year; garaging state; number of registered vehicles per square mile (vehicle density); driver age, gender, and marital status; deductible; and risk" — location is on that list because leaving it in would swamp the effect being measured.

How much does the make and model change the loss picture?

Make and model change the loss picture enormously, and not in the direction most people guess. HLDI's collision report for 2020–22 model years puts countrywide collision claim frequency at 6.1 claims per 100 insured vehicle years, average claim severity at $8,739, and overall losses at $532 per insured vehicle year. Individual vehicle series are then indexed against those totals, with 100 as the all-vehicle result: the highest relative overall collision loss was the McLaren 720S convertible at 752, the lowest the Toyota Tundra double cab long-wheelbase pickup at 48.

Split the index into how often a claim happens and how much it costs, and the folklore falls apart:

Rows come from Tables 2, 3 and 4 of HLDI report R-23; a dash means the vehicle missed that measure's top and bottom ten. The 911 Turbo convertible has the lowest collision claim frequency in the report and a severity more than seven times the average; the Mirage G4, among the cheapest new cars of the period, has the highest frequency. What differs between vehicles is which of the two they are bad at, and a rating plan weighs both.

What is a credit-based insurance score, and is it your credit score?

A credit-based insurance score is not a credit score. The NAIC's topic page separates the two by purpose: "Traditional credit scores predict loan repayment, while insurance scores predict the likelihood of an insurance claim."

The NAIC's breakdown of the FICO insurance score gives the weights as payment history 40%, outstanding debt 30%, length of credit history 15%, pursuit of new credit 10% and credit mix 5%, and rules out race, religion, gender, marital status, age, and income or occupation. The topic page reports FICO's estimate that "about 95 percent of auto insurers and 85 percent of homeowners insurers use them in states where the practice is allowed."

Which states restrict credit-based insurance scoring in auto rating?

Fewer states ban it outright than most lists claim, and the rest restrict it in varied ways. The NAIC's state-by-state chart Use of Credit Reports/Scoring in Underwriting (MC-20), in its Summer 2025 model law compilation, marks each entry reviewed in June 2025. Reading the entries rather than a summary of them produces this:

Two corrections follow. Popular lists of "states that ban credit scoring" fold Utah, Nevada and Oregon in with Hawaii, Massachusetts and Michigan; the NAIC's chart calls those three limits on use, not bans. And California, usually named first, gets there structurally: nothing in Proposition 103 says "no credit," it simply never authorised it.

Hawaii's ban has been enforced. In an April 2002 press statement, the Insurance Division announced that seven auto insurers representing nearly 60% of the state's market had agreed to pay fines totalling $115,500 for using prohibited criteria.

What follows a driver between insurers?

Claims and violations follow a driver between insurers through third-party consumer reporting companies, not through any one insurer's memory. The CFPB's 2025 List of Consumer Reporting Companies, current as of January 2025, describes LexisNexis C.L.U.E. — the Comprehensive Loss Underwriting Exchange — as a claims exchange that "collects and reports up to seven years of auto and personal property claims to help inform pricing and underwriting decisions for the insurance industry." The same list names LexisNexis Telematics OnDemand, which collects "driving behavior data for auto insurance pricing."

Coverage selections are the input that lives inside the policy rather than a database. A higher liability limit raises the insurer's ceiling, which is why the three numbers on a liability limit move a premium, and collision and comprehensive turn a liability-only policy into what people mean by full coverage.

Which factors do drivers believe matter that could not be verified?

Car colour is the headline example, and the honest finding is a negative one. Colour appears nowhere in California's list of permitted rating factors, and section 1861.02(a)(4) requires the commissioner to adopt any additional factor by regulation, so an insurer there could not use it. HLDI reports losses by make, series, class and size, not by paint, and no state insurance department page reviewed here names colour. What could not be found is the mirror image: a regulator or loss-data publisher that has tested colour and published a null result. The defensible claim is that colour is absent from the rating lists and the loss data — not that a study disproved it.

Two other beliefs do fail against a source. "A sports car is the most expensive thing to insure" fails on frequency, per the HLDI table. "My credit score is my insurance score" fails on the NAIC's own definition: different model, different target variable, different permitted inputs.

Where this breaks

HLDI's relative results describe vehicles, not drivers. A relative overall loss of 752 does not mean anyone paid 7.5 times the average premium for a McLaren; it means the vehicle's loss experience sat that far above the all-vehicle average once the standardising variables listed above were held constant.

The state law picture has a shelf life. Every NAIC chart entry used here is marked reviewed in June 2025 and the Hawaii enforcement action is from 2002, so the table is a pointer to a state's own code rather than a substitute for reading it. The 95% adoption figure is FICO's own estimate, not a regulator's census.

Calmorg Insure is a comparison and marketing company — not an insurer, agency or broker, and it holds no insurance licences. What any individual policy costs, and why, comes from the licensed provider that issued it and from that state's insurance department.