Price optimization: why regulators barred charging what you'll pay

Regulators in at least 18 jurisdictions said between October 2014 and December 2015 that an insurer may not set one driver's price higher than another's because of a prediction about who will accept it, as opposed to a prediction about who will cost more in claims. Their reasoning was legal, not moral: nearly every state requires rates to be not excessive, inadequate or unfairly discriminatory, and a charge unrelated to expected losses or expenses is, in their view, unfairly discriminatory. They acted through bulletins and notices to insurers, not new statutes, and told insurers to remove the practice from their filings.

Our companion article, price optimization and loyalty pricing in car insurance, covers the NAIC white paper and a shopper's renewal test. This one takes the regulators' side of the story: the legal standard they applied, what each order required, what they did not forbid, and how the same logic reaches newer tools.

The legal standard: price must follow cost

Rate laws differ in wording, but they share a core. The following texts show the common thread.

The logic regulators applied is a simple syllogism. A rating factor is permitted when it measures a difference in expected loss or expense. A customer's tolerance for a price increase, or how likely the customer is to shop, is not a difference in expected loss. A price built on it therefore sorts customers who cost the same into different prices. That is the definition of unfair discrimination in rating, even though no protected characteristic is involved.

California said so in the plainest words. In its notice of February 18, 2015, sent to more than 750 property and casualty insurers, the Department of Insurance defined price optimization as "any method of taking into account an individual's or class's willingness to pay a higher premium" than others, and declared any use of it in ratemaking or rating plans unfairly discriminatory under California law, according to a law-firm account of the notice. That account added that the Commissioner had earlier told the NAIC that "no differentials allowed based on whether the applicant or insured is more or less likely to look elsewhere." Insurers had to take any such factor out of their next filing, due within six months. California's long-standing Proposition 103 standard, which bars rates that are excessive, inadequate or unfairly discriminatory (Ins. Code 1861.05), gave the notice its footing; see California Proposition 103 explained.

A worked example of cost-based versus demand-based pricing

Say two drivers, Dana and Lee, live in the same ZIP code, drive the same car with the same mileage and have clean records. The insurer's cost model predicts $900 a year in claims and expenses for each. The filed rating plan would charge both $1,000.

A demand model looks at other data: how each responded to past price changes and whether either has ever gotten a competing quote. It predicts Dana will renew 95 times in 100 even after an 8% increase, and Lee only 70 times in 100. The insurer charges Dana $1,080 and leaves Lee at $1,000.

Nothing about the expected cost of insuring Dana changed. The $80 difference reflects only how unlikely Dana is to leave. That is the gap regulators called unfair, and it is hard for Dana to see, because a premium notice shows one number. These figures are invented for illustration, not market data.

Why insurers were doing it, and the definitional fight

Vendors sold the practice as a revenue tool. A press release from consumer groups on Maryland's bulletin, which is an advocacy source, quoted the bulletin as citing a vendor's claim that firms adopting advanced pricing strategies considering customer elasticity could boost revenue by roughly 3 percent and returns on equity by about 1 percent on average. That is a vendor's marketing claim, and I could not verify it independently.

The disputes were mostly about definitions, not the core principle. When Florida's Office of Insurance Regulation issued memorandum OIR-15-04M on May 14, 2015, it conceded that price optimization has no universally recognized definition, and it supplied one: a process for modifying the premium that would otherwise be charged, with the goal of maximizing retention, profitability, written premium or market share, using price elasticity of demand. Insurance Journal reported that the Property Casualty Insurers Association of America called the definition vague and very broad and worried it could sweep in beneficial practices, citing "price tempering," which gradually phases in actuarially justified rate increases to limit the effect on policyholders.

Indiana's bulletin handled the same worry differently, by drawing a line instead of a definition. It described price optimization as using data to predict which consumers will accept higher rates without changing insurers, or varying premiums based on factors unrelated to risk of loss so each insured is charged the highest price the market will bear, and it named demand price elasticity modeling as one component. It then said that when a rating factor or methodology produces classification rates outside a reasonable range of cost-based estimates, or unfairly discriminatory rates, the rates are not in compliance.

What each regulator told insurers to do

The orders were practical. They did not fine anyone in the documents I could open; they told insurers to stop and to prove it in their next filings. The Connecticut legislature's research office counted 13 regulators acting through bulletins, notices, memoranda or advisories as of its 2015 report, and found no state statutes on the topic. A Bressler law-firm survey later counted 18 jurisdictions, also all by guidance. The table summarizes the ones I could read in detail.

Minnesota's announcement, as reported by Insurance Journal, came after the Department objected to an auto insurer's rate filing that included price optimization techniques. The Commerce Commissioner told Insurance Journal that charging more for a bad driving record is acceptable but charging more for shopping habits is not, and noted that seniors and low-income consumers are more likely to stay with their insurer. The Bressler survey also reports that Vermont, Maine and Rhode Island required disclosure in filings, and that Connecticut's December 2015 bulletin required removal from personal lines unless shown to be cost-based.

The enforcement picture is thin. The sources I opened show orders to refile, certify and submit corrective plans, but I did not find a published penalty against a named insurer for price optimization. Consumer advocates noted that commissioners can fine or sanction insurers that ignore lawful orders, which is true of any bulletin, but that is a statement of authority, not a record of use. If you need to know whether a specific insurer has been cited, search your state department's enforcement actions; reading a regulatory action: consent orders, fines and what they mean explains how to read one.

What regulators did not forbid

The bans are narrower than the headlines. Actuarial pricing remains the standard, and these things stay legal where state law allows them:

• Risk-based factors. Age, location, vehicle, mileage, driving record and, where permitted, credit-based scores, if filed and supported by loss experience. See rating factors.
• Competition between insurers. One company may charge less than another for the same driver. The rule applies to differences inside a company's own filed rating system.
• Capping and tempering. Phasing in a justified increase to avoid a sudden jump is a recognized practice, and the PCI objection quoted above shows insurers feared a vague definition might catch it. Whether a particular capping rule is acceptable is a question for your state's filing rules; the companion article discusses capping and renewal gaps.
• Filed discounts. A multi-policy or tenure discount that is in the rating plan is a price rule you can see in the filing.
• Company-level business planning. Forecasting how many customers will renew at a given overall rate level is ordinary management, as distinct from adjusting each person's price based on a prediction about that person.

A worked example of tempering

Say a rate filing shows your risk class should pay $1,200 and your current price is $1,200 while the class indication jumps to $1,440, a 20% increase. The insurer's filed rule caps annual change at 10%. Your premium goes to $1,320 in year one, and to $1,440 in year two, since 10% of $1,320 would exceed the indicated amount. That is a rule you could find by reading the filing, and it applies to everyone in the class. Price optimization differs: the cap would be $1,440 for Lee and $1,600 for Dana because of predicted behavior, with no filed rule that explains the difference. These numbers are invented.

Renewal prices now have to be explainable

Regulators did not create a single national disclosure form. Read together, the orders above point to three working expectations (this is our reading, not a rule any one state states in these words).

1. Every price difference should trace to a filed factor. The filing is where a department checks, which is why Minnesota required a certification and Vermont, Maine and Rhode Island required disclosure. See How to read a car insurance rate filing on SERFF for how to find yours.
2. A renewal increase should come from a base rate change, a changed rating factor, or a filed capping or transition rule. Those are the explanations a department can verify.
3. Behavior-based adjustments need support. Indiana's bulletin lists the kinds of data it worried about: reactions to past rate changes, information collected by an agent or in the application, consumer purchasing patterns and socioeconomic status.

State renewal-notice laws add timing and content rules; see reading a car insurance renewal notice: rate-change notice rules. The point for a consumer is that "your rate changed" is not an explanation, and an insurer should be able to say which of the three categories applies.

How the same logic reaches newer tools

Price optimization was an early case of a wider question: what happens when a model finds a pattern that predicts profit but not cost? The NAIC's 2023 bulletin on artificial intelligence says rates developed with machine learning must not be excessive, inadequate or unfairly discriminatory regardless of the methodology, which is the same principle in a new setting. Colorado's statute and regulations test whether data and models produce unfairly discriminatory outcomes for protected classes, a different question from price optimization but built on a similar idea that a price must be justified by something other than how a model sees the customer. See How regulators oversee algorithms and AI in car insurance. For telematics programs, how regulators review usage-based insurance rating plans covers how rating plans are reviewed.

What shoppers should know

1. Compare, don't audit. You cannot see the models. You can see prices. Get a quote for identical coverage from your current insurer as a new customer and from at least three others. Our renewal rate increase checker and why your premium went up at renewal with no claims help you read the result.
2. Ask the three-category question. "Of this increase, how much is a base rate change, how much comes from my own rating factors, and is any of it from a cap or transition rule?" Ask for the answer in writing.
3. Ask for the filed factor. If the insurer cites a factor you do not recognize, ask for its name and where it appears in the filing.
4. Complain with specifics. If an insurer cannot trace a gap to a filed rule, your state's state department of insurance can ask the company to justify it. Include both quotes and dates.
5. Do not assume the ban means equal prices. A renewal that exceeds a new-customer quote can come from lawful causes, and a lower new-customer quote elsewhere is competition, not discrimination.

Common questions

Is price optimization illegal everywhere?

Not by a single national law. The regulators who acted did so under their states' existing rate laws, and the sources I found (from 2015) show guidance, not statutes. A state that issued no bulletin has not said the practice is allowed; it may not have addressed it. Your state department of insurance can tell you its current position.

Does it matter that it was a bulletin and not a law?

It matters for form, less for force. A bulletin states how a regulator will read the statute and what it will do when reviewing filings. A department can disapprove a filing and take action under its existing authority. Courts, not departments, have the last word on the statute, though I did not find a court ruling on price optimization in the sources I opened.

Did every state ban it?

No. New York asked insurers to report their use rather than banning it, according to the Connecticut legislature's research office. Many states do not appear in the surveys at all. That tells you what was documented in 2015, not what the law is in your state in 2026.

How is this different from a loyalty discount?

A loyalty discount lowers your price by a filed rule. Price optimization raises or lowers it based on a prediction about your behavior, with no cost reason. A rule is visible in the filing; a behavioral adjustment is not.

Can an insurer still raise my rate after many years without a claim?

Yes, for filed reasons such as a base rate increase, a change in your rating factors or an end to a discount. Ask which applies, and shop. See reading a car insurance renewal notice: rate-change notice rules.

Key takeaways

• Between October 2014 and December 2015, regulators in at least 18 jurisdictions said pricing on willingness to pay or likelihood to shop is unfairly discriminatory because it is unrelated to expected loss or expense.
• They acted by bulletin under existing rate laws, telling insurers to stop, refile, certify or file corrective plans; I found no published penalty against a named insurer in the sources I opened.
• Risk-based rating, competition, filed discounts and filed capping rules remain legal; the line is between a filed, cost-based rule and a hidden, behavior-based adjustment.
• The same principle, that a price must be justified by cost, now frames regulators' approach to machine-learning models.
• Ask your insurer which part of an increase is a base rate change, a rating factor or a cap, get it in writing, and compare quotes for identical coverage.