Verisk (ISO) advisory loss costs: the data behind many insurers' rates

Competing insurers' rates often move together because many of them start from the same data. Insurance Services Office (ISO), which is part of Verisk, collects premium and claims records from insurers and uses them to publish advisory prospective loss costs: an estimate of what claims will cost per unit of exposure, before any insurer adds its own expenses and profit. ISO's number is not a price and no insurer has to use it. Each insurer converts it into a final rate with its own multiplier, which is why two insurers can start from the same loss cost and quote you very different amounts.

Two jobs: statistical agent and advisory organization

ISO does two related jobs, and it helps to keep them apart.

A statistical agent collects data. Insurers send it detailed premium and loss records under a statistical plan, and it checks them, compiles them and reports them to state regulators. New Hampshire's law gives a sense of the standard menu. Under RSA 412:24, a statistical agent may develop statistical plans (including territory and class definitions), collect statistical data from members and subscribers, and distribute information that must be filed with the commissioner.

An advisory organization does analysis. RSA 412:23 lists what such an organization may do in that state: develop statistical plans, collect data, prepare, file and distribute prospective loss costs, and prepare factors for classification, territory and increased limits. It may also distribute rating manuals and rules, as long as they do not include final rates, expense or profit provisions, or minimum premiums. That last limit matters, because it marks where the shared work stops and each insurer's own pricing starts.

ISO is not alone. The NAIC's auto insurance database report, in its edition covering 2019 through 2023 data, says its insurance data came from several 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. A 2020 Insurance Journal report described ISO as a statistical agent in 54 jurisdictions and reported that the Texas Department of Insurance had chosen it as statistical agent for the state's catastrophe event statistical plan, adopted by Commissioner's Order No. 6467 in September 2020.

The scale of the database is large. In a February 2022 press release, Verisk said its database held close to 30 billion premium and loss records and that 47 more property and casualty insurers had committed to contribute data in 2022. It said contributors use the data for benchmarking, ratemaking, product development and strategic planning. That is the company's description of its own product; treat the "world's largest" style claims in its marketing as marketing.

Verisk also runs ISO ClaimSearch, a claims database used to spot suspicious claims. That is a different product with a different purpose; ISO ClaimSearch and how fraud analytics flag claims covers it. This article is only about pricing.

From claims to a loss cost, step by step

The NAIC's Loss Cost Memorandum, a guidance document written for advisory organizations and insurers filing with state departments (the copy I used is the 2022 revision), describes the process in general terms. For auto, the chain looks like this:

1. Insurers report. Participating insurers send premiums, exposures and losses to the statistical agent, organized by coverage, territory and class.
2. The data are developed. The advisory organization takes past loss experience and projects it to an ultimate value over time, allowing for known changes in claim values and for claims incurred but not yet reported.
3. Trend is applied. The organization trends the losses forward to the average date of loss for the period when the new rates will be in force. That is how this year's repair costs and medical bills end up in next year's rate.
4. A reference filing goes to the state. The advisory organization files its proposed loss costs, with supporting actuarial and statistical data, in what the memo calls a loss cost reference filing, for the department of insurance to approve or acknowledge.
5. Insurers are notified. After approval, the organization tells its participating insurers, normally distributing manuals of loss costs, rules and other supplementary rating information.

Verisk's own brochure on the product (2019) says ISO's loss costs project average future claim costs and loss adjustment expenses, can be viewed by coverage, class and territory, and can be broken down by state, policy limit and deductible. It adds that because the loss costs are submitted to state regulators for review, an insurer's actuarial and compliance staff can concentrate on other decisions.

What a loss cost is, and what it leaves out

Under the NAIC memo, a loss cost is the expected claims cost per exposure unit, exclusive of expenses and profit. The expenses it leaves out include commissions and brokerage, other acquisition and field expenses, general expenses, and taxes, licenses and fees. Profit means underwriting profit, including a provision for contingencies and expected investment income. In most states and lines the loss cost already includes loss adjustment expense, the cost of investigating and settling claims. When it does not, the insurer must add it.

An exposure unit in auto insurance is typically one car insured for one year, which the NAIC calls a car-year. A loss cost of, say, $400 for a coverage, territory and driver class means roughly "$400 of claims per car-year expected there," and nothing about what the insurer must collect to stay in business. The link to the loss ratio is direct: if expenses and profit take 30% of the final rate, the loss cost is the other 70% of it.

From loss cost to the rate you see

The insurer converts a loss cost into a rate using a loss cost multiplier, or LCM. The NAIC memo says the LCM is generally the product of two factors:

• The loss cost modification factor adjusts for the insurer's own quality of business and past experience. If the insurer expects results 10% better than the advisory number, the factor is 0.900. If it needs no adjustment, it is 1.000.
• The expense multiplier covers expenses and profit. The memo's example: if expenses and profit total 33.3% of premium, the multiplier is 1 divided by (1 minus 0.333), or 1.500.

Multiply the two and you have the LCM. The memo's own example gives 0.900 times 1.500, or 1.350.

A worked example with invented numbers

Say the advisory loss cost for one coverage, territory and class is $400 per car-year. Three insurers each choose their own factors:

All three started from the same $400. Z's rate is about 41% above Y's because Z expects a worse book and carries higher expenses and profit. Now suppose the advisory loss cost rises 10% to $440, and all three have chosen to adopt the new loss costs automatically. The rates become roughly $629, $528 and $745, so each rises 10% while the gap between them stays the same. This is the mechanism behind "everyone's rates went up at once."

These are made-up figures, chosen to make the arithmetic visible. They are not market data.

Automatic adoption

The NAIC memo says that in many states an insurer may ask to keep its LCM on file and automatically adopt each new reference filing. When the state approves a new reference filing, the insurer can do nothing, in which case its rates become the new loss costs times the LCM already on file, usually without filing new rate pages. Or it can file to delay, decline or change its multiplier. The advisory organization files the loss costs, relativities, rules and most supplementary rating information. The insurer files its effective date, its LCM, its expense constant and minimum premium, whether it adopts automatically, and any exceptions or deviations.

Why competitors' rates move in similar directions

Several forces push the same way, and the advisory system is only one of them.

• Shared costs. Every insurer pays for the same body-shop labor, parts, used-car replacements and medical bills. When those rise, every insurer's claims data show it. The why car repairs cost so much more than they used to article covers the repair side.
• Shared data. Insurers that use the same pooled experience will see the same trends at about the same time.
• Automatic adoption. An insurer that has chosen to follow the advisory loss costs moves when they do.
• Market cycles. When losses run high, insurers collectively raise prices and tighten underwriting; hard and soft markets: why car insurance prices move in cycles explains the cycle.

Forces that push the other way are why the same driver gets spread-out quotes. Each insurer picks its own modification factor, expense load, class plan and tiers. The why two insurers quote the same driver very different prices article goes through them. Some insurers do not use the advisory system at all. In general, an insurer with a very large book of its own data can price from that, and a small one may lean on pooled data; I could not find a published count of which insurers use ISO loss costs for personal auto, so do not assume any particular company does.

Oversight and antitrust limits

Rates and rating organizations are regulated state by state. Under the federal McCarran-Ferguson Act, the business of insurance is subject to state law relating to its regulation, and 15 U.S.C. 1012(b) provides that the Sherman Act, Clayton Act and Federal Trade Commission Act apply to the business of insurance only to the extent that state law does not regulate it. Section 1013(b) separately says the Sherman Act still applies to any agreement to boycott, coerce or intimidate. The result is a limited exemption, not a blank check. The longer story is in McCarran-Ferguson antitrust exemption: what insurers may do jointly.

In practice the line the states draw is the one in the New Hampshire statute above: pooling historical data and publishing trended loss costs is permitted, and the shared materials stop short of final rates, expense and profit provisions and minimum premiums. The NAIC memo adds that in the early 1990s most advisory organizations stopped promulgating advisory final rates and moved to loss costs, so each insurer must use its own decision-making to set the final rate it files. States set their own rules for advisory organizations and what they may publish. A few have systems of their own entirely, as with North Carolina's rate bureau and Reinsurance Facility: a system apart.

How smaller insurers use the data

A small insurer has a credibility problem. With a few thousand policies, its own claims in one territory and class can swing wildly from year to year on a handful of large losses. Pooled data from thousands of insurers smooth that noise. Verisk's brochure presents the loss costs as a benchmark that an insurer can combine with its own data to calculate the rates it needs for its profit targets and for reinsurance and reserving decisions.

A typical small or regional insurer might therefore take the advisory loss cost as its starting point, apply a modification factor from its own experience and an expense multiplier from its own costs, and file the LCM. A large national insurer has enough of its own experience to build its own loss models, and may use the advisory data mostly as a cross-check.

Seeing it in a rate filing

Rate filings are public in many states, and you can look for the advisory system in them. A step-by-step approach:

1. Find the filing. Use your state department's filing search or SERFF; How to read a car insurance rate filing on SERFF shows how.
2. Search the text. Look for "loss cost," "loss cost multiplier," "LCM," "reference filing," "advisory organization" or "ISO."
3. Check the filing type. A filing that adopts the advisory loss costs usually shows a multiplier and an effective date, not a full rebuild of the insurer's rates. A filing built on the insurer's own data usually shows base rates and relativities with actuarial exhibits instead.
4. Read the multiplier's parts. The NAIC's "Loss Cost Filing Document" form breaks an LCM into its modification and expense parts; the memo says this form or a similar state-specific one should be included in the insurer's filing.
5. Compare to the state's average. An LCM well above or below its peers means the insurer is loading expenses and profit differently, or expects a different quality of business.

A filing's headline figure is an average change across all policyholders, not the change you will see. loss ratios and rate filings: how insurers justify an increase covers how insurers justify an increase.

Common questions

Does ISO set my car insurance rate?

No. ISO publishes advisory loss costs and rating information. Your insurer decides whether to use them, applies its own multiplier, and files its own rates with the state. The price you see reflects your insurer's choices as much as ISO's data.

Why do insurers all raise rates at about the same time?

They face the same repair, medical and vehicle costs, many rely on similar pooled data, and some automatically adopt updated loss costs. Each also decides its own timing and size, which is why the increase varies from insurer to insurer.

Is it legal for competitors to share data this way?

State law allows advisory organizations and statistical agents to collect data and publish loss costs under regulator oversight, as the New Hampshire statute shows. Federal antitrust law still applies to the extent state law does not regulate the conduct, and to boycotts. Specific questions are for a licensed attorney.

Can I see the ISO loss costs for my state?

The loss costs themselves are filed with state regulators, and in many states the filing is public. The place to start is your state department's filing search.

Key takeaways

• ISO, part of Verisk, collects insurers' premium and loss data and publishes advisory prospective loss costs: expected claims cost per exposure unit, before expenses and profit.
• Insurers turn loss costs into rates with a loss cost multiplier, which combines a modification for their own results with an expense and profit multiplier. The NAIC's own example is 0.900 times 1.500, or 1.350.
• Shared costs, pooled data and automatic adoption explain why rates move together; different multipliers explain why prices still differ by large amounts.
• Federal law applies antitrust rules to insurance only to the extent states do not regulate it, and states allow loss-cost publishing but not shared final rates.
• To see the system in a filing, search for "loss cost multiplier" or "reference filing," and remember that insurers using their own data will not show one.