If your credit data has characteristics that correlate with high claims risk for an insurer, insurance-scoring models may penalize you. The result may be a higher premium.

Known as credit-based insurance scores, these numbers are computed from bill-paying and loan data collected by the major credit bureaus. In recent years, the scores have become as important in determining your annual premium as your driving record and the neighborhood where you live.

An analysis by Consumer Reports found that scoring could cost many consumers hundreds of extra dollars. Here's a typical example of how scores can hurt: a 28-year-old single male from Orlando, Fla., with a clean driving record and no accidents would normally pay State Farm Mutual $1,251 a year for a new policy. With one at-fault accident, his premium would rise to $1,447. If the same driver instead fell into the lowest ranking in State Farm's credit-scoring system, however, his premium would shoot up to $2,600.

Even a driver with a great credit score whom lenders would normally bless with a low-interest mortgage could wind up with a less favorable insurance score and thus a high premium. That's because formulations for insurance scores weigh credit data differently from traditional lender scores. Indeed, insurance scores can penalize consumers who use credit reasonably. For instance, Progressive Auto Pro's Financial Responsibility Score will give premium-boosting black marks to a customer whose credit-bureau information says he opened three credit accounts within the previous year, including one credit card in the previous four months, and then made two or more additional loan inquiries without accepting the credit.

Such a system may seem bizarre, but insurers contend that there is method to their math. Because there is a statistical correlation between scores and claims, scoring "shifts costs from drivers who file fewer claims to those who file more," says Robert Hartwig, chief economist for the Insurance Information Institute, an industry trade association.

Insurance officials say that most consumers are paying lower premiums more precisely related to risk. "Before, there were only a few rating tiers," says Keith Toney, president of InsurQuote, a division of ChoicePoint, an Alpharetta, Ga., provider of insurance scores. "Credit scoring allows companies to sort customers into hundreds of tiers."

Even though scoring has been in use for more than a decade, it is a mystery to most consumers. About two-thirds of 1,578 consumers surveyed for a report last year by the Government Accountability Office did not know that their credit histories could affect their insurance premiums.

And how could they know? Few insurers routinely disclose scores or what role they play in setting premiums. To fill that void, Consumer Reports sought and obtained scoring models filed with regulators in Florida, Michigan, and Texas used by 9 of the 10 largest U.S. auto insurers. (See Secrets of scoring methods revealed.) Then, with the help of experts, we deciphered the cryptic wording and algorithms that affect the price of insurance.

What we found was a mishmash. Hundreds of insurers use scoring models created by ChoicePoint and Fair Isaac, the Minneapolis company that invented credit scoring. Other insurers developed their own systems. The scoring models emphasize bits of credit data that to the average person would seem to have little to do with a driver's propensity to make claims. There are no standards: Each company uses different models and weighs different credit-report information. Some big companies find scoring useful only for new customers, not renewals, while others may use it for both.

Moreover, the credit data from which the scores are derived have a reputation for being inaccurate and out of date. And several studies have shown that insurance scoring adversely affects blacks, Hispanics, and low-income consumers. Despite such problems, most states allow insurance scoring, and efforts to limit or ban it have been met with aggressive lobbying by insurers.

SCORING: FACTS AND FICTION

Insurers have long used statistics to determine premiums. That's how they figured out that drivers under age 25 have more accidents than older drivers. The traditional rating factors have been age, sex, marital status, ZIP code, driving record, and three-year history of at-fault accidents. Insurers determine how much each factor affects the frequency and size of payouts and create a formula for calculating a premium based on your characteristics. The formula starts with a dollar base rate for each type of coverage, then multiplies, adds, or subtracts amounts based on each of the rating factors.

In the 1990s, Fair Isaac worked with several insurers to test its theory that credit scores might predict homeowners- and auto-insurance claims losses. Statistical analysis of archived data from more than a million credit files found that 30 of 100 or so items in the reports correlated with payouts. That finding led to the creation of homeowners- and auto-insurance scores.

A study conducted in 2000 by James Monaghan, a research strategist at Metropolitan Property and Casualty Insurance Company, found, for example, that people whose oldest account on their credit report dated back 25 to 29 years subsequently filed only $60 worth of claims for every $100 of premiums paid over the next three years. But people whose oldest account was only a year old filed $95 in claims per $100 of premiums over the ensuing three years.

Neither insurers nor the credit-scoring companies that discovered the relationship know what causes it, except to suggest that those with subpar credit are themselves subpar. "People with a pattern of irresponsible financial behavior and poor credit history have a much greater chance of being in an accident or filing a claim," says Joseph Annotti, a spokesman for the Property Casualty Insurers Association of America, a trade group. The American Academy of Actuaries said in its 2004 recommendations to the Federal Trade Commission that "aggressiveness" and "willingness to take risks" go along with a poor driving record. "The correlation with fraud is striking," says Gordon Stewart, president of the Insurance Information Institute. But the Monaghan study, which reviewed those long-standing inferences, says that links between responsible financial management and future expected losses are "unsupported."

Steven Parton, general counsel for the Florida Office of Insurance Regulation, says, "What they're really looking to see with insurance scores is who is most likely to file a claim, not who will most likely have an accident. If I have the money, I won't file a claim, because my rates will go up. People of low economic status don't have that luxury." Parton adds, "Insurance companies are looking at whether they're relying on their insurance in case they have an accident, which is what they're buying insurance for to begin with."

PREMIUMS ALL OVER THE PLACE

How many benefit as a result of scoring and how many lose is a matter of dispute. Lamont Boyd, insurance market manager at Fair Isaac, says, "Two-thirds to three-quarters of customers are getting better premiums because of credit-based insurance scores." A study by the Texas Department of Insurance in 2004 found that half paid more and half paid less than they would have without scoring. Among insurers we interviewed, only the Farmers Insurance Group told us how scoring affected premiums for its customers: an $80 annual reduction, on average, for the 58 percent who saved and a $109 increase for the rest.

To see how insurance scores affect premiums, an actuary worked with Consumer Reports to calculate premiums charged by preferred/standard-risk companies run by eight of the largest U.S. insurers operating in Florida for a 28-year-old single man with a clean driving record in Orlando, Fla., who owns a 2005 Toyota Camry LE. First, the actuary calculated the premium the driver would pay if no credit score was used—considered a "neutral" score. Then he recalculated the premium using each insurer's best and worst possible insurance score. Finally, each insurer verified our calculations.

The results were erratic. (See Differences among insurer premiums.) For example, if insurance scores were neutral, our hypothetical customer would pay roughly the same annual premium at Nationwide and GEICO, about $1,150. But if the driver received the worst possible insurance scores, the premium would increase 29 percent to $1,468 at GEICO and 47 percent to $1,706 at Nationwide. At Birmingham Fire, scoring from best to worst increased the premium by $3,166.

Such variations raise questions about whether scoring closely customizes price to each driver's loss risk, as insurers contend. How precise can scoring be when our hypothetical customer with the best score gets a 31 percent discount on his annual premium at Progressive but only a 19 percent discount at Birmingham Fire? Or when USAA charges our hypothetical customer with the worst score 32 percent extra, why is State Farm charging him 108 percent more?

Leslie Kolleda, a spokeswoman for Progressive, says that the variation in pricing has "nothing whatsoever to do with credit scores" and is typical of insurer-to-insurer price differences. Progressive's comparison of 90,000 multicompany premium quotes in 2002 showed a $524 difference, on average, on six-month policies.

But some state regulators still have doubts. "You can't say this is the best predictor we have, but at the same time we all do it completely differently," says Joel Ario, the insurance administrator of Oregon. "Either there's a core to this or it's a farce."

AT-FAULT CREDIT BEHAVIOR

Insurers say, as Allstate does on its Web site, that a consumer's "financial difficulties" may indicate a tendency toward greater risk-taking behavior. But insurance scoring also punishes people in no particular difficulty. For example, only 40 percent of Fair Isaac's Assist insurance score is based on payment history, says Boyd. The other 60 percent weighs balances and credit limits, the age of your earliest account, whether you shopped for loans, and the types of loans you have.

A version of Assist used by the Farmers Insurance Group in Michigan until last winter rated customers on 11 factors. With a nearly perfect score of 827, a 28-year-old single male from Ann Arbor would have landed in the top tier and paid $1,027 a year in premiums if he engaged in optimal credit-use behavior as defined by Assist: Having exactly two major bank credit cards, not shopping for a loan in the previous year, and paying bills on time, among other things.

But if he shopped for loans, opened three new credit-card accounts, and temporarily ran up balances—all no-no's in Assist's scoring system—his insurance score would have plunged like the Dow Jones Industrial Average on a bad day—207 points. That would have kicked him three tiers down Farmers credit rating factor table and raised his premium $664.

This year, however, partly because Farmers found Fair Isaac's model difficult to explain to consumers, Farmers switched to an in-house scoring model called Fire & Auto Combined Evaluation Tool, or FACET. Surprisingly, the credit characteristics that cost our model customer a bundle last year may not raise his premium a dime this year.

Are Farmers customers from last year entitled to a rebate? No, says Bill Martin, a Farmers vice president: "It's not that their rate or score was wrong. It's just that we've changed how we do it. Two actuaries can look at the same set of data and make different decisions about risk." CR's examination of credit models used by large U.S. auto insurers in Florida, Michigan, and Texas found numerous inconsistencies. (See Secrets of scoring methods revealed.)

Progressive's A24 credit-scoring model looks at 12 items on credit records and will bite you for opening one new credit card in the previous four months or having a credit-card balance higher than 40 percent of your limit, neither actions indicative of grave financial problems.

The ChoicePoint Attract scoring model, with dozens of separate factors, has the most complicated formula of any we analyzed. It dings you for having department-store charge cards and auto loans from automaker finance companies such as GMAC, for not having an oil company credit card, and for so much as touching finance-company credit, which may include incentive installment loans from appliance, electronics, or jewelry stores.

Why does ChoicePoint deem such activities "bad?" The company says it doesn't know why, just that there is a correlation. "It's certainly statistically true that people who have finance-company accounts have higher losses," says John Wilson, an assistant vice president who works on credit-scoring-related analytics.

Your insurance score may also take a hit if you frequently apply for loans to see what rate you qualify for. Progressive's model concludes that you are desperately seeking and being rejected for credit and lowers your score. Monaghan's research shows that the average number of inquiries on credit reports has doubled from a decade ago.

CRUNCHING IMPERFECT DATA

The reliability of insurance scores in predicting claims depends on the reliability of the data on which they are based, and the verdicts are mixed. Studies in recent years by the Consumer Federation of America, the Federal Reserve, U.S. Public Interest Research Group, and Consumer Reports have together found significant percentages of credit reports with erroneous delinquencies, out-of-date balances, and incorrect credit limits, all of which could lower your insurance score. Steven Katz, a spokesman for TransUnion, one of the three major credit bureaus, says, "We believe that credit reports are highly accurate." A study by the GAO in 2003, however, declared that it didn't have enough data to draw any conclusions.

In any case, a credit report is a snapshot of a moment in time. "Our models can only rely on the data existing in the credit report as of the day it's accessed," Fair Isaac's Boyd says. Lenders don't update balances with credit bureaus every day. A reporting time lag can show a balance weeks to months after you've paid it, according to a 2004 report by the Federal Reserve. If you've run up your balance on a vacation, for example, an insurance-scoring model could downgrade you for using too large a percentage of your credit line even if you always pay your balance in full. "The model doesn't know that you're planning to pay that bill off," Boyd says.

AN ADVERSE EFFECT

Some factors in credit-scoring models seem to target low-income consumers. For example, 18 percent of lower-income families use finance-company credit, according to Federal Reserve research conducted in 1995. The ChoicePoint model penalizes such credit. Similarly, 67 percent of all families have a major credit card, while only 45 percent of lower-income families do. ChoicePoint may downrate those who don't carry major plastic. "Credit scoring is a proxy for race and income, and it is a tool to allow insurers to get around redlining restrictions," says Birny Birnbaum, executive director of the Center for Economic Justice, an Austin, Texas, advocacy group for low-income and minority consumers.

Studies by officials in Missouri, Texas, and Washington over the past few years have shown that insurance scores have a disproportionately adverse effect on blacks, Hispanics, and low-income families. A 2004 Texas Department of Insurance study found that, depending on the credit model, members of minority groups have credit scores that on average are lower than those of whites, by 10 percent to 35 percent for blacks and 5 percent to 25 percent for Hispanics.

But Texas concluded it had no legal authority to prohibit scoring, because a follow-up study presented in 2005 confirmed the strong relationship between scores and claims. "Credit scoring, if continued, is not unfairly discriminatory as defined in current law because credit scoring is not based on race," wrote Jose Montemayor, then insurance commissioner of Texas, in a letter accompanying the report. In June 2006, in a settlement preliminarily approved by a Texas judge, Allstate agreed to change its credit-scoring model nationwide in response to a class-action lawsuit alleging that the company's insurance-scoring model discriminated against blacks and Hispanics. In court papers, Allstate denies that it has discriminated against the plaintiffs or violated the law. Allstate did not respond to our requests for comment.

Some state insurance officials question scoring fairness. Joel Ario, Oregon's insurance administrator, asks, "Should disproportionate impact be enough to mean it should be limited or banned? In the case of race, we say yes." So in Oregon, insurers can use scoring only with new customers, not existing ones who have a track record.

In May, Florida put insurers on notice that by Dec. 1 they will have to show that their use of credit scores does not disproportionately affect consumers because of race, color, religion, marital status, age, gender, income, national origin, or place of residence.

Insurers vigorously argue that scores are not discriminatory. "We don't ask for customer ethnicity or income level, so we're blind to all that," says Dick Luedke, a State Farm spokesman.

An FTC study may shed new light on the issue. "We plan to dig a little deeper into the discrimination issue," says Jesse Leary, an assistant director in the FTC's bureau of economics. "We're going to be looking at how well scores predict risk within racial categories. Are scores a proxy for race or ethnicity?"

SOME STATES PUSH BACK

Insurers have no legal obligation to disclose their scores, and few do. Progressive and Farmers give customers details about their score on request. And ChoicePoint sells insurance scores to a consumer for about $13, but it won't reveal which insurers might use them.

Consumers are largely left to protect themselves. (See How to get a lower premium.) Most states allow the use of scoring, and insurance lobbyists have successfully repelled numerous efforts to ban or restrict it. Twenty-two states have passed model legislation endorsed by the insurance industry, which provides only weak consumer protections. Chiefly, the laws prohibit scoring models from counting delinquent medical accounts. Maryland, New Mexico, Oregon, and Washington have added stronger restrictions, while California, Hawaii, and Massachusetts ban scoring for all insurance. (Maryland has banned it for homeowners policies.) But this year, insurance lobbyists helped to squelch legislation to end scoring in Colorado, Delaware, and Minnesota.

Despite that, some insurers are having second thoughts. David Snowden, a spokesman for USAA, says, "After two years, our experience with the policyholder proves to be more valuable than the insurance score." And Bill Martin of Farmers says that if other insurers drop scoring, his company might also. "If we can't explain what we're doing, we shouldn't be doing it," he says.


What affects your score

If your credit data has characteristics that correlate with high claims risk for an insurer, insurance-scoring models may penalize you. The result may be a higher premium. But many items meticulously weighed by one scoring model are vastly different or not even considered in another. Below is a sampling of three scoring models we examined and what credit-use characteristics they penalize consumers for.

Characteristics consumers are penalized for
Scoring model Consumer-initiated credit inquires: Number of new accounts opened: Age of applicant when first loan was taken out: Number of major credit cards:
Fair Isaac 1 or more in the last 12 months Depends on customer Not considered Any number but two
ChoicePoint 1 or more in the last 6 months 1 or more in the last 24 months Not considered None
Progressive 5 or more in the last 24 months 1 bank card in the last 4 months, 3 or more accounts in the last 12 months 26 years and up Not considered


Prying the lid off insurers' secret scoring systems

Four years ago, Consumer Reports asked insurers to what degree credit data figured in scoring formulas used to set car-insurance premiums. We were politely rebuffed. "Our insurance-scoring model is proprietary," said Michael Trevino, an Allstate spokesman.

Last year, we took a different tack and asked insurance regulators in Florida, Michigan, and Texas to give us the rate-setting manuals that contain the scoring formulas. Little did we realize that a request for such records would set off a legal skirmish in Florida.

Later that year, AIG, Liberty Mutual, Nationwide, and State Farm sought injunctions against Florida and Consumers Union, our publisher, to block our access to their credit-scoring models, which they said were trade secrets. Indeed, Liberty Mutual said that keeping its model secret was "in the public interest." Insurers learned of our request from the Florida Office of Insurance Regulation, which gives companies claiming confidentiality an opportunity to object before release.

Insurers haven't exactly been forthcoming about scoring. When its use became prevalent in the mid-1990s, they "did a horribly poor job" telling customers, says Joseph Annotti, a spokesman for the Property Casualty Insurers Association of America, a trade group. Many states have since forced insurers to tell consumers that their credit information may be used.

While the Florida court actions proceeded, Michigan and Texas sent us copies of scoring models filed there, and three of the Florida insurers settled with us and handed over documents. Nationwide had not settled by press time but verified our calculations nonetheless.

A mystery remains. Why would insurers fight for secrecy in Florida when the models are public in other states?

In an affidavit filed with one of the insurers' complaints, Fair Isaac claimed that details of its algorithms "have never been disclosed to any end-user clients in the insurance industry," because insurers would steal them "without having to pay Fair Isaac." Also, if competitors such as ChoicePoint saw the models, Fair Isaac contended, its $140 million-per-year scoring business "would suffer significant competitive injury." But Farmers Insurance had the model, and Fair Isaac later conceded in an interview that ChoicePoint has had access and years to study its models.

The people most affected by insurance-scoring models, consumers, may be the last to know, but at least now the secret is out.


Polish your score and get a lower premium

Insurance credit scoring might raise your premiums, but Consumer Reports' analysis shows that you can take steps to protect yourself when you apply for a car-insurance policy.

Shop harder than ever before. Our rate comparisons in Florida showed a wide range of premiums that eight major insurers charged the same customer. You have no way of knowing whether your score will give you a lower or higher premium, because each insurer calculates scores differently. Only by getting quotes from several insurers can you be sure to find a low rate.

Monitor your credit reports. You should make sure that they are accurate by taking advantage of your right to a free report from each of the credit bureaus once a year. To get the information, you can go to www.annualcreditreport.com.

Avoid certain types of credit. These include department-store credit cards; credit provided by stores to help move big-ticket items; credit accounts at your local tire dealer, auto parts store, or service station; and finance-company credit. And watch out for retailer credit cards that are issued by finance companies, not banks.

Use credit that insurers favor. Scoring models prefer oil-company credit cards. They also like national bank credit cards such as American Express, Discover, MasterCard, and Visa.

Ask about your score. Farmers and Progressive both give details but only if you ask. Information can help you complain if you think you've been incorrectly scored.

Ask for exceptions. Credit reports can be ruined by circumstances beyond your control. Progressive says that it may rescore you if your score has been adversely affected by divorce, Hurricanes Katrina or Rita, job loss, the death of a family member, or serious medical problems.

Demand a recount. If your score is low, ask to be rescored as frequently as your state's law allows, which is usually once a year.

Keep credit-card balances in check. Maintaining any balance can count against you, but the higher the balance, the more points you lose on your score. Avoid letting balances ride, especially on department-store charge cards with a low credit limit. Pay your balances in full each month, and consider making payments every week, perhaps over the Internet, to keep the balance from being reported as high relative to your total credit limit.

Try not to add to your credit. Scoring systems look askance at frequent new credit and penalize you for the number of inquiries related to the shopping and application process. Installment loans will also boost total balances relative to your total credit lines and lower your score.

Pay your bills on time. This is another money-management given that will prevent score penalties. Consider automatic bill payment from your checking account so that you don't have to worry about payments getting lost or delayed in the mail.


Good score, bad score

Depending on each insurer's method of calculating insurance scores, the approximate annual premium for a hypothetical 28-year-old male in Orlando, Fla., who drives a new Toyota Camry can vary from $782 to $4,755.

Annual Premium
Insurer Without insurance score With the best score Discount With the worst score Surcharge
Birmingham Fire (AIG) $1,960 $1,589 19% $4,755 143%
Allstate Property & Casualty 1,098 790 28% 1,490 36%
GEICO 1,142 947 17% 1,468 29%
Liberty Mutual 1,687 1,410 16% 2,361 40%
Nationwide General 1,159 943 19% 1,706 47%
Progressive Auto Pro 1,127 782 31% 1,868 66%
State Farm Mutual 1,251 1,109 11% 2,600* 108%
USAA 1,015 1,015 >0% 1,336 32%
Sources: Auto insurers.

* State Farm would place a customer with the worst score into State Farm Fire & Casualty, which generally has higher rates than State Farm Mutual.

The full Ratings and recommendations for more than 200 vehicles, along with the latest information on thousands of other products and services, are available to ConsumerReports.org subscribers. Find out how to subscribe today.