Credit Check.

Insurers planning to use credit histories in personal-lines pricing must think before deciding how to collect, analyze and use the data.

Studies show that a person's credit history can predict insurance losses affecting that person's private-passenger automobile and homeowners policies. Insurers that incorporate credit into rating plans for these lines can achieve more accurate pricing. Progressive Corp., Allstate Group, MetLife Auto & Home and Nationwide Group already use credit data to price auto insurance.

When insurers investigate using credit in rating plans, they must consider data collection, data analysis and operational issues. Each project phase affects what occurs in the others. Insurers also must realize that credit frequently will be re-evaluated after its introduction.

Collect All Necessary Data

An insurer studies the relationship between credit and losses in a retrospective analysis. This requires segmenting policyholders into groups with similar credit histories and studying the relationship between these policyholder groups and their subsequent loss experience.

Before collecting any data, it is important to understand the data needs for the entire project, including those for filing with regulators and those for implementing the credit model. Building an analysis file will be the most complex and time-consuming portion of the credit project. Time invested at this stage, however, will pay big dividends later. Company insurance data and external credit data both are required.

Two types of insurance data are essential. The first is the data the insurer uses to search for insureds' historical credit reports. Policyholders' names and addresses are required; birth dates and previous addresses also can be helpful. A person's Social Security number enables additional policies to be matched with credit reports.

Insurers also need individual policy experience data--including premium, exposure and claims data--to evaluate the relationship between credit history and loss experience.

Should a commercially available credit score or a custom method be used to segment the risks? Several commercial scores are available. Developing a custom method requires an insurer to use information directly from credit reports and requires substantial in-house statistical expertise.

Insurers can purchase archived credit information from the major credit-reporting agencies or from data vendors that provide credit reports. Data cost should be considered as it applies to the retrospective study and to the use of the credit model in future rating of policies. By predominantly using one vendor, insurers may earn volume discounts.

Organizing Insurance Data

Insurers can organize premiums and exposures by either calendar year or policy year. Claim data is organized by calendar year, accident year or policy year. Each organization method has advantages and disadvantages.

The selection of an organization method depends on the line of business. For example, loss-development considerations weigh more heavily for personal auto than homeowners. Auto losses frequently take longer to be paid, because they are often third-party liability claims. Also, the availability of internal data may determine the organization method selected.

Complying With Federal Laws

The federal Fair Credit Reporting Act is designed to protect the privacy of consumer report information and to guarantee that the information supplied by credit-reporting agencies is as accurate as possible, according to the Federal Trade Commission's October 1998 publication: Consumer Reports: What Insurers Need to Know.

To ensure that federal requirements are not violated, credit-reporting organizations encrypt or omit information that identifies specific individuals when they return reports to insurers for retrospective studies. Since policy numbers are removed from the data, the insurer must collect all the necessary information up front. Additional data cannot be appended at a later date using policy number, name, address or Social Security number.

Insurers must follow federal requirements when using credit history in rating. In particular, the act contains duties of users taking adverse action--such as denying, decreasing or canceling coverage, or increasing price--based on information in a credit report.

State laws and regulations regarding credit vary widely and evolve constantly. Insurers must research the requirements in the various states in which credit will be used.

How to Analyze Credit Data

The analysis stage begins once the data is collected, balanced and prepared. Many analysis decisions depend upon how credit will be used in future rating of policies.

Each common loss statistic offers different benefits. To the degree that factors such as territory relativities and driver-class deductibles are accurate, the loss ratio adjusts for rating variables that already are incorporated in the rating plan. To the extent that changes have been made in the rating plan, however, the insurer must bring premium to the current rate level before completing a loss-ratio analysis. This requires the collection of historical rating characteristics and factors.

Using pure premium avoids the problems associated with bringing premium to the current rate level. The insurer must make adjustments, however, for the impact of the existing rating variables. It may be beneficial to analyze both the frequency and the severity components of pure premium. Analyzing claim frequency will avoid distortions from unusually large claims. If insurers rely solely upon frequency, however, they may miss relationships that exist between credit and claim size.

Creating a Custom Credit Model

To develop a custom model, the insurer must determine which credit characteristics to use. Credit characteristics are variables with values determined by the information contained in credit reports. Depending on the credit-reporting agency, there may be hundreds of characteristics. Selecting those with the greatest correlation to the loss statistics can narrow the number of characteristics. Another method is selecting the credit characteristics that best divide the sample population into groups with homogeneous credit experience. Applying multiple techniques may provide the best approach.

It is likely that the insurer will decide to create a limited number of discrete credit classifications. These credit groups must be large enough to have credible experience and small enough to be relatively homogeneous.

Adjusting for Variable Interaction

The interaction of credit with other rating variables requires adjustments. Several accepted techniques allow the analysis of credit simultaneously with the existing rating variables.

Including multiple states in the research database creates one potential bias. Adjusting the data for each state or analyzing each state individually avoids distortion in the results.

Considering the What-Ifs

Finally after the analysis has resulted in a credit model and a method for incorporating it into the rating plan, the insurer must decide how to apply credit to rate specific policies.

What if credit information is not available for a policyholder or only minimal credit information is available? These credit "no-hits" and "thin-files" could be young people who haven't established a credit history.

If there is a single person on the policy and credit information is not available, will that policyholder get the best rate, the worst rate or some other rate? If there is more than one person on the policy, can the credit information of the other household members be used to substitute for the missing credit information? Another possibility is allowing the credit "no-hits" and "thin-files" to form distinct credit classifications.

How often will the insurer order credit information for a policy? It is important to understand the volatility of credit history and the movement between classifications over time.

Protecting Proprietary Information

When an insurer files a custom credit model in a particular state, the insurer may be able to keep the model private, or it may have to be disclosed as part of the public record. If it must be disclosed, the insurer may consider a commercial credit score, or an altered version of the custom model for use in that state so it doesn't have to divulge the knowledge that has been gained through the data-collection and analysis stages.

Using Credit

Credit provides a very useful way to further segment business into groups with substantially different loss experience. Credit history is more objective and verifiable than some traditional rating variables, such as annual mileage and vehicle use. Pricing for credit differences may allow an insurer to serve markets that it has not served before. But be prepared for lots of work, many challenges and quite a few frustrations.

Susan E. Kent is an associate actuary with Nationwide Insurance in Columbus, Ohio. Philip A. Baum is a senior actuarial officer at Nationwide.

Knowing the Score

Credit histories gain favor with personal-lines underwriters.

Credit histories, which have sparked some controversy when used as an underwriting tool by personal-lines insurers, now are finding their way into the setting of rates for individual consumers' policies. Called "credit scoring" by some in the business, the practice is unanimously praised by insurance professionals but is prompting scrutiny from regulators and consumer groups. Regulators in Illinois, for instance, recently have begun to examine the practice in response to a growing tendency among the state's biggest insurers to apply the practice to auto premiums.

Robert Hartwig of the Insurance Information Institute, a nonprofit research organization for the insurance industry, said a more proper term for the practice is "insurance scoring," which avoids confusion as to whether the client is being considered for a line of credit.

Insurance scoring takes into account personal credit histories--including bankruptcies, foreclosures, late bill payments and outstanding debt--in setting premium rates or making underwriting decisions. Hartwig said insurance scoring has been used for decades to write policies for businesses and has been highly useful in that market.

Loss Predictor

Allstate Insurance Co. has used credit histories for auto and homeowners underwriting purposes in Illinois since 1994. In January, the insurer started applying insurance scoring to auto premiums for rating purposes, said spokeswoman Sharon Cooper. In a recent nationwide review of about 11 million auto policies, Allstate found that people with evidence of serious financial instability incurred about 40% more auto insurance losses, she said.

"It's a significant predictor of future loss for auto insurance," she said of credit histories.

Cooper added that different insurers use insurance scoring in different ways to determine premiums. "The use f credit histories varies significantly from company to company," she said. "There are huge differences in what companies do and how they do it."

Allstate tends to use credit histories for underwriting purposes in homeowners insurance; in auto insurance, Allstate uses them for both rating and underwriting purposes, Cooper said.

Le Roi Brashears, a spokesman for Safeco Insurance Cos., said Safeco uses insurance scoring in auto and homeowners underwriting decisions, but "we don't have plans to set rates" using credit histories.

"The credit score is woven into the insurance score, and that's woven into our technical underwriting process," he said.

Safeco began using credit scoring on its in-force book of business last December, which has led to higher premiums for some customers with poor credit. The company began the project in 15 states and expects to complete its rollout before the end of the year.

State Farm's auto underwriting model uses prior-loss history and "certain credit characteristics" to develop an underwriting score for customers, a spokesman said.

"Our model makes fairly unique use of credit characteristics," said Dick Luedke, a spokesman for State Farm Mutual Automobile Insurance Co. "We look at certain aspects of a credit record, but we don't look at everything. You can't use our model to determine creditworthiness, as a bank would."

Luedke cited studies done by the Virginia Bureau of Insurance, the Arizona Department of Insurance and the Casualty Actuarial Society, all of which, he said, demonstrated a "strong correlation between credit history and the likelihood of future loss." State Farm, he said, also conducted studies that reached similar conclusions.

Hartwig said the practice is "quite commonplace," fair and effective in determining rates.

"It's a very good predictor of future loss," he said. "In fact, it's a better predictor than many others." Hartwig added, however, that insurance scoring doesn't replace any of the other common predictors insurers employ; "it's just an added tool."

Regulatory Scrutiny

Nan Nases, spokeswoman for the Illinois Department of Insurance, said the practice has been common for homeowners insurance, since banks use credit histories to make mortgage and related decisions. "I think there was a natural spillover in that area," she said. But Nases said auto insurers have begun using the practice only recently.

"We're definitely aware of the practice, and we're keeping an eye on it," Nases said. "Currently, there is nothing in Illinois law that prohibits the practice. We want to make sure it isn't used for discriminatory purposes."

Nases said there was no proposed legislation related to insurance scoring but that some state legislators are considering holding hearings on the subject.

Insurers say they have not had any problems with regulators, and complaints from consumers have thus far been negligible. Nases also said the department hadn't received any complaints from consumers, but she cautioned that may change if the practice becomes more widespread.

"It is nothing new," Hartwig said. "It is relatively new to the individual consumer, but it doesn't lend itself to unfair practices." He said research has shown that upper-income consumers with bad credit histories are just as likely to generate high claims as lower-income policyholders.

Consumer Resistance

There has been some resistance from consumer groups related to credit-scoring by insurers. A housing group in Richmond, Va.--Housing Opportunities Made Equal Inc.--and five other fair-housing groups filed a federal lawsuit against Citigroup, alleging that its Travelers Property Casualty Insurance subsidiary used credit ratings to determine eligibility for homeowners coverage. The group claims that practice discriminates against residents of low-income neighborhoods.

The lawsuit includes fair-housing groups in Washington, D.C.; New Orleans; Milwaukee; and Toledo, Ohio.

A similar suit filed by Housing Opportunities Made Equal in 1996 against Nationwide Mutual Insurance Co. ended with a $17.5 million settlement and an agreement by Nationwide that it would work to reach further into urban markets. Nationwide did not admit any wrongdoing in the settlement. (See "Urban Renewal," Best's Review, October 2000.)

Hartwig insists, however, that insurance scoring is not a vehicle for discrimination.

"Insurance scoring can work in good and bad ways," he said. "But studies have shown there is a lack of correlation between income and credit scoring. A good credit history can encourage an insurer to accept someone who otherwise might not seem desirable."

Function of Technology

Luedke agreed, adding that State Farm has found that using credit histories can work to the advantage of lower-income drivers when it comes to auto policies. "Our auto model is an underwriting, rather than a rating, tool that can work to the advantage of someone who had an auto accident, say, but otherwise can be shown to be responsible," he said. "Basically, what it does is help to determine whether they go into a standard or a preferred policy."

Luedke said State Farm had been using insurance scoring extensively for auto policies but just recently expanded its use in homeowners policies. He said State Farm's insurance scoring model is being used in 41 states, while the homeowners model is being tested through a pilot program in eight states.

Hartwig said the spread of insurance scoring to the individual auto and homeowners market is a "function of the technology age," since credit reports are much more easily compiled and accessed through computer and communications systems.

Safeco's Brashears agreed that insurance scoring has allowed the company to automate and speed up the underwriting process, while making online underwriting feasible.

David Pilla