HB 47-INSURANCE DISCRIMINATION BY CREDIT RATING Number 0048 CHAIR ANDERSON announced that the first order of business would be HOUSE BILL NO. 47, "An Act prohibiting discrimination by credit rating or credit scoring in certain insurance rates; and providing for an effective date." Number 0118 SAM SORICH, Western Regional Office, National Association of Independent Insurers, explained that his group is an association of insurance companies. He testified that 100 of his member businesses are doing business in Alaska, and they account for about 60 percent of the car and homeowners' insurance written in the state. He stated: The business of insurance is a unique business. When we sell our product, we don't really know the ultimate price of that product; we don't know who is going to file a claim, and when a claim is filed, what the ultimate cost is. What we do is we estimate our costs and charge premiums. One approach is to charge everybody the same premium and just average things, but that wouldn't be fair. So what the insurance industry has done over the years is to identify factors that are proven predictors of loss. So insurance companies use the claims history of a person, that type of car a person drives, how a person uses a car, the type of home construction as factors that are good predictors of whether or not a person is going to have a loss. MR. SORICH said: Some insurance companies have recently introduced the use of credit information as a predictor of loss. As Mr. Lo will show you, this factor is based on a solid, scientific evidence. The Fair Isaac Company has looked at millions of credit reports and millions of insurance policies and the loss history on those policies and have determined, have seen a clear relationship between certain credit factors and the likelihood that a person is going to have an insured loss. And it's not just the Fair Isaac research that has established this relationship. There's been a body of research, most recently a report that was done by the University of Texas Business School, not paid for [by] the insurance industry, not paid for by the insurance agents, but paid for by the Texas legislature. This report came out a month ago from the University of Texas and it confirms the relationship between loss experience and credit characteristics, specifically an insurance score. And Mr. Lo will explain what an insurance score is. Insurance companies have a responsibility to our customers to consider this evidence. Because if we are forced to reject this evidence, we are going to be forced to charge people more than they should be paying for their insurance based on proven cost predictors ... Number 0351 MR. SORICH continued: First on the legal basis, the Federal Fair Credit Reporting Act was enacted in 1970. It specifically allows insurance companies to use credit information. Now states can enact laws that restrict the use of this information, but any state enactment cannot be inconsistent with the federal law. No state has absolutely prohibited the use of credit information. There are many states, though, that have restricted the use. In terms of the consumer benefits, the use of credit information helps insurance companies make decisions that are objective. An insurance score does not consider a person's income, where a person lives. What kind of car a person drives, a person's race, a person's income, is not considered in an insurance score. An insurance score is based on the objective information in a credit report. MR. SORICH testified: Secondly, the use of credit ... gives insurance companies more information about our policyholders so we can make fairer decisions. We are a fact-based industry. We make our decisions based on information. The more information we make, the more well informed our decisions are and the fairer our decisions will be. Third, the use of this tool helps us to make our decisions more quickly. When people apply for insurance, we are able to make a decision more quickly based on the insurance score of how much to charge a person and whether or not the policy will be written. That helps to lower the cost of insurance. Fourth, the use of this information helps us to establish a higher level of equity and fairness. Again, the body of evidence that establishes this relationship does show that this is a good predictor. If we are forced to ignore this tool, our rates will be less fair. Number 0497 MR. SORICH noted: And finally, the use of this tool is helping insurance companies to make insurance more available. We are in the business of writing business. We're not in the business of not writing business. Insurance companies want to offer coverage. And providing them this additional information gives insurance companies the security of knowing ... the likelihood of whether or not a person will file a claim is. ... There is some controversy, no doubt about it. But I think there are good reasonable answers to some of the controversy. MR. SORICH stated: The National Conference of Insurance Legislators (NCOIL) last November adopted a model act, the NCOIL model. The NCOIL model was adopted not by insurance companies, not by insurance agents, but by legislators, state legislators just like you. The NCOIL model addresses many of the concerns that have been raised, and the NCOIL model is reflected in the committee substitute (CS) for HB 47. A number of states have already passed the NCOIL model: North Dakota, Nebraska, Kansas, Oklahoma. And just yesterday, the Georgia legislature passed a bill that's modeled on the NCOIL model, and that's on the governor's desk. So I would encourage the committee to consider adopting ... committee substitute HB 47 because it's based on NCOIL and it's got a solid body of research behind it. MR. SORICH continued: ...There has been some talk about what Hawaii's law is or isn't. Hawaii's insurance code does not prohibit insurance companies from using credit completely. What the Hawaii law says, and it's only applicable to auto insurance, there's no restriction on the use of credit information for homeowners' insurance in Hawaii. The Hawaii law does prohibit insurance companies from using credit information to develop rates. There is a question about whether the law also applies to an underwriting decision, and that issue is before the Hawaii Supreme Court. ... But again, no state has completely prohibited the use. Number 0671 EDDY LO, Insurance Manager, Fair Isaac Corporation, testified about modeling that is based on credit information. He referred to page 5 of the presentation package. Mr. Lo reiterated that the Fair Credit Reporting Act allows the use of credit information. His company, Fair Isaac, as a modeler, started studying the use of credit over 30 years ago. But in the last 10-12 years, he said, the company has applied technology to the use of credit information, predicting losses in personal, auto, and homeowners' insurance. The company works on the individual policy level, matching the premium and loss to credit information on a one-to-one basis. Company officials examine the set of credit characteristics that distinguish whether there was a loss, and based on that likelihood, they predict the loss potential for that same class of business with similar credit characteristics. Those are the basics behind [the modeling], he explained. Number 0755 MR. LO emphasized several preliminary points. He explained that Fair Isaac has built many kinds of predictors. One predictor is the FICO (Fair, Isaac and Company) score that predicts the likelihood of repayment of loans, auto loans, and mortgage loans, but he emphasized that he was not talking about those predictors today. Fair Isaac also has developed a different set of credit-based predictors for personal, auto, and homeowners' insurance, which are the subject of today's presentation. He said that the key relationship from the credit data is that whenever there is an increase in financial obligation as reflected in a credit report, there is an increase in actual losses and an increase in the prediction of future losses. He stressed that all the issues that he will discuss are based on this found relationship between increased financial obligation and future losses. Number 0830 MR. LO highlighted the graphic on page 5, upper left column, which lists five credit characteristics that figure in insurance losses. The middle box in the left hand column shows the number of months since the most recent adverse public record, that is, a bankruptcy, foreclosure, judgment, or lien. In this example, when the losses happen within the last four years, the loss ratio, relative to the group that has no such adverse public record, is 68 percent higher. He said that very significant difference would be useful to an underwriter. The first box on the right hand side looks at the number of adverse public records -- the number of bankruptcies, foreclosures, judgments, and liens in the past. He explained that 96 percent of people have no such adverse public record, so the majority of people are not affected by this particular credit characteristic. But for the remaining 3-4 percent, the loss ratio is simply a measure of loss performance on a policy. So when loss ratio is high, then either the actual loss is high or the expected loss is high. On that basis, he said the company finds that anyone with more than one adverse public record has a loss ratio of 64- 68 percent, a very significant number relative to the group that has no such adverse public records. Number 0970 MR. LO pointed to the middle box on the right hand column, the number of trade lines that are delinquent. He explained that a trade line is any entry on a credit report such as a credit card, a mortgage loan, or an auto loan. The company looks at how many of those were more than 60 days delinquent in the last two years. Basically, 89 percent of the population does not have such delinquencies on their credit reports. But of the remaining 11 percent [of the population], 15 percent of the time, the more delinquency there was in the past, the higher the losses in the past. Therefore, [the losses] are expected to be higher in the future. MR. LO said that another credit characteristic is the number of collections. He said that 97 percent of the people have no such collections on their credit report. For those [people] that do, the loss rate shown is again very high. He said there's a very clear distinction. Number 1039 MR. LO explained that the fifth characteristic is the number of trade lines opened in the last 12 months, the bottom box in the right hand column. These trade lines are not actual financial obligations; they are indications that someone was seeking obligations. The more a person seeks financial obligation or the use of credit, the higher the losses were in the past, and will be in the future. MR. LO summarized that in using these five characteristics, there is a clear distinction. He said there are more characteristics, but he is only allowed to show five in a year to the public. There are other models out there -- Fair Isaac is only one model -- which are very creative in the use of credit information as the model attempts to produce very powerful results. He turned to the results of using five characteristics. The bottom box, left hand column on page 7 [deals with] the number of adverse public records, using that information to separate risk. Since the majority of the people have no such adverse public record, they are assigned 30 points. People with one or more adverse actions are assigned zero. This one characteristic and the two attributes make it possible to separate out risk. Focusing on the months since the most recent adverse public record, points are assigned as follows: 30 points to people with no public records, zero points for less than four years, and 10 points for more than four years. He summarized how this very simple numeric scheme can separate out the risk. With the characteristic of delinquent trade lines [illustrated in the bottom box on the right side], he said it is known that the more delinquency in a person's credit record, the worse the person's risk [for an insurance loss]. Points for delinquency are assigned as follow: 25 for no deliquency, 10 to one delinquent trade line, and zero for two or more. For the number of collections, 20 points are assigned to a person with no collection history. The last characteristic, number of trade lines opened on the last 12 months, is shown on page 8, [top left-hand box]. On these five characteristics, this model predicts a risk between the lowest score of zero and the highest score of 125. On that basis, he said his company is able to tell one risk from the other. If the company uses more than five characteristics, the model is even more powerful with finer [delineations of risk]. Number 1232 MR. LO turned to a commercial value model on page 2, the bottom right hand corner. He pointed to the smooth, downward sloping set of bars, each representing a group of policies in a certain score range. The low scores are on the left side, and the high scores are on the right side. He explained that when the score value is low, the policy could generate 70 percent higher or worse [losses] than the average. The score values on the right are high and they predict that the policy will have future losses 50 percent better than average. The swing from 70 percent worse to 50 percent better is a separation that really helps an underwriter [decide whether to] accept, reject, or tier the risk into the proper rating structures. Number 1283 MR. LO concluded that FCRA gave the legal base to collect information and statistics throughout the years, and models were built, not just for insurance but for other businesses as well. For insurance, the model was built, validated, proven, used, and implemented in a number of states. The insurers can use [the model] to underwrite on a selection basis or can assign a risk into a certain tier by a score range. CHAIR ANDERSON announced that HB 47 and questions pertaining to it would be held until next week, when the sponsor can present the bill. Witnesses will be able to testify then also.