THE PREDICTIVE VALUE OF CREDIT-BASED INSURANCE SCORES

Transcription

1 THE PREDICTIVE VALUE OF CREDIT-BASED INSURANCE SCORES Abstract The application of consumer credit information 1 is widespread throughout the United States, used predominantly by financial services institutions. Knowledge of this information enables lenders to grant credit more equitably and allows insurers to more accurately evaluate risk levels for underwriting and rating purposes. This paper examines the predictive power of creditbased insurance scores employed by personal lines insurers and the impact on insurance prices and availability if credit history is no longer allowed in the risk assessment process. Executive Summary Credit-based insurance scores have been widely used by insurance companies for more than a decade, if not longer. As mentioned in a Federal Trade Commission study released in July 2007, the 15 largest automobile insurers have adopted this method. 2 In fact, the majority of insurance companies now use credit-based insurance scores to some extent, because they are highly predictive of future insurance losses. Insurance scores measure the insurance risk not the credit risk of consumers; they are generated from credit histories and insurance claims data, and are distinct from scores based solely on credit histories intended for other business purposes such as making decisions on loans. The development and use of insurance scores vary from company to company, depending on the type of scoring model and the parameters and weights applied. In the highly competitive personal lines insurance markets, insurance scores can be used for different purposes by different insurers, e.g., underwriting; pricing; tier (i.e., rating level) placement within the same company; and placing a risk in a separate company within the same group. Although insurance scores are not the sole criteria underlying these decisions, they have been found to be among the three most important personal automobile risk factors used because of their value to the accuracy of the risk assessment process. 3 Since its inception, there has been much debate surrounding the use of credit information in personal automobile and homeowners insurance. 4 These discussions have prompted various studies by academicians, actuaries and government agencies that explore the impact of one s credit history on the 1 Today, the three national credit bureaus that provide credit scores are Equifax, Experian and TransUnion. Fair Isaac Corporation is another leading credit information provider as well as a statistical modeler that develops FICO credit scores for individuals. These scores, which are most widely used in the mortgage industry, are based solely on information in consumer credit reports maintained at the three bureaus. 2 Federal Trade Commission, Credit-Based Insurance Scores: Impacts on Consumers of Automobile Insurance (July 2007). The 15 largest automobile insurers represented 72 percent of the market in EPIC Actuaries, LLC, The Relationship of Credit-Based Insurance Scores to Private Passenger Automobile Insurance Loss Propensity, June Although some insurers may use credit information to evaluate business risks, the scope of this paper deals with personal insurance products (i.e., personal automobile and homeowners insurance).

2 Page 2 ability to obtain insurance and the amount paid for coverage. Analyses have found a strong correlation between credit history and the risk of insured loss. The use of insurance scores increases coverage availability for more consumers and helps keep insurance policies in force; companies are able to accept applicants or renew policyholders who probably would not be underwritten or renewed if the use of scores were not allowed. This practice also allows pricing to be more competitive and more advantageous for policyholders due to greater accuracy in matching the cost of insurance to the level of risk. As such, lower-risk customers pay lower premiums and higher-risk customers pay higher premiums for their automobile and homeowners coverages. More consumers benefit from the use of credit-based insurance scores than not. If restrictions on the use of insurance scores were imposed, then fundamental economic and mathematical principles indicate that increased insurance premiums would be given to one group of policyholders to offset decreased premiums given to another group. In this case, since more policyholders have better-thanaverage credit history and hence insurance scores, it is the majority who would have to pay more to subsidize the decrease given to the remaining minority. Not only would the majority of policyholders (i.e., those less likely to file a claim due to lower risk exposure) be unfairly overcharged if restrictions were placed on the use of insurance scores, but the involuntary market might grow. In the past, insurers in the voluntary market have been less willing to write those individuals and households who were not paying their fair share of premiums. An ancillary effect may be a greater increase overall in insurance losses and hence premiums since there would be less correlation between prices and expected costs, and individuals would therefore have less incentive to act more responsibly themselves in reducing losses. In a competitive environment where there are fewer restrictions in the ability to use actuarially sound factors such as insurance scores, more companies are likely to enter or stay in markets. A growing number of carriers providing a wider array from which to select can only benefit the insurance-buying public. Introduction to Credit-Based Insurance Scores An individual s credit history can range from a few to potentially hundreds of different credit characteristics such as the length of credit history, current payment status, amounts owed, history of delinquencies, proportion of available credit used, credit-seeking behavior, and public record data (e.g., bankruptcies, tax liens, foreclosures, and judgments). 5 The introduction of credit-based insurance scores (i.e., insurance scores) began to take hold in the early 1990s, although many property/casualty insurers did not adopt this practice until the latter part of the decade. 6 Insurance scores are part of the risk assessment process adopted by companies and used to determine the likelihood of filing claims and their insured costs. 5 Personal information such as ethnicity, religion, gender, marital status, nationality, age, income and address is typically not reflected in a credit score provided by credit bureaus. Insurance companies do not use data on race, ethnicity, national origin or income when screening applicants or existing customers, nor do they have access to such data unless required by law. 6 According to Conning Research and Consulting, Insurance Scoring in Personal Automobile Insurance: Breaking the Silence, 2001, Fair Isaac reported that the number of insurers using its product has increased from roughly 200 insurers in 1995 to more than 300 today (in 2001). In 2005, about 350 firms obtained credit scores from Fair Isaac.

3 Page 3 Using high quality statistical tools, i.e., credit scoring models, insurance scores based on information from credit reports and insurance claims data are generated using a mathematical formula. While many companies use statistical scoring tools created by modeling firms such as Fair Isaac and ChoicePoint, more companies are now designing proprietary models using their own data and methodology. 7 From an actuarial perspective, credit-based insurance scoring models (that comply with legal, regulatory and business standards) are developed using a statistical analysis of the relationship between the credit histories of individuals and the claims histories of those same individuals for a given insurance product, in this case automobile or homeowners insurance. Further work is then done to combine the credit characteristics and statistical weighting factors to produce an insurance score for each individual that is highly correlated with the likelihood that he or she will file an insurance claim. An insurance score, typically shown as a three-digit figure, is essentially treated like traditional personal lines insurance underwriting and rating factors (e.g., driving record, accident and violation history, use of vehicle, make and model of vehicle, home construction, amount of fire protection, etc.). However, one s credit history differs from data underlying other rating factors to the extent that it comprises incredibly rich, detailed and robust information reflecting multiple and ongoing responsibilities to cover the mortgage, utilities, insurance, student loans, etc. In contrast, insurance company data are slow to change or are relatively thin, statistically. For example, typical payment transactions occur very infrequently, perhaps once or twice a year, and most policyholders do not file claims so their records show little or no data related to losses, accidents, and violations, etc. Furthermore, one s marital status, age class, and use of vehicle or fire protection, usually do not change rapidly, if at all. Credit information is thus unique in the sense that different models can be used to evaluate large amounts of data and produce scores that are typically not applied to other factors characterized by limited data. Because credit-based insurance scores successfully identify relatively high-risk and low-risk individuals even when measured against traditional underwriting and rating classifications, these scores contribute to a more predictive and reliable risk classification system. This has two major impacts: (1) the use of insurance scores improves underwriting equity by allowing more accurate risk evaluation and provides insurers with a greater opportunity to write business in all markets. Since they can develop more refined risk classification plans, they are more inclined to enter new lines or product niches; and (2) superior risk classification systems are a tremendous competitive advantage for the insurers using them, as they help to identify risks that do not qualify for lower premiums based on traditional rating factors only. With a system that uses credit-based insurance scores, companies are able to give a better price to those risks that are more financially responsible. Research Findings on the Use of Credit-Based Insurance Scores Since the introduction of credit-based insurance scores, many have wondered, What does my ability to pay bills have to do with the way I drive or maintain my home? There also are concerns that the use of insurance scores may affect the availability and affordability of coverage, and factors included in scoring models may have adverse effects on certain populations, particularly minorities and lowincome groups. It may be surprising to learn that since most people have relatively good credit history, 7 According to the Federal Trade Commission, Credit-Based Insurance Scores: Impacts on Consumers of Automobile Insurance, July 2007, the use of credit-based insurance scores has grown dramatically some of this growth is attributable to changes in technology and industry practices that have made it easier for companies to develop and use these scores. This made it easier for them to develop proprietary scores

4 Page 4 the application of this factor has affected the price and availability of automobile and homeowners insurance in a positive way for them. Based on survey results reflecting about 3 million personal lines policies, the Arkansas Department of Insurance 8 found that more than 9 out of 10 insured consumers in the state (91%) had lower premiums than they would have otherwise due to their credit history, or were not affected at all by the use of insurance scores. In particular, almost every single homeowner (95%) had some level of cost savings because of insurers use of credit information or was unaffected by its usage. Results of another survey conducted by the National Association of Independent Insurers (NAII) 9 several years ago found that the use of insurance scores enabled nearly nine out of 10 personal lines carriers to offer lower rates for their policyholders who present less risk. In addition, three-fourths (75.7%) and two-thirds (66.0%) of survey participants that use credit information were respectively able to accept applicants and renew policies that would probably not be accepted or renewed if credit histories were not allowed. How Use of Insurance Scores Has Affected Companies Personal Auto and Homeowners Insurance % 66.0% % 34.0% Accept More Applicants Renew More Policies Yes No These findings are supported by the Alaska Division of Insurance 10 in its questionnaire of insurance companies that concluded: In the aggregate... the number of policyholders increased by approximately 8% from 1999 to Likewise, New Jersey s 2003 automobile insurance reform package that permitted the use of credit histories influenced additional large carriers decisions to enter the market and one leading carrier s decision to temporarily postpone its withdrawal. Furthermore, over 500 new agents were appointed by insurance companies, making automobile coverage more accessible in this state. 8 Arkansas Department of Insurance, Use and Impact of Credit in Personal Lines Insurance Premiums Pursuant to Ark. Code Ann , July 9, 2008, for the State House and Senate Committees on Insurance and Commerce 9 National Association of Independent Insurers, 2002 Survey on the Use of Credit Information in Personal Lines Insurance; NAII is one of the predecessor organizations to the PCI. 10 State of Alaska Division of Insurance, Insurance Credit Scoring in Alaska, February 21, 2003

5 Page 5 Other research efforts by the insurance industry, academicians, actuarial consulting firms and insurance departments varying in scope, methodology and data have been conducted to analyze the effectiveness of insurance scores. Common findings among several different correlation analyses 11 in the past several years include the following: There is a positive relationship between one s credit history and insurance claiming behavior; i.e., people who are worse credit risks tend to file more claims than people who are better credit risks. Insurance scores successfully distinguish individuals and households with higher risk of insurance claims from those with lower risk of claims. The use of insurance scores is actuarially justified and improves the relationship between the insurance risk and the related price, so that those with a higher risk exposure pay more and those with a lower risk exposure pay less. Individuals and society as a whole can therefore make good tradeoffs between risk acceptance, risk avoidance, and risk mitigation. Insurance scores are unique in that they offer additional explanatory predictive information about risk not found in other variables. They add predictive ability even when all traditional underwriting and rating risk classification factors have been considered. No traditional risk factor can be used as a replacement for the accuracy of the credit-based insurance score. Even the federal government has studied this issue. Pursuant to Section 215 of Title II, the Fair and Accurate Credit Transactions Act (FACTA) of 2003, the effects of credit-based insurance scores on the insurance market were examined by the Federal Trade Commission (FTC). 12 Similar to earlier studies, the FTC 13 found that companies ability to evaluate risk with greater accuracy means they are more willing to offer auto insurance at an appropriate price to higher-risk consumers who may otherwise not be able to obtain coverage in the private marketplace. Because insurance scores can make the process of granting and pricing coverage quicker and less expensive, the FTC confirmed that cost savings may be passed on to consumers in the form of lower premiums. Furthermore, insurance scores were shown to have only a small effect as a substitute or proxy for membership in racial, ethnic and income groups in estimating insurance risk. They remain strong predictors of risk when controls for these variables are included in statistical risk models. After trying a variety of approaches, the FTC was not able to develop an alternative credit-based insurance scoring model that would continue to predict risk effectively and lower the differences in scores on average among racial and ethnic groups. Although a model might be constructed that meets 11 Studies include: (1) The University of Texas at Austin, A Statistical Analysis of the Relationship Between Credit History and Insurance Losses, March 2003; (2) EPIC Actuaries, LLC, The Relationship of Credit- Based Insurance Scores to Private Passenger Automobile Insurance Loss Propensity, June 2003; and (3) Texas Department of Insurance, Supplemental Report to the 79 th Legislature, Use of Credit Information by Insurers in Texas: The Multivariate Analysis, January 31, The Federal Reserve Board was also mandated to study the use of credit information and its impact on the credit market. 13 Federal Trade Commission, Credit-Based Insurance Scores: Impacts on Consumers of Automobile Insurance (July 2007)

6 Page 6 both of these objectives, the agency firmly believes that no readily available scoring model currently exists. 14 Impact of Banning or Restricting the Use of Insurance Scores on Consumers The majority of people have insurance scores that currently afford them better insurance rates overall than what they would be expected to receive if credit histories were not permitted. Eliminating or placing constraints on this risk assessment factor would penalize these individuals and harm the competitive efficiency of the insurance marketplace. The result would be premium increases for those who are more likely to pay their bills and less likely to incur insurance losses so that premium decreases can be given to others. According to former Texas insurance commissioner Jose Montemayor, 15 Effectively, we would ban risk-based pricing and underwriting and revert to a pricing system where we homogenize the risk and essentially charge everyone the same price regardless of risk. That would be a set-back to all Texans, of all races, especially those of moderate to lower income whose risk remains low. Moreover, consumers may not be able to obtain insurance in the voluntary market and, as a last resort, find it necessary to use the involuntary market mechanism where coverage selection may be limited and prices may be higher. Restrictions on the use of insurance scores would make it more difficult for companies to write in certain markets, depriving policyholders of the benefits of a truly competitive environment. Abolishing the use of this predictive tool may also provide little incentive for individuals to act more responsibly themselves in preventing or reducing future losses (e.g., driving more cautiously or installing fire extinguishers or security systems in their residences). Such a distortion in incentives would tend to cause insurance costs and, hence, premiums to rise even more. Threats to underwriting and rating standards have been ongoing for more than forty years. No doubt, insurers have a good understanding of the impact that these types of changes will have on their business models and already know to some extent the effect on their policyholders if the use of insurance scores is banned or restricted. In general, one insurer in the NAII survey stated, (Removal of the use of insurance scores) would cause a flattened premium structure because one factor to pinpoint rates for a customer would be lost. Other respondents in the 2002 study provided more specific consequences affecting the company and their own customers if restrictions had been put into place:.we would need to raise the rates of our preferred business by as much as 7 percent over the next five years in order to maintain break-even profitability; We would be forced to charge a higher pooled rate to our policies that benefit from the use of credit; 14 The Federal Reserve Board followed with its own independent study on credit scoring, Report to the Congress on Credit Scoring and Its Effects on the Availability and Affordability of Credit (August 2007). Its findings support those of the FTC, concluding that there is no compelling evidence that any particular group has experienced markedly greater changes in availability and affordability than other groups due to credit scoring. Moreover, credit characteristics included in scoring models do not serve as substitutes or proxies for race, ethnicity or sex. 15 Texas Department of Insurance, Supplemental Report to the 79 th Legislature, Use of Credit Information by Insurers in Texas: The Multivariate Analysis, January 31, 2005

7 Page 7 two-thirds to three-fourths of insureds who now benefit from lower premiums would see premium increases to subsidize those who have poor scores and higher losses; We would have to eliminate the use of our preferred auto tier and eliminate discounts based on credit scores; and It would have a major impact in our risk selection process. The quality of our book of business would not be as sound; thus, this would result in higher loss ratios. In turn, we would be required to increase our rates for all insureds. The rate increase scenario was precisely what occurred in Maryland, when the use of credit information by homeowners insurers was prohibited beginning in October Customers who received discounted insurance rates because of high credit scores were said to face double-digit percentage increases in their homeowners premiums. After the ban, one company analyzing the cost impact found that 75 percent of its homeowners policyholders would see rate increases. 17 Other comments pertaining to insurance-scoring restrictions made by NAII survey participants follow: (There would be) increased rates for many and considerable administrative programming costs would change. (We would have to) tighten underwriting guidelines; We would probably lose new growth as well as in-force (renewals); and It would prevent us from charging adequate rates for a large portion of our new business, leading to subsidization of those who don t maintain favorable credit histories by those who are responsible in how they handle their credit. Failure to include the insurance credit score will inevitably lead to rate inequities and subsidization. In conclusion, the credit-based insurance score is one of the most accurate and equitable factors used in pricing and underwriting personal lines insurance; it provides information about a risk that no other factor can provide. Consumers benefit from the use of credit-based insurance scores, since credit risk evaluation allows more people to obtain and keep their coverage and receive increased savings in their insurance premiums. If this factor were no longer allowed, policyholders would be placed into less refined and less accurate rating categories and there would be a redistribution of premiums so that more financially responsible insureds would be overcharged to subsidize less financially responsible insureds. Clearly, the majority of the insurance-buying public would not be well served if restrictions were placed on this valuable, predictive risk assessment tool. The Property Casualty Insurers Association of America (PCI) is a trade association consisting of more than 1,000 insurers of all sizes and types, and representing 35.9 percent of the nation s property/casualty business. PCI members comprise 43.4 percent of the personal automobile market and 29.6 percent of the homeowners market. 16 Maryland Ins. Code , as amended by Ch. 580, The Washington Times, Maryland Ban on Credit Scores Raises Home Insurance, June 2003