Credit Scoring

Credit Scoring

THE TOPIC

AUGUST 2012

The goal of every insurance company is to correlate rates for insurance policies as closely as possible with the actual cost of claims. If insurers set rates too high they will lose market share to competitors who have more accurately matched rates to expected costs. If they set rates too low they will lose money. This continuous search for accuracy is good for consumers as well as insurance companies. The majority of consumers benefit because they are not subsidizing people who are worse insurance risks—people who are more likely to file claims than they are.

The computerization of data has brought more accuracy, speed and efficiency to businesses of all kinds. In the insurance arena, credit information has been used for decades to help underwriters decide whether to accept or reject applications for insurance. New advances in information technology have led to the development of insurance scores, which enable insurers to better assess the risk of future claims.

An insurance score is a numerical ranking based on a person’s credit history. Actuarial studies show that how a person manages his or her financial affairs, which is what an insurance score indicates, is a good predictor of insurance claims. Insurance scores are used to help insurers differentiate between lower and higher insurance risks and thus charge a premium equal to the risk they are assuming. Statistically, people who have a poor insurance score are more likely to file a claim.

Insurance scores do not include data on race or income because insurers do not collect this information from applicants for insurance.

RECENT DEVELOPMENTS

  • Federal Activities and Congress: Legislation has been introduced in the House of Representatives to ban the use of credit scores to underwrite and rate personal auto insurance. The bill, HR 6129, would amend the Fair Credit Reporting Act. It has been referred to the House Committee on Financial Services.
  • In 2008, the Federal Trade Commission (FTC) asked nine of the largest homeowners insurance companies to provide information that it says will allow it to determine how consumer credit data are used by the companies in underwriting and rate setting. The Fair and Accurate Credit Transactions Act, passed in 2003, directed the FTC to consult with the Office of Fair Housing and Equal Opportunity on how the use of credit information may affect the availability and affordability of property/casualty insurance, whether the use of certain factors by credit scoring systems could have a disparate impact on minorities and, if so, whether the computer models used could be modified to produce comparable results with less negative impact. The final report is expected in 2012, according to the Property Casualty Insurers of America.
  • In a similar study, the FTC found that auto insurers’ use of insurance credit scores leads to more accurate underwriting of auto insurance policies in that there is a correlation between insurance scores and the likelihood of filing an insurance claim. The FTC report, Credit-Based Insurance Scores: Impacts on Consumers of Automobile Insurance, released in July 2007, also states that credit scores cannot easily be used as a proxy for race and ethnic origin. In other words, credit scoring predicted risk for members of minority groups in much the same way that it predicted risk for members of nonminority groups.
  • The Federal Reserve also studied the use of credit scoring. Although looking at credit scoring to quantify risk posed by a borrower rather than an applicant for insurance or a policyholder, the Federal Reserve said in a report issued at the end of August 2007 that credit scores were predictive of credit risk and were not proxies or substitutes for race ethnicity or gender, underscoring the FTC study.

State Activities

  • NAIC: The National Association of Insurance Commissioners (NAIC) has developed a set of questions for state insurance departments to use in the event that they want to survey auto insurance companies operating in their state on risk classification. The information could be used to determine the degree of variance in the use of different factors, including credit scores. States that decide to carry out a survey may also decide whether or not to publish the results. The voluntary use of the risk classification survey is an outgrowth of an earlier, narrower project on credit scoring alone.
  • Legislation: In Michigan, where the issue of credit-scoring was the subject of court actions for many years, legislation has been signed by Governor Rick Snyder that will protect consumers experiencing extraordinary life circumstances such as a serious illness or being unemployed for three months or more, based on the National Conference of Insurance Legislators (NCOIL) model act, see below. Michigan is one of the few states where credit scoring is used only to provide premium discounts. HB 4595 lists the type of events that would qualify as extraordinary and what actions the insurer may require to validate the request for an exception. 
  • Fewer states are considering insurance scoring bills. This year only two states, Michigan and Kentucky, passed legislation and both enacted measures that include NCOIL’s extraordinary life circumstances language.
  • Arkansas: A report from the insurance department shows the impact of insurance scoring on calculations of the final premium in 2011 for some 3.2 million personal lines policies. In 43 percent of those policies, the use of credit information resulted in a decrease in the final premium and in 12 percent of cases, it resulted in an increase. Credit scoring was a neutral factor—meaning it did not affect the outcome—in the remaining 45 percent of policies. Policies for which credit information decreased the premium outnumbered policies for which it increased the premium by 3.4 to 1. Of the insurers reporting premium income derived from personal lines insurance, 133 companies writing $2.1 billion in premium used credit information, compared with 46 companies writing more than $39 million in premium that did not.
  • When analyzed by type of insurance policy, the data showed the use of credit scoring impacted more auto insurance policies positively (44.7 percent) than homeowners policies (40.9 percent). The percentage impacted negatively was about the same for both types of policies. This is the eighth report issued by the Arkansas Insurance Department as required by the Use of Credit Information in Personal Insurance Act. In 2010 about the same number of policyholders saw a decrease in premium.
  • Texas: In May 2011 the Texas Supreme Court ruled that credit scoring, which uses race-neutral factors, is allowed under Texas law even if it has an unintended impact of creating higher insurance premiums for some minority policyholders. The lawsuit was brought by an African-American attorney, Patrick Ojo, on behalf of himself and other minorities whose homeowners insurance premiums had increased because of information acquired through the insurer’s automated credit scoring system. Ojo had never filed a claim. Ojo argued that the credit-scoring system results in a racially disparate impact and therefore violates the federal Fair Housing Act (FHA). The court rejected the plaintiffs’ arguments and said that Texas law supersedes the FHA and that the state’s insurance code does not provide for disparate impact liability.
  • Texas law prohibits the use of race, color, religion or national origin in underwriting and setting rates. Insurers do not collect such information.
  • NCOIL: The National Conference of Insurance Legislators (NCOIL) voted in July 2009 to amend its model law on credit-based insurance scoring. The revision requires insurers using insurance scoring for underwriting or rating purposes to consider “extraordinary life circumstances,” such as serious injury or illness; divorce; death of a spouse, child or parent; identity theft; temporary involuntary loss of employment for three months or more; or a federal or state-government declared catastrophic event. Policyholders would have to document their request for an exception to the insurer’s rules or practices and demonstrate that circumstances they described had had a direct impact on their credit score. The NCOIL model law is in effect in more than half of the states. Several others have laws that recognize the extraordinary life circumstances concept.
  • Only California and Massachusetts ban the use of insurance credit scoring for rating purposes. Hawaii does not permit its use for auto insurance and Maryland bans it for homeowners insurance.
  • Kansas: In April 2010 the Kansas legislature passed a measure that amends the existing insurance credit scoring law instead of banning the use of credit scoring as had originally been proposed by the House in 2009.
  • The legislation included the NCOIL extraordinary life circumstances provision, which requires insurers to take into account personal hardships such as a serious injury that led to usually high medical bills, long-term unemployment or the death of a spouse, see above and Background section. In addition, it removed the requirement that insurers reexamine policyholders’ credit history every three years, making the credit check on policy renewals voluntary. In this way, consumers can opt to have their score re evaluated when they think their score has improved. Not having to check credit history every three years has the added advantage of lowering insurers’ expenses.
  • The bill also modified the trigger for an adverse action notice. Adverse action notices are generally sent when a policyholder’s insurance policy is not renewed or the rate is increased. In Kansas adverse notices were also required if the customer did not receive the best possible rate.
  • Adverse Action Notices: In Washington State, the state insurance department promulgated a rule that requires insurers to provide specific and detailed reasons and explanations, in plain, unambiguous language, whenever they inform consumers of an “adverse action.” Adverse actions include nonrenewal of an existing insurance policy, an increase in premium or a refusal to issue a new policy, based on credit scores. Under the rule, insurers must provide a description of the element in the credit history that adversely affects the consumer’s insurance score. They must also explain how this affects the insurance score and what consumers can do to improve this aspect of their score.
  • In June 2007, the U.S. Supreme Court overturned appeals court rulings in two cases that centered on when insurers are required to send consumers notices to comply with the Fair Credit Reporting Act (FCRA). Siding with insurers on the issue, the high court said that the companies were not breaking the law. The 9th Circuit court had ruled in the first case that an insurer must issue adverse action notices whenever a consumer’s credit information does not result in the consumer receiving the best possible rate but the high court said that such actions would result in too many notices being sent that were likely to be ignored. In the second case, it said that the company had not acted recklessly in willful disregard of the law.

BACKGROUND

Insurance scores are confidential rankings based on credit history information. They are a measure of how a person manages his or her financial affairs. People who manage their finances well tend to also manage other important aspects of their lives responsibly, such as driving a car. Combined with factors such as geographical area, previous crashes, age and gender, insurance scores enable auto insurers to price more accurately, so that people less likely to file a claim pay less for their insurance than people who are more likely to file a claim. For homeowners insurance, insurers use other factors combined with credit such as the home’s construction, location and proximity to water supplies for fighting fires.

Insurance scores predict the average claim behavior of a group of people with essentially the same credit history. A good score is typically above 760 and a bad score is below 600. People with low insurance scores tend to file more claims. But there are exceptions. Within that group, there may be individuals who have stellar driving records and have never filed a claim just as there are teenager drivers who have never had a crash although teenagers as a group have more accidents than people in other age groups.

Credit Report Information—Who Wants It? It is becoming increasingly important to have an acceptable credit record. Whether we like it or not, society equates the ability to manage credit responsibly with responsible behavior, even if individuals have a bad credit record through no fault of their own. Landlords often look at applicants’ credit records before renting apartments to see whether they manage their finances responsibly and are therefore likely to pay their rent on time. Banks and other lenders look at the credit records of loan applicants to find out whether they are likely to have loans repaid. Some employers also look at credit records, especially where employees handle money, and view a good credit record as a measure of maturity and stability.

In some insurance companies, underwriters have long used credit records in cases where additional information was needed. Before the development of automated scoring systems, underwriters would look at the data and make decisions, often erring on the overly cautious side that disadvantaged many more people. Automated insurance scoring and underwriting systems eliminate the weaknesses inherent in someone's personal judgment and have allowed more drivers to be placed in preferred and standard rating classifications, saving them money. With the development of these scoring models, the use of credit-related information in underwriting and rating for many insurers has become routine. Insurers use insurance scores to different extents and in different ways. Most use them to screen new applicants for insurance and price new business.

Why Insurers Need It: Insurers need to be able to assess the risk of loss—the possibility that a driver or a homeowner will have an accident and file a claim—in order to decide whether to insure that individual and what rate to set for the coverage provided. The more accurate the information, the closer the insurance company can come to making appropriate decisions. Where information is insufficient, applicants for insurance may be placed in the wrong risk classification. That means that some good drivers will pay more than they should for coverage and some bad drivers will pay less than they should. The insurance company will probably collect enough premiums between the two groups to pay claims and expenses, but the good drivers will be subsidizing the bad.

By law in every state, insurers are prohibited from setting rates that unfairly discriminate against any individual. But the underwriting and rating processes are geared specifically to differentiate good risks from bad risks. Since insurance is a business, insurers favor those applicants that are least likely to suffer a loss. One of the key competitive aspects of the personal lines insurance business is the ability to segment risks and price policies accurately according to the likely cost of claims generated by those policies.

Insurance scores help insurers accomplish these objectives. Actuarial studies by Tillinghast, an actuarial consultant firm, have shown a 99 percent correlation between insurance scores and loss ratio—the cost of claims filed relative to the premium dollars collected. In other words, people who have low insurance scores, as a group, account for a high proportion of the dollars paid out in claims. Insurance scores developed by the insurance scoring company Fair Isaac involve a set of 15 to 30 credit characteristics, each with an assigned weight, that produce a score ranging from 100 to 999. The lower the score, the greater the risk. According to Fair Isaac, 76 percent of consumers exhibit good or fair credit management behavior. Only four percent of the population are so-called “no hits” with no credit history. This small group would include the very young, who have not yet established a credit history; those who might not use credit on personal or religious grounds; and retirees who have probably paid off their mortgage.

The reasons behind the predictive value of credit scores appear to be behavioral. The character trait that leads to careful money management seems to show up in other daily situations in which people have to make decisions about how to act, such as driving. People who manage money carefully may be more likely to have their car serviced at appropriate times and may also more effectively manage the most important financial asset most Americans own—their house—making routine repairs before they become major insurance losses. But of course, there are always exceptions to the rule. For example, there are people who have filed for bankruptcy that have never filed an insurance claim. Furthermore, a low insurance score doesn't predict that a person will have an accident.

The information used in insurance scoring models does not include personal data such as a person’s ethnic group, religion, gender, family or marital status, handicaps, nationality, age, address or income. The scoring process relies on information in a person's credit record. Particular emphasis is placed on those items associated with credit management patterns proven to correlate most closely with insurance risk, such as outstanding debt, length of credit history, late payments, collections and bankruptcies, and new applications for credit. Credit-related activities within the last 12 months are given most weight.

Common Misunderstandings about Credit Scoring: Many people have no idea they are beneficiaries of insurance scoring. More than 50 percent of policyholders have a lower premium because of good credit, insurers say, although consumers themselves, when asked, think most people do not benefit.

Some consumers are disturbed by the fact that, when applying for insurance, one insurer will reject their application based on their insurance score yet another company will find it acceptable. They ask how insurers' responses can be so different when they are all working from essentially the same credit report information. Many large insurance companies have now developed their own insurance scoring model, using their own proprietary information in combination with standard actuarial data. Even when insurers use the leading vendors of insurance scoring models they may have the model tailored to their own target market. Not all insurers are looking to insure the same kind of drivers or homeowners. Some may target only those with the best scores, with no recent accidents or traffic violations, while others may seek out people with a less than perfect record.

Since virtually all companies use credit information in different ways, insurance scoring fosters competition among insurance companies and more choices for the consumer.

Most people think that insurers can obtain all the information they need from state motor vehicle departments and that reportable accidents, speeding tickets, convictions for drunk driving and other traffic violations are automatically in this age of electronic communication, instantaneously recorded. But, in fact, much of that data is missing from motor vehicle records (MVRs).

A 2002 Insurance Research Council study found that MVRs are typically inaccurate. One in five convictions may be missing. An earlier study found that on average only 40 percent of reportable accidents appeared on MVRs. An analysis of current laws shows the amount of useful information is very limited. Some states don't require records of information that show how drivers perform, such as convictions for drunk driving. If a driver is found guilty of an out-of-state infraction, that information is not automatically provided to the state where the licensed driver or vehicle is registered. Other states offer drivers an opportunity to obtain a lesser sentence or to avoid having information noted in the official record. By contrast, credit records are generally complete and where they are not or are inaccurate, there is a clearly defined review process for correcting the deficiencies.

In short, credit information is generally more accurate, and that works to the advantage of the majority of insurance consumers. With this information available to insurers, a majority of policyholders will pay less for home and auto insurance.

Research: A 2004 study commissioned by the Texas Department of Insurance on the use of credit information by insurers doing business in the state found a strong relationship between credit scores and claims experience. The study also found that the use of insurance scores significantly improves pricing accuracy in predicting risk when combined with other rating variables such geographical area and age of driver. Although there was a consistent pattern of differences in credit scores among different racial/ethic groups, with blacks and Hispanics having worse scores than whites and Asians, on average the results were actuarially supported and not unfairly discriminatory. This means that all drivers with the same credit rating characteristics would be charged the same amount, regardless of race, income or ethnic background. The research, which was required by law, was conducted by the insurance department with assistance from the University of Texas and the Texas A&M University as well as the Office of Public Insurance Counsel. The findings, which were published in December 2004 and January 2005, confirm the results of other studies.

Another earlier Texas study published in March 2003 found a strong correlation between credit history and the filing of an auto insurance claim—both the size and frequency of claims. The Bureau of Business Research at the University of Texas found that when credit scores were matched up with claim data, those with the worst credit scores had claim losses that averaged $918—53 percent higher than the expected average—and those with the best credit score had losses that averaged $558—25 percent less than the average.

A June 2003 study by EPIC Actuaries conducted for the insurance industry also found that overall, insurance scores significantly increase the accuracy of the risk assessment process. Insurance scores, their study showed, are among the three most important risk characteristics for each of the six major automobile coverages. For example, for property damage liability coverage, those with the worst insurance scores had expected losses of 33 percent above average. Those with the best had losses 19 percent below average. Some 2.7 million records were studied.

Some states have examined the issue of whether credit scores have an adverse impact on low-income or minority populations. A February 2004 report issued by the Maryland Insurance Administration (MIA) found that there was insufficient data to conclusively determine whether the use of credit scoring has an adverse impact on these communities because insurers do not collect information on an applicant’s race or income. Without such data, it is not possible to match premiums paid to any socioeconomic group.

The Missouri Department of Insurance claimed in February 2004 that low-income households and minorities are adversely affected by insurance scoring. However, the department’s findings were based on flawed methodologies. For example, it aggregates ZIP code credit score data for everyone in a ZIP code area, whether they own cars or homes and therefore purchase auto or homeowners insurance or not.

Typical Provisions in Legislation Regulating Insurers’ Use of Credit Information:

  • Need to File a Model with the Department of Insurance. Insurers are required to file their underwriting model based on insurance scores with the state’s department of insurance.
  • Restrictions on Factors. An insurance score uses information from an individual’s credit history that has been shown to statistically correlate with claim costs. Restrictions on factors that may be used vary from state to state but may include:  
    1. Total available credit
    2. Disputed items under review
    3. Number of credit inquiries (credit card or loan applications)
    4. Debt from financing payments to hospitals or for health reasons
    5. Use of certain types of credit (personal loans, credit cards)
  • Limits on Use. Different states have proposed varying thresholds, but in general they allow insurers to accept or reject an application based on an insurance score.
  • Prohibitions on Penalizing Consumers with No Credit Histories. While credit cards, mortgages and other debt instruments are widely used today, there are still segments of the population (some elderly people, certain religious sects and some low-income individuals, including students) that have no experience with credit. Regulations generally require insurers to consider an applicant with a so-called “thin” or “no-hit” file an average risk.
  • Sole Use Rules. Insurance companies are usually barred from using insurance scores as the sole determining criteria in making underwriting or rating decisions.
  • Disclosure Rules. Insurers are required to inform consumers they are using credit information in the underwriting/ratemaking process. If that is a deciding factor in rejecting the application for insurance or another adverse decision, in accordance with the Fair and Accurate Credit Transactions Act of 2003 (FACT), the insurer must notify the individual that credit report information was used and may have to make a copy of the credit report available to the consumer free of charge.

Even before the recent surge of interest in the use of credit reports, many states already required insurers to notify their policyholders if credit histories were used or played a role in adverse decisions, such as raising rates or placing a policyholder in a higher rating tier. Many also already barred insurers from using insurance scores as the sole determinant in underwriting — the process of deciding which applicants to accept and classifying those selected —or pricing/rating decisions. As the issue of credit has assumed a higher profile, additional states have passed such laws. Most are based on a model law passed in December 2002 by the National Conference of Insurance Legislators (NCOIL). Among other things, the model legislation requires insurers to disclose to consumers that a credit report may be used and to notify the policyholder in compliance with the federal Fair Credit Reporting Act when credit is the basis for an adverse action. The model law prohibits the use of credit information as the sole basis for refusal to insure or to nonrenew or cancel. It also bars the use of disputed information or information identified as medical collection accounts in the credit report. And it encourages insurers to take into account extraordinary life events, such as catastrophic illness or the death of a spouse.

Another area is how lack of credit history—“no-hits” and “thin files”—should be dealt with. The NCOIL model law says that in such cases either the credit score should be considered “neutral,” or average, or credit as an underwriting factor should not be used at all. As a third option, it allows the insurer to follow a procedure of its own. The justification for this must be provided to the insurance department.

A few states have very restrictive rules. A law passed in Washington State in March 2002 prohibits cancellations after 60 days and nonrenewals based in whole or in part on credit history. Maryland, which had previously allowed the use of information from credit histories, bans the use of credit in homeowners policies and in auto insurance underwriting decisions on existing business. And while credit-related information may be used in rating decisions about new insurance policies, the law imposes a cap on discounts and surcharges related to credit of 40 percent.

Only one state, Hawaii, has a law on the books that bans the use of credit reports for auto insurance underwriting and rating. In California, the use of credit is not permitted under Proposition 103 for rating auto insurance policies unless specifically allowed by the regulator and in Massachusetts, although not banned, regulators will not approve rate filings for auto or homeowners insurance that include the use of credit scoring. According to the Property Casualty Insurers Association of America, 26 states have adopted laws on credit or regulations based largely on the National Conference of Insurance Legislators’ model law.

Federal Activities: In December 2003, H.R. 2622, the Fair and Accurate Credit Transactions Act of 2003 (FACTA) was signed into law, permanently reauthorizing the expiring Fair Credit Reporting Act. FCRA was first created in 1970 and amended in 1996. The law preempts state privacy laws, some of which are more stringent than the federal law. Banks, insurers and others who use credit information can now work under a uniform set of federal rules. The law gives consumers new fraud and identify-theft protections. It allows them to opt out of information sharing among affiliates if the purpose of the sharing is for marketing. The law also entitles one free credit report a year upon request from the three major credit reporting agencies, Equifax, Experian and TransUnion. Consumers can obtain their free reports from http://www.annualcreditreport.com , a service funded by the three agencies.

The law directed the Federal Trade Commission (FTC) to conduct a study on the use of credit information by financial services companies, including insurers’ use of insurance scoring. The FTC was required to consult with the Office of Fair Housing and Urban Development, part of the Department of Housing and Urban Development, in researching this issue. The study will evaluate whether the use of credit information has an effect on the affordability and availability of financial services products, including the degree to which it may have a “disparate impact” on various demographic groups. The bill requires the FTC to make recommendations for legislative or administrative actions.

© Insurance Information Institute, Inc. - ALL RIGHTS RESERVED.