© 2009 National Association of Insurance Commissioners
AGENDA
PROPERTY AND CASUALTY INSURANCE (C) COMMITTEE
MARKET REGULATION AND CONSUMER AFFAIRS (D) COMMITTEE
THE USE OF CREDIT-BASED INSURANCE SCORES
PUBLIC FACT-FINDING HEARING AGENDA
TIME TOPIC
8:45 am-9:00 am
Introductory Remarks from the Chairs
9:00 am-9:40 am
How Credit Scores are Developed and Used
Panel Discussion with the Credit Reporting Agencies
Panelists
• Chet Wiermanski, TransUnion
• Eric J. Ellman, Consumer Data Industry Association
• Jon Burton, LexisNexis
• Lamont Boyd, FICO CBIS
• Birny Birnbaum, Center for Economic Justice
9:40 am-10:30 am
Member Questions to Credit Reporting Agencies
10:30 am-10:45 am Break
10:45 am-11:15 am
Data Quality in Credit Reports and Actuarial Standards
Panel Discussion with Actuarial Representatives
Panelists
• Jeff Kucera, American Academy of Actuaries
• Robert P. Hartwig, Insurance Information Institute
• Mike Miller, Actuary
• Birny Birnbaum, Center for Economic Justice
• J. Robert Hunter, Consumer Federation of America
11:15 am-12:00 pm
Member Questions to Actuarial Representatives
12:00 pm-12:45 pm
Lunch
12:45 pm-1:15 pm
How Insurers Develop and Use Credit-Based Insurance Scores
Panel Discussion with Insurer Representatives and Insurance Producers
Panelists
• Dave Snyder, American Insurance Association;
• Alex Hageli, Property Casualty Insurers Association of America
• Neil Alldredge, National Association of Mutual Insurance Companies
• Charles Neeson, Westfield Insurance
• Wesley Bissett, Independent Insurance Agents and Brokers of America
1:15 pm-2:00 pm
Member Questions to Insurer Representatives and Insurance Producers
2:00 pm-2:30 pm
Consumer Perspectives on the Use of Credit-Based Insurance Scores
Panel Discussion with Consumer Representatives
• J Robert Hunter, Consumer Federation of America
• Birny Birnbaum, Center for Economic Justice
• Pat Butler, NOW
• Gregory Squires, George Washington University
2:30 pm-3:15 pm
Member Questions to Consumer Representatives
3:15 pm-3:30 pm Break
3:30 pm-4:30 pm
Open Discussion on Various Issues
Impact of the Current Economy on Credit Scores
Reliability of Credit Reports
Fairness to Consumers
Correlation v. Causation
4:30 pm-5:00 pm
Concluding Remarks and Next Steps
THE USE OF CREDIT-BASED INSURANCE SCORES
PUBLIC FACT-FINDING HEARING
April 30, 2009
Arlington, Virginia
Written Testimony Submitted April 27, 2009:
• Chet Wiermanski, TransUnion
• Eric J. Ellman, Consumer Data Industry Association
• Jon Burton, LexisNexis
• Lamont Boyd, FICO CBIS
• Birny Birnbaum, Center for Economic Justice
• Jeff Kucera, American Academy of Actuaries
• Robert P. Hartwig, Insurance Information Institute
• Mike Miller, Actuary
• Birny Birnbaum, Center for Economic Justice
• J. Robert Hunter, Consumer Federation of America
• Dave Snyder, American Insurance Association
• Alex Hageli, Property Casualty Insurers Association of America
• Neil Alldredge, National Association of Mutual Insurance Companies
• Charles Neeson, Westfield Insurance
• Wesley Bissett, Independent Insurance Agents and Brokers of America
• J Robert Hunter, Consumer Federation of America
• Pat Butler, NOW
• Gregory Squires, George Washington University
• Commissioner Karen Weldin Stewart, Delaware Department of Insurance
• Senator James Seward, NCOIL
Testimony of
Chet Wiermanski, Group Vice President, Analytical Services, TransUnion LLC
On The Use of Credit-Based Insurance Scores
Before The
National Association of Insurance Commissioners
Property and Casualty Insurance (C) Committee
Market Regulation and Consumer Affairs (D) Committee
April 30, 2009
Director McRaith, Commissioner Holland and distinguished committee members, thank you for
providing TransUnion with the opportunity to speak with you today at this important hearing on
credit-based insurance scores. We appreciate your leadership on the insurance scoring issue,
this open forum, and your objectivity surrounding the often studied, but frequently
misunderstood topic of credit-based insurance scores. I hope to shed some light on two of your
objectives for this hearing: to understand the development of credit-based insurance scores and
the trending of these scores in light of the current economic recession.
By way of background, TransUnion is a global leader in credit and information management and
one of three global consumer credit reporting companies (CRAs). Headquartered in Chicago,
TransUnion provides objective credit reports and credit-based insurance scores to insurers. As
such, we do not determine rates or premiums, nor do we accept or reject applicants or
policyholders. Our credit-based TransUnion Insurance Risk Scores (TUIRS) are used by
insurers across the country. The security and accuracy of our information are our highest
priorities.
Testimony of Chet Wiermanski, TransUnion LLC April 30, 2009
2
TransUnion Insurance Risk Scores are developed to be completely transparent at all levels of
the policy cycle. Thus, agents and consumers have a clear understanding of the credit
characteristics impacting each score and how scores may potentially be improved. With
each TUIRS adverse action reason code message, we provide an explanation detailing why the
score is less than ideal. All characteristics and algorithms used to create TUIRS are available
upon request, providing a clearer understanding of all the credit elements that impact a
consumer’s insurance score.
TransUnion Insurance Risk Scores are based exclusively on objective, factual, accurate credit
report information, including consumer accounts such as credit cards, retail store cards,
mortgages, and auto loans. Also included in our scores is public record information, including
bankruptcies, liens and judgments, and collection accounts. Additionally, TUIRS takes into
consideration consumer initiated inquiries associated with their request for new credit accounts.
Multiple consumer generated credit inquiries associated with the shopping for a mortgage or
auto loan are deduplicated to minimize the impact on their score. All of this factual credit
information is received from tens of thousands of financial institutions, retailers, and court
houses on a monthly basis. I should also note what is not included in the credit report and or in
the calculation of a consumer’s TUIRS: medical history and records, consumer buying habits,
checking and savings information, income, or any prohibited basis characteristics identified by
the Comptroller of the Currency, which includes information regarding marital status, race, age,
religion, family status, color, receipt of public assistance, disability, gender or national origin.
It is important to note that while the term credit score is often used interchangeably by many for
credit and insurance decisioning, credit-based insurance scores and credit risk scores are not
synonymous. Credit-based risk scores are designed to predict the likelihood that an individual
will satisfactorily repay their credit obligations, while insurance scores are designed to predict
claims loss ratio. TransUnion Insurance Risk Scores were developed to meet the needs of our
insurance customers who seek a transparent, objective, and accurate predictor of consumer
insurance risk. TUIRS was developed from a pool of insurance policies collected directly from
many different insurance companies. Approximately 1.1 million consumers were analyzed,
accounting for about $741 million in premiums, with claim amounts totaling $539 million from
127,000 claims. The average loss ratio was 73% and the average premium was $650. Claims
frequency totaled .11 per consumer and the average claim was $4200.
Testimony of Chet Wiermanski, TransUnion LLC April 30, 2009
3
When developing the TransUnion Insurance Risk Score, TransUnion looked at approximately
2000 predictor candidate credit characteristics derived from consumer credit reports obtained
one year prior to establishing each consumer’s loss ratio. Using logistic regression we selected
the credit characteristics that best predicted consumer loss ratios and based upon statistical
analysis assigned the appropriate weights to each characteristic value to optimize the model’s
ability to estimate a consumer’s loss ratio. Thus, our insurance scoring models are highly
interpretable, multidimensional, consistent and objective.
TUIRS contains over seventy unique credit characteristics. Some of the credit characteristics
used in TUIRS include the number of collections within five years, percentage of all accounts
with balances greater than 50% of limit, months since oldest bankcard account has been
opened, average balance of financial installment accounts, number of previous bankruptcies,
and ratio of total balance to credit limit for all credit accounts. What is important to note is that
each credit characteristic is highly correlated to loss ratio, and this correlation has been studied
and verified by our customers, independent actuaries, state departments of insurance and
federal regulators.
TransUnion continues to study our models and their performance in light of changes in the
economic landscape. Thus, for this hearing, I will provide our perspective on two major
questions, including, the trend and volatility of credit-based insurance risk scores and how
actions taken by lenders to minimize their risk exposure are impacting credit-based insurance
risk scores.
To better understand how recent economic conditions and changes in lending practices affect
credit-based insurance scores, TransUnion analyzed a random sample of approximately 28
million consumers from each of the twelve most recent quarterly archived credit files. Each of
the different consumers sampled was scored by all three TransUnion developed credit-based
insurance risk models. In addition to appending a score from each TransUnion developed
credit-based insurance risk model, thousands of credit characteristics were also appended to
the approximately 340 million unique consumer credit reports in this analysis.
Testimony of Chet Wiermanski, TransUnion LLC April 30, 2009
4
Between the fourth quarter of 2005 and the fourth quarter of 2008 the national average
TransUnion Insurance Risk Score for each of the three proprietary TransUnion developed
insurance models exhibited a very small fluctuation. As an example, during this time period the
national average TransUnion Auto Insurance Risk Score, which ranges on a scale of 150 to
950, shifted from a low of 840.7 in the first quarter of 2006 to a high of 843.7 in the first quarter
of 2008. The most recent national average TransUnion Auto Insurance Risk Score, as reflected
of the fourth quarter of 2008, is 842.7. The national average scores for TransUnion’s Property
and combined Auto/Property Insurance Risk Scores showed similar fluctuations.
In general, the national average TransUnion Insurance Risk Scores exhibit far less fluctuation
than the national average for credit risk scores. Credit risk scores are generally more volatile
because they tend to rely more upon various forms of revolving credit utilization, recent new
account openings and recent delinquency, than TransUnion Insurance Risk Scores. Although
different aspects of utilization, account openings, and delinquency are contained within
TransUnion’s Insurance Risk Scores, these credit characteristics are defined differently and are
not weighted as heavily as TransUnion’s credit risk scores. Generally speaking TransUnion’s
Insurance Risk Scores, when compared to TransUnion’s credit risk scores, tend to place more
emphasis on credit characteristics that demonstrate a consumer’s depth of credit history as
reflected by the number and type of accounts maintained over time and a longer term view
towards account delinquency.
A recent concern regarding credit-based scoring systems, in particular insurance risk models, is
that proactive actions taken by lenders to reduce potential losses by lowering revolving credit
limits may artificially lower a consumer’s insurance score, which penalizes consumers in the
form of higher premiums and less favorable terms to the consumer. Based upon TransUnion’s
analysis it appears that from an insurance risk score perspective, the action of lowering
revolving credit limits has not played a significant role in the small fluctuations observed in the
national average for TransUnion Insurance Risk Scores. This is attributed to the manner in
which credit utilization credit characteristics are designed and weighted within TransUnion’s
Insurance Risk Scores. For example, revolving credit utilization credit characteristics are
included in credit risk models, but by themselves, they are not included in the calculation of
TransUnion Insurance Risk Scores. Based upon empirical evidence uncovered when
developing TransUnion’s Insurance Risk Scores, only a relatively few credit utilization
characteristics, of the dozens tested, were highly correlated to insurance loss ratio and
Testimony of Chet Wiermanski, TransUnion LLC April 30, 2009
5
subsequently included within the models. In addition, a majority of the credit characteristics
calculate credit utilization as a function of a consumer’s revolving credit limits combined with
original installment loan amounts. This different approach dilutes the potential impact associated
with the lowering of revolving credit limits.
As you continue to review the subject of insurance scores in this hearing and beyond, we ask
you to consider a few points:
• Credit-based TransUnion Insurance Risk Scores are completely transparent.
• TransUnion Insurance Risk scores do not use any variables that unfairly discriminate
against classes of consumers.
• A number of valid studies show a high correlation between credit data and future
insurance losses, and that credit data are highly predictive of such losses.
• Analysis of credit-based TransUnion Insurance Risk Scores shows that they are not
volatile; rather, scores are stable.
Once again, thank you for the opportunity to speak with you regarding the topic of credit and
insurance, and I stand open for any questions you may have.
1090 Vermont Avenue, NW ● Suite 200 ● Washington, DC 20005 ● Fax (202) 371-0134 ● www.cdiaonline.org
Writer’s Direct Dial: 202-408-7407
Writer’s Email: eellman@cdiaonline.org
April 27, 2009
The Honorable
Michael McRaith
Chair, NAIC Property and Casualty Insurance Committee
The Honorable Kim Holland
Chair, NAIC Market Regulation and Consumer Affairs Committee
Dear Commissioners McRaith and Holland:
I write on behalf of the Consumer Data Industry Association (CDIA) to offer comments in connection
with your review of credit-based insurance scores.
CDIA was founded in 1906 and is the international trade association that represents over 200 consumer
data companies. CDIA members represent the nation’s leading institutions in credit reporting, mortgage
reporting, check verification, fraud prevention, risk management, employment reporting, tenant screening
and collection services. Our members help their customers more effectively manage risk using precise,
current, and reliable information.
We note that the call for the April 30, hearing did not specifically seek information concerning the
underlying credit data that goes in to a credit-based insurance score. Since credit information is the
building block of a credit-based insurance score, you might be interested in CDIA’s comment.
Credit report information is heavily regulated by federal and state law, the credit report information is
proven to be reliable by those that use consumer reports, those that regulate consumer reports, and
consumers themselves. Credit information sets a solid foundation upon which credit-based insurance
scores are built.
I. Background on Insurance Scoring
Sometimes confused with credit scores that determine creditworthiness, an insurance score is designed to
measure risk of loss. Insurance scores often, but not always, contain credit information. In a credit-based
insurance score, credit information is one just one part of an overall score that also may include
information like application information, MVA/DMV data, claims history, home or auto information, and
more. Credit histories by themselves, or as part of insurance scores, are used by insurers because they are
highly predictive of risk of loss. We will let others speak more directly and greater length to the actuarial
predictiveness of insurance scores. For this comment CDIA will focus on the credit data.
II. Building a Credit Report
A credit report is like a financial biography of a consumer and for most consumers the story is
extraordinarily positive. Equifax, Experian, and TransUnion receive three billion updates every month
from 18,000 data furnishers on 200 million Americans. More than 90% of all consumers have no adverse
information on their files. For those consumers with information that is likely to be viewed as negative,
in general that information cannot stay on a credit report beyond seven years.
Credit reports will contain credit information, like car loans, credit cards, and similar items. Consumers
may also find preapproved offers of credit or insurance on their credit reports. By law, only that
consumer sees such offers, better known as inquires. No other lender, insurer, or other user is allowed to
see such non-consumer initiated offers of credit or insurance.
Just as important as what is on a credit report is what is not on a credit report. Credit reports do not
contain information about gender, race, religion, creed, color, national origin, or income.
1
Credit reports
also do not contain medical histories.
III. Credit Information and the Law
The federal Fair Credit Reporting Act (FCRA), 15 U.S.C. § 1681 et seq. heavily regulates the consumer
reporting industry, including those that furnish data to and use data from consumer reporting agencies.
The most comprehensive changes to the FCRA were in 2003 when Congress passed the Fair and Accurate
Credit Transactions Act of 2003 (FACTA or FACT Act). The FCRA, which now has over 23,000 words,
substantially controls the intake and output of consumer reporting data. Many states also have their own
credit reporting laws.
The touchstone of the FCRA is the accuracy obligation of consumer reporting agencies. The law requires
that consumer reporting agencies maintain reasonable procedures to assure maximum possible accuracy.
15 U.S.C. § 1681e(b). Determining what is accurate may sometimes be a challenge and may mean
different things to different people, but it is important to note that even the Federal Trade Commission
believes that credit reports cannot be completely error free.
2
A more thorough review of accuracy is
found in Appendix I.
In addition to the accuracy obligations imposed upon consumer reporting agencies by federal and state
law, federal law imposes restrictions on those that furnish data to consumer reporting agencies. For
example, furnishers cannot provide data they know or have reasonable cause to believe is inaccurate.
Furnishers are required to correct and update information. Id., § 1681s-2(a).
1
The committees should be aware of two studies in 2007 on credit scoring by the Federal Reserve Board and the
Federal Trade Commission. Among the Fed’s conclusions was this: “credit characteristics included in credit history
scoring models do not serve as substitutes, or proxies, for race, ethnicity, or sex”. In fact, “[c]redit scoring likely
increases the consistency and objectivity of credit evaluation and thus may help diminish the possibility that credit
decisions will be influenced by personal characteristics or other factors prohibited by law, including race or
ethnicity.” While the FTC’s study was focused on credit-based insurance scores, it too found that “[c]redit-based
insurance scores appear to have little effect as a ‘proxy’ for membership in racial and ethnic groups in decisions
related to insurance.”
2
16 C.F.R. Part 600 App. (2000). The Federal Trade Commission acknowledges that the law does not contemplate
error free consumer reports. If a consumer reporting agency accurately transcribes, stores and
communicates consumer information received from a source that it reasonably believes to be reputable and
which is credible on its face, the agency does not violate this section simply by reporting an item of
information that turns out to be inaccurate.
Consumers have a right to dispute information on their credit reports with consumer reporting agencies
and the laws require dispute resolution in not more than 30 days (45 days in certain circumstances). If a
dispute cannot be verified then the information must be removed in the consumer’s favor. Id., § 1681i.
Under the FACT Act, consumers will soon be able to file disputes directly with the data furnisher. Id., §
1681s-2(a)(8).
In 2004 the FTC reported to Congress that the FACT Act “imposed a host of new requirements that, when
fully implemented, should further enhance the accuracy and completeness of credit reports.”
3
A summary
of all of the new FACTA consumer protections, consumer reporting agency obligations, and data
furnisher obligations is available at Appendix II.
IV. Accessibility of Credit Reports
A. Credit Reports are Accessible to All Consumers
Credit reports are easy to obtain and are usually free. Consumers who wish to access their free annual
credit reports can do so by going to www.annualcreditreport.com
. In addition to the one free report per
year, per nationwide consumer reporting agency, consumers are entitled to a free credit report if they
believe they might be a fraud victim, are unemployed and seeking employment, on public assistance, or
have been denied credit or insurance on the basis of a credit report. As noted below in more detail
between 2004 and 2006, more than 52 million free credit reports were provided to consumers who
exercised their free credit report rights under the FACT Act.
4
B. Dispute Resolution is Fast and Efficient
Dispute resolution is treated uniformly and electronically via the Online Solution for Complete and
Accurate Reporting (e-OSCAR) system. This system works to resolve disputes in an even faster, more
efficient, user-friendly, and accurate manner than ever before. A majority of all disputes are resolved
within seven days or less and an additional 18% are resolved between days 8 and 14.
Consumers who disagree with the result of the dispute process are entitled by the FCRA to place a 100-
word statement on their files explaining the dispute. The GAO found that 18% of consumers disputed
information in their file and just 30% of that 18%, or 1% of those surveyed, submitted a dispute statement
to the consumer reporting agency.
5
It is important to remember that a dispute is not synonymous with an error that will lead to an adverse
result. The General Accountability Office found in 2005 that 10% of consumers who disputed
information on their credit reports disputed personal information like their name and address.
6
A dispute
must always be viewed in the context of whether that item, had it been correct, would have led to a
different credit or insurance result.
3
Federal Trade Commission, Report to Congress Under Sections 318 and 319 of the Fair and Accurate Credit
Transactions Act of 2003, Dec. 2004, vii.
4
The GAO undertook a study of financial literacy and issued a report in March 2005. The GAO found that 79% of
consumers who looked at their credit reports “felt that the information on their reports was very or somewhat easy to
understand.” General Accounting Office, Credit Reporting Literacy; Consumers Understood the Basics but Could
Benefit from Targeted Educational Efforts Child Support Enforcement; Better Data and More information on
Undistributed Collections are Needed , GAO-05-223 (March 2005), 22. Hereinafter, “2005 GAO Report”.
5
Federal Trade Commission, Board of Governors of the Federal Reserve System, Report to Congress on the Fair
Credit Reporting Act Dispute Process, (Aug. 2006), 22, 24.
6
2005 GAO Report, 30.
Another reason for consumer disputes are attempts to “credit repair” a file. Although the FTC tells
consumers that “[n]o one can remove accurate negative information from your credit report”
7
, tens of
thousands try each year to flood consumer reporting agencies through credit repair operators who promise
to assist consumers by disputing inaccurate or “unverifiable” information and trying to have accurate,
predictive derogatory data removed. Our nationwide consumer reporting agency members estimate that
on average approximately 30% of disputes filed are tied to credit repair.
V. Reliability of Credit Reports
A. Industry Standards to Ensure Maximum Accuracy
Federal law imposes accuracy standards on consumer reporting agencies and data furnishers. In addition
to legal standards, there are operational standards in place to ensure reliability. The Metro 2 data
reporting format is the data standard used by thousands of furnishers to report information to consumer
reporting agencies. Approximately 82% of all data furnished to consumer reporting agencies is sent using
the Metro 2 format; a 63% increase since 2005.
B. Economic Incentives to Ensure Maximum Reliability
In addition to legal obligations and industry standards, there is an even bigger incentive for credit reports
to be reliable. Simply put, there is a “market incentive[] to maintain and improve the accuracy and
completeness of [credit] reports.”
8
There are approximately 200 million Americans with credit reports
and credit reports are requested from Equifax, Experian, and TransUnion 27.4 million times each day. If
credit reports were not reliable they would not be used by banks, insurance companies, or others.
Furthermore, the American economy is a credit-based economy. The consumer reporting industry is the
foundation upon which that economy is built and because of that importance, credit reports must be
reliable and predictable.
C. Debunking the Public Interest Reports
Often cited to perpetuate the myth of inaccuracies are reports issued by the U.S. Public Interest Research
Group (PIRG) and Consumers Union (CU). The first PIRG report, issued in 1998, reviewed 133 files of
88 people (out of 200 million Americans with credit histories). The second PIRG report in 2004
reviewed the credit reports of 154 people, most of whom were PIRG members or staffers. The sample
sizes were not representative of the population nor were the conclusions drawn statistically sound. For
example, PIRG did not seek the input of creditors with regard to likelihood of an adverse credit decision,
and based its conclusions on its own staffs’ opinions as to who would or would not receive credit.
Consumers Union’s report was based on its asking 57 employees and their relatives to obtain their credit
reports and identify anything they thought was wrong, regardless of whether it might actually impact the
credit decision and again based on the consumers’ own conclusions.
The Federal Trade Commission reviewed the PIRG and CU reports and found that not only have
“questions…been raised about the sample size and representativeness of the samples”, but neither of these
“relied on the participation of all of the…key stakeholders in the credit reporting process.”
9
7
“Credit Repair: How to Help Youself” http://www.ftc.gov/bcp/edu/pubs/consumer/credit/cre13.shtm (viewed April
20, 2009).
8
Federal Trade Commission, Report to Congress Under Sections 318 and 319 of the Fair and Accurate Credit
Transactions Act of 2003, Dec. 2004, 7.
9
Federal Trade Commission, Report to Congress Under Sections 318 and 319 of the Fair and Accurate Credit
Transactions Act of 2003, Dec. 2004, iii.
The General Accounting Office reviewed available literature on perceived inaccuracies in consumer
reports and concluded that
the studies did not use a statistically representative methodology, examining on the credit files of
their employees who verified the accuracy of the information, and counted any error as an
inaccuracy regardless of the potential impact. Similarly, these studies use varying definitions in
identifying errors, and providing some obscure explanations of how they carried out their work.
10
D. What the Regulators Say About Reliability
The Federal Reserve has reviewed the reliability of consumer reports and made several observations.
First, the Federal Reserve, which looked at over 300,000 credit reports, noted that
Overall, research and creditor experience has consistently indicated that credit reporting company
information, despite any limitations that it may have, generally provides an effective measure of
the relative credit risk posed by prospective borrowers.
11
The report also noted that
Available evidence indicates that these data and the credit-scoring models derived from them
have substantially improved the overall quality of credit decisions and have reduced the costs of
such decision-making. Almost certainly, consumers would receive less credit and the price of the
credit they received would be higher, if not for the information provided by credit reporting
companies.
12
In the context of the use of credit for insurance purposes, the 1997 NAIC White Paper Credit Reports and
Insurance Underwriting cited the FTC. While the White Paper stated that “…various studies have
indicated different results of the accuracy of credit reports,” it went on to state that
[a] representative of the FTC, speaking to regulators on October 26, 1995, stated that the FTC is
only able to estimate the accuracy of credit reports based upon the volume of complaints it
receives. The number of complaints has been decreasing, thus the FTC assumes that the accuracy
of credit report information is improving.
Indeed. Between 2004 and 2006, more than 52 million free credit reports were provided to consumers
who exercised their free credit report rights under the FACT Act. Just 10% (520,000) of consumers had
questions about that report or filed a dispute. Of the 10% who filed a dispute, just 1.98% (102,960) of
disputes resulted in a deletion of data.
According to the Wall Street Journal
of July 26, 2002: “The Federal Trade Commission, responsible for
monitoring the activities of both the bureaus and data providers under the Fair Credit Reporting Act, says
the different players in the credit-reporting process overall do a good job of making sure credit reports are
accurate.”
10
General Accounting Office, Consumer Credit – Limited Information Exists on Extent of Credit Report Errors and
Their Implications for Consumers, GAO-03-1036T (July 31, 2003), 9-10.
11
An Overview of Consumer Data and Consumer Reporting, Federal Reserve Bulletin, Feb. 2003, 50-51 (citations
omitted); See also, Credit Reporting Accuracy and Access to Credit, Federal Reserve Bulletin, Summer 2004, 320.
12
Id., 70 (citations omitted); See also, Credit Reporting Accuracy and Access to Credit, Federal Reserve Bulletin,
Summer 2004, 320.
E. What the Users of Consumer Reports Say About Reliability.
In 2001, Allstate ordered over 17 million credit reports. The number of written requests from
consumers disputing information on their credit report totaled less then 3,000, or .017 percent of
the total number of reports ordered. Of that small number, only some of the disputes were
legitimate. Of the number of legitimate disputes, only some would have any bearing on the
insurance score because we only look at certain characteristics. Of the number affecting the
insurance score, only some would affect the discount amount because the score must change by a
certain amount to move into another discount category. Thus, the number of inaccurate credit
reports that affect the premium charged is at most a subset of a subset of a subset of .017%.
13
Conclusion
The use of credit report information for insurance purposes is lawful and heavily regulated, commercially
accepted by businesses and consumers, and statistically proven. Credit reports are reliable because the
law requires it, industry tests it, and the economy demands it. Credit information sets a solid foundation
upon which credit-based insurance scores are built.
Sincerely,
Eric J. Ellman
Vice President, Public Policy and Legal Affairs
Attachments
13
Allstate Insurance Company’s Additional Written Testimony: Allstate’s Use of Credit Scoring, before the
Michigan Office of Financial and Insurance Services, July 23, 2002.
How has the Economy Impacted
Credit Models in Insurance
NAIC Public Fact-Finding Hearing
The Use of Credit-Based Insurance Scores
1
2
Agenda
Credit Models 101
Do We See a Shift in Credit Score Changes?
Credit Trends
3
Credit Models 101
• Insurance Models ≠ Financial Models
Insurance Models are developed on
historical insurance losses
Insurance Score rank order loss
propensity
Financial Models are developed
on bad debt/deliquencies
Financial Scores rank order
credit “bads”
Credit Models 101
Insurance Models use attributes to rank
order loss propensity
Credit Seeking Behavior-inquiries
Account Age-length of time account is open
Credit Utilization-ratio of balance to limit
Payment Behavior-delinquencies
Derogatory Public Records- # of
bankruptcies/foreclosures
4
5
Higher Scores indicate better risks
Credit Score Naturally shift over time
Average scores shifted between -1.0% and 1.0% annually between 2004 and 2007
6
The average Attract Home Score was flat from 2007 to 2008
The average Attract Auto Score actually rose between 2007 and 2008
7
Higher Index=lower score=higher risk=higher premiums
No significant change has occurred from 2007 to 2008
8
Average Account Age
Higher Index=lower scores=lower risk=lower premiums
Avg Age is increasing from 2007 to 2008 due to credit tightening
Average Inquiries
Lower Index=lower scores=lower risk=lower premiums
Avg Inquiries are significantly decreasing from 2007 to 2008 due to credit
tightening
All Segments are trending consistently with overall trends
Consumer Segment Credit Trends
9
Insurance credit based scores continue to remain stable over time
Carriers do not use credit in isolation of other factors when setting rates
The current credit crisis is affecting consumer credit positively and negatively
Increased delinquency, foreclosures and bankruptcies
Decreased inquiries, balances and overall debt burden
Insurance credit based scores continue to rank order loss propensity
Reducing the information available to insurance companies for measuring risk
Forces carriers to loosen their underwriting standards
Run the very real risk of not properly pricing their products
The end result is deficient capital for paying claims.
Summary
10
• Contact: Jon Burton
Senior Director, Government Relations
678-694-3383
11
April 24, 2009
Director Michael McRaith, (IL) Chair
The Property and Casualty Insurance (C) Committee
Commissioner Kim Holland (OK) Chair
The Market Regulation and Consumer Affairs (D) Committee
National Association of Insurance Commissioners
444 N. Capitol Street, NW, Suite 701
Washington, DC 20001-1509
(Via Email – Eric Nordman, [email protected]
and Pam Simpson, [email protected])
RE: NAIC Public Hearing on Credit-Based Insurance Scores – April 30, 2009
Dear Director McRaith, Commissioner Holland,
Fair Isaac Corporation (FICO™) is pleased to offer the following comments in response to two questions
asked by your Committees in their hearing announcement:
• What constitutes a credit-based insurance score?
• How have current economic conditions affected credit-based insurance scores?
What is a credit-based insurance score?
FICO
®
Credit-Based Insurance Score models, introduced in 1993, evaluate credit data available on a
consumer’s credit report to produce a score that indicates the risk of an auto or home loss (measured by
loss ratio relativity) for new applicants and existing policyholders.
These models are mathematical
algorithms that use legally acceptable data to predict future behavior. FICO CBIS models are updated as
required to meet statutory and/or regulatory requirements in each state and are evaluated regularly to
ensure continuing predictive value.
It is important to understand the difference between FICO
®
Credit-Based Insurance Scores and FICO
®
Credit Risk Scores.
FICO CBIS scores predict likely future insurance loss ratio relativities, while FICO
Credit Risk Scores predict the likelihood of future serious delinquencies or defaults on credit obligations.
While our analytic and model development techniques are similar, the models are developed to predict
completely different outcomes and, as such, the influence of different credit variables can vary greatly.
FICO CBIS scores generally range from the 100s to the 900s, with the higher the score, the lower the
likely loss ratio relativity and the better the risk performance.
We believe our CBIS scores are effectively
used by insurers in combination with other relevant factors for underwriting and pricing decisions.
Each
insurer determines how best to implement its use of CBIS scores to match its market objectives, based on
the competitive environment as well as state statutes and regulations.
As was revealed by the Federal Trade Commission Report to Congress on Credit-Based Insurance
Scores (July 2007), among many other independent studies, insurers using CBIS scores are able to
Director Michael McRaith (IL) Chair and Commissioner Kim Holland (OK) Chair
April 24, 2009
Page 2
effectively offer premium discounting to the majority of consumers.
By relying on predictive, objective and
consistent risk segmentation provided by CBIS scores, in concert with other key risk variables, the
industry has expanded the availability and affordability of auto and home insurance coverage to
consumers in all markets.
While FICO develops and maintains our CBIS models, the models are programmed into the processing
systems of our partners, the consumer reporting agencies (CRAs), where the consumer credit data
resides.
FICO CBIS models are regularly audited by FICO and the CBIS scores are available to insurers
from the CRAs as:
• InScore
®
via Equifax
• Experian/Fair Isaac Insurance Scores via Experian (delivered by ChoicePoint)
• Fair Isaac
®
Insurance Risk Scores (formerly known as ASSIST) via TransUnion
To develop our CBIS models, FICO follows rigorous statistical methodology and gathers depersonalized
credit data on millions of consumers and multi-millions of dollars in insurance premiums and losses. In the
model development process, advanced proprietary technology is used to empirically determine the
correlation of hundreds of credit variables with subsequent claim performance.
Those credit variables
found to be most predictive of future losses are used to build the models.
As the FTC’s 2007 Report to Congress and several independent studies have shown, the CBIS models
are demonstrably efficient and accurate in predicting insurance losses. Without credit-based insurance
scores, the risk selection process would likely be less objective, less consistent, and less accurate and
we believe a majority of consumers—most of whom are good insurance risks—would have to subsidize
those consumers whose greater level of risk is not able to be fully and effectively considered.
How have current economic conditions affected credit-based insurance scores?
In spite of the current economic climate, recent analysis of FICO CBIS scoring models shows that
average CBIS scores have remained virtually the same over time for the general population (see
attachments). This is especially noteworthy when the number of people who are delinquent in repaying
creditors has grown in recent months. We believe the overall stability of FICO CBIS scores may be
caused by a greater number of consumers becoming even more credit conscious—making certain to pay
all bills on time, paying down outstanding balances, and perhaps not seeking additional credit obligations.
While a small but growing number of consumers have experienced recent financial hardships, such as
mortgage foreclosures, it is impossible to generalize about the impact of such an event on an individual’s
credit-based insurance score.
FICO CBIS models consider the interrelationship of all credit information in
a consumer’s credit report, including any foreclosure information. Scores may change when lenders
reduce credit limits, but FICO CBIS scores assess a wide variety of data on credit reports, so the impact
from a single factor like credit limit reductions will depend on what other data is on the credit report and
the amount of line reduction taken by a lender. While, in many cases, credit cardholders don’t control their
credit limits, they can control their account balances. Recent data shows that a notable number of
consumers have reduced their revolving credit usage, helping to minimize any effect from lenders
reducing their account limits.
FICO research will continue as the economic climate continues to change and as FICO CBIS score
stability is important in helping insurers make objective, consistent and accurate underwriting and pricing
decisions.
Director Michael McRaith (IL) Chair and Commissioner Kim Holland (OK) Chair
April 24, 2009
Page 3
Thank you for the opportunity to present this information. I look forward to responding to questions from
your Committees.
Sincerely,
Lamont D. Boyd, CPCU, AIM
Director, Product Management
Insurance Scoring Solutions
602-485-9858
Attachments
Birny Birnbaum
Center for Economic Justice
NAIC Hearing on Insurance Credit Scoring
April 30, 2009
Credit Reports
Issues Presented
1. What Is And Is Not In A Credit Report
2. Non-Traditional Credit Information
3. Merged Reports For Mortgage Lending, Not For Insurance
4. Errors In Credit Reports -- Errors Of Commission And Omission
5. Missing Information -- Lender Choices
6. Variation Among Credit Bureaus And Resulting Score Impacts
7. Problems/Difficulties With Fixing Errors In Credit Reports
8. Problems With Adverse Action Notices And Explanations To Consumers
9. Credit Reports Reflect And Perpetuate Historical Inequities
1
NAIC PUBLIC HEARING ON CREDIT-BASED INSURANCE SCORES
APRIL 30, 2009
My name is Jeff Kucera. I am here today representing the Casualty Practice Council of the
American Academy of Actuaries.
1
I am employed as a senior consultant with EMB America
LLC, an actuarial consulting firm. I am a fellow of the Casualty Actuarial Society and a member
of the American Academy of Actuaries. I will be addressing actuarial practice applicable to risk
classification and specifically, the use of credit-based insurance scores for rating and
underwriting purposes. I am also here to offer the assistance of the Casualty Practice Council in
your continued exploration of credit-based insurance scores.
In particular, my comments will demonstrate that the use of credit-based insurance scores allows
the insurer to better segment insurance risks for the purpose of charging appropriate rates. I will
address the following items:
• Current economic circumstances;
• Definition of what constitutes a credit-based insurance score;
• Evaluation of how insurers use credit-based insurance scores; and
• Discussion of how current economic conditions have affected policyholder premiums
related to credit-based insurance scores.
Most companies now use credit-based insurance scores in the rating of personal lines such as
private-passenger automobile or homeowners’ insurance. The use of credit-based insurance
scores helps insurance companies charge those risks that are likely to generate greater costs
higher premiums, while those likely to generate lower costs get lower premiums. The removal
of such insurance scores will not lower overall insurance premium; rather, it will redistribute the
premium charges so that those risks with lower expected costs will pay more than is actuarially
fair, while those with greater expected costs will pay less than is actuarially fair.
1
The American Academy of Actuaries is a 16,000-member professional association whose mission is to serve the
public on behalf of the U.S. actuarial profession. The Academy assists public policymakers on all levels by
providing leadership, objective expertise, and actuarial advice on risk and financial security issues. The Academy
also sets qualification, practice, and professionalism standards for actuaries in the United States.
2
Current Economic Circumstances
As we are all aware, the United States is suffering from a major economic crisis, which has
imposed considerable hardship on both individuals and businesses. A significant aspect of the
current economic crisis is the severe tightening of the credit markets. This may suggest that
credit standards are being tightened by banks and other sources of commercial credit. This
comes at a time when increasing numbers of Americans are experiencing loss of income,
including decreases in the value of many of their assets and unemployment. These problems are
significant and ongoing, and they raise questions regarding the use of credit rating in insurance.
These issues span multiple lines of insurance, but for individuals, they have the greatest impact
on private-passenger auto and homeowners’ insurance.
The American Academy of Actuaries is the public policy organization for actuaries practicing in
all specialties within the United States. A major purpose of the Academy is to act as the voice of
the profession on public policy issues. The Academy regularly prepares testimony for
Congress, provides information to federal elected officials, comments on proposed federal
regulations, and works closely with state officials on issues related to insurance.
The purpose of my presentation on behalf of the Casualty Practice Council today is to assist the
NAIC in its analysis of these questions and to offer to work with the NAIC in its continuing
study of these issues. The Casualty Practice Council has a history of working with the NAIC on
this and many other topics. In fact, the Risk Classification Subcommittee of the Academy’s
Products, Pricing, and Market Committee presented the NAIC with a report, “The Use of Credit
History for Personal Lines of Insurance,”
2
in November 2002, which is still relevant today.
The NAIC has identified three issues to serve as a basis for discussion. Our comments will
provide an actuarial context for each of these issues.
Definition of What Constitutes a Credit-Based Insurance Score
An insurance score is a numerical score or ranking assigned to an insurance risk (i.e., a
prospective insured) based on that risk’s underlying characteristics. A common purpose of
insurance scoring is to generate useful information in underwriting and pricing insurance for the
individual risk being scored. The score provides a relative measure of the expected cost to the
insurance company associated with the risk.
A credit-based insurance score utilizes various attributes found in a typical individual’s credit
report. There are several different scoring models currently in use to calculate credit-based
insurance scores, including models developed by third-party vendors and proprietary models
built by individual insurance companies. The type of credit attributes generally having the
2
http://www.actuary.org/pdf/casualty/credit_dec02.pdf (last visited on Apr. 24, 2009).
3
greatest effect on an individual’s insurance score include: number of inquiries into opening new
accounts, accounts 30 days or more past due. While the attributes and relative values are not
identical for all companies, generally the higher the credit-based insurance score, the better an
individual’s credit rating.
The importance of credit-based insurance scores is that there is a strong correlation between
them and the expected costs associated with the risk. In other words, in a group of insureds who
are identical in every other way, insureds with favorable insurance scores are significantly more
likely to have better loss experience than insureds with unfavorable insurance scores.
Consequently, credit-based insurance scores are a statistically reliable tool for segmenting risks
into different groups with different expected cost levels. This has been demonstrated in a
number of studies and reports, some of which we have listed in Appendix A.
Evaluation of How Insurers Use Credit-Based Insurance Scores
Most state insurance laws prohibit the use of insurance rates that are excessive, inadequate, or
unfairly discriminatory. Principle 4 of the Casualty Actuarial Society’s Statement of Principles
Regarding Property and Casualty Insurance Ratemaking states that, “A rate is reasonable and
not excessive, inadequate, or unfairly discriminatory if it is an actuarially sound estimate of the
expected value of all future costs associated with an individual risk transfer.”
3
Thus, the overall
average rate level should be set so that the total premium collected from all risks is sufficient to
cover the total expected costs. Additionally, the individuals’ rates should be set such that the
premium collected from each individual risk, or group of similar risks, reflects the expected costs
for that individual risk (or group of similar risks).
In a 2001 survey, 90 percent of the responding insurers (from the top 100 personal lines
companies) indicated that they were using credit data.
4
According to the survey, the use of credit
data is a relatively recent trend; more than half of the responding insurers using credit said that
they began using credit in 1998 or later. Today, the number of companies using credit is likely
even greater. Some insurers use insurance scores simply to determine whether a prospective
insured qualifies to be written by the company. More typically, insurers also use insurance
scores to help segment risks into different groups with similar expected costs for the purpose of
rating. In such cases, the insurer may use the insurance score directly as a rating factor, also
called a “risk classification factor,” similar to an amount of insurance for homeowners’ insurance
or prior violations for private-passenger auto insurance. Alternatively, an insurer with multiple
“tiers” representing different levels of expected cost may use the insurance score to help assign
risks to the appropriate tier. Whether insurance scores are being used as a risk classification or
3
http://www.casact.org/standards/princip/sppcrate.pdf (last visited on Apr. 22, 2009), Statement of Principles
Regarding Property and Casualty Insurance Ratemaking, Casualty Actuarial Society, May 1988.
4
“Insurance Scoring in Personal Automobile Insurance—Breaking the Silence,” Conning & Company, 2001.
4
tiering factor, the impact is the same: insurance scores are being used to segment risks into
homogenous groups so that appropriate premiums can be charged.
With respect to insurance scores as a risk classification or tiering factor, the actuary is guided by
Actuarial Standard of Practice (ASOP) No. 12, Risk Classification.
5
Rating plans for individual
lines of insurance generally include several different risk classifications. For example, private-
passenger auto lines use such risk classifications as the make and model of the car, age of the
driver, prior traffic violations and accidents, etc. For homeowners’ insurance, examples of risk
classification include amount of insurance, type of home construction, prior loss history, etc.
The key section of ASOP No. 12 that is applicable to the use of insurance scores is section 3.2.1.,
which reads in part as follows:
Relationship of Risk Characteristics and Expected Outcomes—The actuary should select risk
characteristics that are related to expected outcomes. A relationship between a risk
characteristic and an expected outcome, such as cost, is demonstrated if it can be shown that
the variation in actual or reasonably anticipated experience correlates to the risk
characteristic. In demonstrating a relationship, the actuary may use relevant information
from any reliable source, including statistical or other mathematical analysis of available
data. The actuary may also use clinical experience and expert opinion.
Rates within a risk classification system would be considered equitable if differences in rates
reflect material differences in expected cost for risk characteristics. In the context of rates,
the word fair is often used in place of the word equitable.
The actuary should consider the interdependence of risk characteristics. To the extent the
actuary expects the interdependence to have a material impact on the operation of the risk
classification system, the actuary should make appropriate adjustments.
The summary of articles on credit in Appendix A includes several studies that have shown that
credit scores reflect significant differences in expected loss costs. Thus, credit scores are
appropriate tools for risk differentiation. Rates based on groups differentiated by insurance score
are not excessive, inadequate, or unfairly discriminatory.
The removal of such insurance scores will not lower overall premium collected; it will only
redistribute the premium collected such that risks with lower expected costs will pay more, and
those with greater expected costs will pay less.
While the evidence may only be anecdotal, most companies report that the use of insurance
scores, along with multivariate rating and other new rating factors, have allowed them to write
more risks from the general population than before these features were introduced.
5
http://www.actuarialstandardsboard.org/pdf/asops/asop012_101.pdf (last visited on Apr. 22, 2009), Actuarial
Standard of Practice No. 12, Risk Classification (for All Practice Areas), adopted by the Actuarial Standards Board,
Dec. 2005.
5
If the NAIC determines that further studies may be appropriate, the Casualty Practice Council
would be pleased to assist the NAIC in such studies.
Discussion of How Current Economic Conditions Have Affected Policyholder Premiums
Related to Credit-Based Insurance Scores
While our current economic condition is certainly on everyone’s mind, it is still uncertain exactly
how this will affect overall insurance costs and, therefore, overall insurance prices. Some
regulators or other public officials may be concerned that if the current economic crisis causes
insurance scores to worsen, it will lead to unwarranted premium increases. It is important to
consider both the impact on the aggregate premium and on individuals’ premium.
First, it is important to consider the impact on the aggregate premium. Insurers use insurance
scores to determine appropriate rate relationships between risk classes, not to determine overall
premium need. Assume for a moment that insurers continue to maintain the same rate
relationships for different insurance score ranges, and that the current economic crisis causes
every insureds’ insurance score to worsen. The actuary would observe this distributional shift or
change and adjust overall rate levels so that the total premium collected by the insurance
company remains the same and the integrity of the rate relationships among risks remains intact.
This is no different than any other distributional shift, such as an increase in the average value of
homes, which an actuary has to consider when setting the overall rate level. Part of a typical
actuarial rate review is an analysis of any shifts in distributions that affect the premium level.
The actuary would adjust for these shifts in determining appropriate future rates. As a result of
this standard ratemaking practice, any shift in insurance scores due to the current adverse
economic conditions will not result in any long-term impact on overall premium collected.
Second, it is important to consider the impact on the individuals’ premium.
6
As stated earlier,
studies have demonstrated that insurance scores are an effective means of segmenting risks.
Because of this, many companies now vary the rates charged to risks with different insurance
scores. Some regulators or other public officials may be concerned that a dramatic shift in credit
scores could disrupt the current relative rates among risks with insurance scores; in other words,
perhaps the difference in expected cost levels among insureds with favorable and unfavorable
scores will be less significant.
This, too, is not a problem that is unique to insurance scores. The gender and age of drivers have
long been recognized as important rating characteristics for personal automobile insurance.
There have been, and still are, very significant differences between the rates charged to young
6
It is important to remember that any distribution shift is likely to have a smaller effect on renewal business than on
new business, because some states and/or companies only permit the use of such scores for renewals if it results in a
more favorable rate for the individual insured.
6
males and young females, reflecting the higher cost of auto insurance for young male drivers
compared to young female drivers. However, over time, the driving habits of young males and
young females have become more similar, and while the difference in risk is still significant, it is
not nearly as large as it was in the past. As this trend has developed, insurers adjusted
classification plans to reduce the rate differentials to reflect it. If the actuary regularly analyzes
the indicated rate differentials for different insurance score ranges, the rate differentials will be
changed if more recent data suggests it. This potential shift in group differentials, and
motivation or intent to be competitive, provide incentives for companies to regularly review their
rate differences.
One of the other roles of an actuary is to regularly review the data to decide whether the overall
average rate level is appropriate and whether the rate differentials for risks with different
insurance scores need to be adjusted. By doing this, the actuary can ensure that the rates are
actuarially sound,
7
regardless of the effect the current economic crisis has on personal insurance
scores.
It is possible that a sudden or immediate distribution shift could result from the current economic
conditions, and that, by the time it works its way into the actuary’s data, many insureds will have
already been harmed. While we have been suffering through the current economic conditions for
approximately six months, we are unaware of any quantifiable evidence that has surfaced to
demonstrate that such a dramatic shift has been occurring. It is our opinion, based on anecdotal
evidence, that any shift thus far has been minor. This could be because renewal business, which
makes up the majority of any company’s business, is less likely to be affected by a shift.
Ascertaining whether an actual shift of any significance has occurred would require a study to
look at the distribution of insurance scores of several companies over a period of time. The
Casualty Practice Council is willing to assist the NAIC should it decide to pursue such a study.
On behalf of the Academy and the Casualty Practice Council, I thank you for the opportunity to
speak to you today. To the extent that we can further assist the NAIC in its endeavors on this
topic, the Casualty Practice Council volunteers its services. We look forward to working with
you.
If time permits, I am happy to answer any questions you may have.
7
http://www.casact.org/standards/princip/sppcrate.pdf (last visited on Apr. 22, 2009), Statement of Principles
Regarding Property and Casualty Insurance Ratemaking, Casualty Actuarial Society, May 1988.
7
Appendix A – Summary of Additional Articles on Credit Scoring
Several studies have already been conducted on the use of credit for rating and underwriting for
both homeowners’ and private-passenger auto insurance. In particular, the following studies
may warrant review:
• Predictiveness of Credit History for Insurance Loss Ratio Relativities by Isaac Fair,
(1999).
• Use of Credit Reports in Underwriting by the Commonwealth of Virginia, State
Corporation Committee, Bureau of Insurance (1999).
• The Impact of Personal Insurance Credit History on Loss Performance in Personal Lines
by James D. Monaghan (2000).
• Insurance Scoring in Personal Automobile Insurance – Breaking the Silence by Conning
& Company (2001).
• Use of Credit Information by Insurers in Texas by the Texas Department of Insurance
(December 2004).
• Use of Credit Information by Insurers in Texas – the Multivariate Analysis by the Texas
Department of Insurance (January 2005).
• Credit-Based Insurance Scores: Impact on Consumers of Automobile Insurance by the
Federal Trade Commission (July 2007).
• Report to the Congress on Credit Scoring by the Board of Governors of the Federal
Reserve System (2007).
PUBLIC HEARING ON CREDIT-BASED
INSURANCE SCORES
Testimony as Delivered
by
Robert P. Hartwig, Ph.D., CPCU
President & Economist
Insurance Information Institute
New York, NY
National Association of Insurance Commissioners
Property and Casualty Committee
Market Regulation and Consumer Affairs Committee
April 30, 2009
Arlington, VA
2
Thank you, Director McGraith, Commissioner Holland and members of the Committee.
Good morning. My name is Robert Hartwig and I am President and Economist for the
Insurance Information Institute, a national property/casualty insurance trade association
based in New York City.
1
I am also a Chartered Property Casualty Underwriter (CPCU)
and have worked on a wide variety of insurance issues during my 16 years in the
property/casualty insurance and reinsurance industries. Over the past decade I have
devoted considerable time and attention to various questions arising from the use of
credit information in the underwriting of personal lines insurance. I have authored
reports on the issue, made presentations before insurance, real estate and mortgage
lending groups, testified in several states, conducted agent and management training
seminars and conducted hundreds of media interviews.
The Committee has asked today’s witnesses to address three fundamental issues:
(i) An explanation of what constitute a credit-based insurance score;
(ii) An explanation of how insurers use credit-based insurance scores; and
(iii) A discussion of how current economic conditions have affected policyholder
premiums related to credit-based insurance scores.
All three are important issues. In my capacity as a professional economist with many
years of experience in the property/casualty insurance industry, I will focus my
comments today on item (iii), providing a detailed discussion on the relationship between
current economic conditions, credit scores and the cost of insurance.
Current Economic Conditions
The United States economy is currently in the midst of a deep recession which began in
December 2007. During the fourth quarter of 2008, the nation’s real gross domestic
product (GDP) declined by 6.3 percent, the largest drop in nearly 27 years (FIGURE 1).
In March 2009, the unemployment rate reached 8.5 percent, its highest level since
January 1984 and more than 4 points above the cyclical trough of 4.4 percent in March
1
Contact information: Tel: (212) 346-5520; Email: bobh@iii.org.
3
2007 [FIGURE 2]. Current projections call for the economy to continue to contract
through the first half of 2009 with a modest recovery beginning in the second half of the
year and continuing into 2010. Unemployment is expected to peak at approximately 9.6
percent in early 2010 and begin to decline thereafter.
Recessions are not uncommon, unusual or unexpected. They are an unpleasant but
unavoidable part of the business cycle, correcting many of the excesses that occur during
the expansionary phase of the cycle (such as the recent housing and credit bubbles).
Since the end of World War II, the United States economy has experienced 11 recessions
lasting an average of 6.4 months, followed by extended periods of economic growth
lasting 60.5 months on average—nearly 5 times longer than the preceding contraction
[FIGURE 3]. Thus recessions are recurrent but temporary phenomena. Growth, jobs and
incomes that are diminished during recessions always recover once the economy resumes
its long-term expansionary trajectory.
What’s Different About This Recession: Credit Market Conditions
Most recessions are triggered by a build-up of excess capacity and inventories in the
economy. Firms seek to correct these imbalances by reducing production and
investment, trimming expenses and cutting jobs. Layoffs (actual, anticipated or fear of)
cause consumer confidence to wane. As consumers—who account for two-thirds of all
spending in the economy—reduce spending, the overall economy suffers with recession
being the likely result.
This same sequence of events is present in the current recession. However, the
precipitating event was a collapse in the credit markets, which had experienced a massive
and unsustainable expansion during the middle part of this decade. This credit “bubble”
was most pronounced in the mortgage sector which grew by an astonishing 58 percent
from $9.35 trillion at year-end 2003 before peaking at $14.74 trillion in mid-2008
[FIGURE 4]. Mortgage debt outstanding began to decline during the second half of
2008. The decline in mortgage debt is part of a much broader pullback in the credit
markets.
4
The Economic Crisis and Consumer Credit Profiles
As the average American consumer pulls back on spending, contributing to the current
deep recession, they will also pull back on their use of credit financing of purchases.
Indeed, recent data from the Federal Reserve indicate that the process of consumer
“deleveraging” or debt reduction is already well underway [FIGURE 5]. Mortgage debt,
which increased at a 12.8 percent average annual pace between the first quarter of 2004
and mid-2006, declined at an average annual rate of 1.4 percent during the final three
quarters of 2008. Likewise, consumer credit fell at an average annual rate of 3.2 percent
during the fourth quarter of last year after expanding at nearly 5 percent per year on
average from 2004 through mid-2007.
The retrenchment of the American consumer is a principal contributor to the current
recession. At the same time, there is little disagreement that millions of Americans were
overextended in terms of their use of credit, especially with respect to mortgage debt.
Thus a silver lining of the current financial crisis is a change in the credit profile of the
average American household whereby outstanding debt is reduced to more manageable
levels. This should lead to an improvement in the health of the typical consumer’s (and
family’s) balance sheet. Indeed, this change is clearly already underway, as illustrated in
Figure 5. This also implies that credit scores (and credit-based insurance scores),
contrary to popular belief, are not headed uniformly downwards despite the current
recession. Likewise, credit scores do not head uniformly upwards during boom times.
Can Consumer Deleveraging and Better Credit Management Lead to Higher Scores?
The fact that many households are becoming more conservative in their use of debt and
are better managing the debt that they already have is reflected in a recent report by the
credit monitoring and management service Credit Karma [FIGURE 6]. Among their
clients, 43 percent saw their credit score rise in March 2009, 27 percent saw their scores
fall while the scores of the remaining 30 percent remained unchanged. The report clearly
illustrates that a large proportion of consumers are benefiting from rising credit scores
even in the current economic. It is also interesting to note that the 30 percent whose
5
score remained unchanged also had the highest average score, suggesting significant
stability within this group.
Does Scoring Work When the Economy is in Recession?
Credit-based insurance scores are highly accurate predictors of risk (expected loss)
irrespective of the economic environment. Over the past 10 to 15 years during which the
use of insurance scoring has become nearly universal, the economy has experienced two
recessions and enormous variations in economic growth and unemployment across states.
Insurance scores have been proven to remain highly accurate predictors throughout the
entirety of this period.
The reality is that that US economy and the economies of individual states are in a
constant state of flux. Insurance scores are robust in the sense that they remain predictive
of future loss across the entire economic cycle. Specifically, since 1995 (near the
beginning of the period when insurance scoring became more commonplace), the
national unemployment rate has ranged from a low of 3.8 percent in April 2000 to a high
of 8.5 percent in March 2009. In that month (March 2009), state unemployment rates
ranged from 4.2 percent in North Dakota to 12.8 percent in Michigan. In 2000, when the
US unemployment rate reached its low, state unemployment rates ranged from 2.3
percent in Connecticut and Virginia to 6.2 percent in Alaska. Scoring remained
continuously and highly predictive before, during and after these two cyclical extremes
for both auto and home insurance in every state in which it is used.
It should be noted that no study has ever shown insurance scores to be anything other
than highly predictive of future loss. Studies recently conducted by the Federal Reserve
and Federal Trade Commission as well as several state insurance departments confirm the
predictive accuracy across states and over time. There are no exceptions to this finding.
This implies that insurance remains strongly predictive irrespective of economic
circumstances.
6
Why Does Insurance Scoring Work Even When the Economy is in Recession?
The fact that insurance scoring is predictive of future loss irrespective of the state of the
economy is enormously critical. It underscores the fact that insurance companies
consider only those items from credit reports that have been shown to correlate with
future insurance loss potential. Unlike a lender, an insurance company is not assessing a
customer’s ability to repay a loan. Insurers are interested exclusively in those factors that
relate to future loss. Consideration of superfluous factors—factors that do not
statistically correlate with expected future loss—would be a costly waste of time.
Moreover, research conducted by credit bureau TransUnion indicates that actions taken
by lenders appear to have little impact on insurance risk scores or insurance risk indices.
Wouldn’t a Recession Push Everyone’s Credit and Insurance Score Down?
It is a common misconception that during a recession virtually consumer’s credit score
and hence insurance score will fall. As discussed previously, some credit scores rise even
during a recession, some fall and others stay the same. The same is true with insurance
scores. Examining this extreme situation is nevertheless instructive in that it reveals how
the predictive ability of credit-based insurance scores is preserved even in the case when
100 percent of the population experiences a decline in their credit score. Statistically,
such a shift in the entire population would likely have little impact on insurance rates
(except to the extent that actual loss performance deteriorates). This is because credit-
based insurance scores are used to help differentiate risk among groups with varying
degrees of loss expectancy. These groupings would still exist as would the ability of
insurers to differentiate between them statistically even if a bad economy pushed down
every member of the population’s credit and insurance scores. The argument also works
in reverse. If an improving economy raises all credit and insurance scores, the insurer’s
ability to distinguish among groups is not in any way diminished.
Credit-Based Insurance Scores: Just One of Many Factors Considered by Insurers
With so much focus on the potential impact of the economy on credit standing, it is
important to recall that insurance scores are just one of many factors insurers consider
when assessing the risk associated with a potential (or existing) policyholder. Credit-
based insurance scores are never used as the sole underwriting criterion. Auto insurance
7
premiums, for example, are based on a myriad of factors such as driving record, previous
losses, the type of car driven, miles driven and where the consumer lives. A homeowners
insurance policy premium is based on where the consumer lives, previous losses, the type
of home, and the cost to replace it, type of construction, proximity to a fire department,
among other factors.
What Are the Consequences of Bans or Severe Restrictions on the Use of Credit-
Based Insurance Scores?
Higher Premiums and Unfair Subsidies
Prohibiting insurers from using credit-based insurance scores would instantaneously
result in inherently unfair outcomes: higher rates for people with lower risk, and lower
rates for those with a higher likelihood of submitting claims. In other words, bans or
severe restrictions on insurer use of credit-based insurance scores would lead to massive
subsidies for people who impose greater costs on the system. A 2007 study by the
Federal Trade Commission found that two-thirds of consumers receive lower premiums
than they otherwise would when credit-based insurance scoring is included as a rating
factor.
Prohibiting the use of credit-based insurance score would exact a steep toll on the typical
American family. A family owning a single home and two cars and currently earning a
10 percent discount related to a strong credit would have to pay approximately $259 more
per year in the event that a ban on the use of insurance scoring forced them to forfeit their
discount.
2
It gets worse. The hard-earned $259 lost by the family with good credit and few, if any,
insurance claims will go straight into the pockets of people who file more claims and
impose greater costs on the insurance system in the form of a subsidy.
2
Based on estimated 2009 auto insurance expenditures of $875 per vehicle and average homeowners
insurance premiums of $841. Estimates are based on historical NAIC data and projected to 2009 using
Consumer Price Index data from the US Bureau of Labor Statistics.
8
There is no question that bans or severe restrictions on the use of insurance scoring lead
directly higher costs for people who impose few costs on the system and subsidies for
those impose greater costs. The aggregate subsidy is potentially very large. Focusing on
auto insurance and using Illinois as an example, a loss of an average 10 percent credit
discount on two-thirds of the 7.7 million insured vehicles on the road in that state would
result in a subsidy from good drivers to bad drivers of some $382 million. Even if the
average discount is just 5 percent, the subsidy is still $191 million.
3
Less Insurance
In addition to millions of drivers and homeowners paying more for their coverage,
insurers in some cases may simply back away from some risks. That is because the
additional predictive power of insurance scoring increases insurers’ understanding of risk
and enables them to write certain risks that in the absence of scoring would be too
uncertain to accurately price. Indeed, the same 2007 Federal Trade Commission study
that affirmed the predictive value of insurance scores as well as the economic value they
bring to consumers through discounts also concluded that the use of credit-based
insurance scores enables insurance companies to offer coverage to more consumers than
they had in the past. This effect is most pronounced for high-risk drivers, many of whom
in the era before the use of insurance scores had no other option but to seek coverage
through state-run auto plans for the highest risk drivers. Insurance scoring is one of the
primary reasons for the 60 percent drop in the number of cars insured through these
costly markets of last resort between 1996 (when 2.9 percent of drivers were insured in
these plans) and 2006 (when just 1.2 percent of drivers obtained coverage though state-
run plans). It’s not that there are fewer bad drivers today; it’s simply that insurance
scoring has helped insurers understand how to price high-risk drivers more accurately
than they could in the past.
Lessons to Be Learned from the Banking Sector’s Experience in Ignoring Risk?
It is ironic that calls to ban or severely restrict the use of credit-cased insurance scores are
being voiced during the current financial crisis. Insurers use credit-based insurance
3
The estimated subsidy of $382 million is based on 7.7 million insured vehicles in Illinois (2006 figure
from AIPSO), two-thirds of which are assumed to be eligible for credit-related discounts (per 2007 Federal
Trade Commission study) and an average auto insurance expenditure of $740 (2006 figure from NAIC).
9
scores to help them quantify and price the risk inherent in the products they sell. It is a
proven risk management tool. Property/casualty insurers’ adherence to the basic
principles of risk management (which include risk-based pricing) has allowed them to
weather the crisis better than almost any other segment of the financial services industry.
Legislation and regulation that would remove or compromise the ability to thoroughly
assess, quantify and price risk is tantamount to an endorsement of the disastrous risk
management strategies employed by many banks, whose willful ignorance of risk is at the
very core of the current financial crisis.
The experience of property/casualty insurers compared with that of banks since the
beginning of the financial crisis in 2007 could not be more stark. Whereas 50 banks have
failed (including the largest bank failure in US history), not a single property/casualty
insurance company has failed as a result of the crisis. Hundreds of those banks that
remain have had to seek billions of dollars in government bailout funds. Not a single
property/casualty insurance company has received bailout money. Throughout the
entirety of the financial crisis, not a single valid claim has gone unpaid. Moreover,
insurers continue to write new coverage, renewal existing policies and roll out new
products. Banks, on the other hand, are turning away families and businesses that need
access to credit, are scaling back their product offerings and are raising fees. The
property/casualty insurance industry even managed to earn a small profit in 2008, despite
the crisis and near-record catastrophe losses of $26 billion—the fourth highest total in
history.
Summary
Insurance scoring is a proven, accurate, objective and consistent risk assessment tool used
widely in the underwriting of auto and homeowners insurance. The data supporting its
use are statistically irrefutable, and the benefits to consumers are significant. Moreover,
the use of credit information leads directly to a fairer and equitable premium charge for
all policyholders because scoring allows premiums to be more closely matched to risk.
The current economic recession has created hardships for many American families, but
there is no evidence to suggest that insurer use of credit-based insurance scores has in any
10
systematic way increased costs for policyholders. Insurance scores have repeatedly been
proven to remain highly predictive of future loss over the course of the entire economic
cycle—recessions and expansions alike—as well as across the very diverse range of
economic environments and experiences found among the states. Importantly, insurance
scores incorporate use only those elements from credit reports that correlate with future
loss.
Finally, calls to ban, severely restrict or suspend the use of credit-cased insurance scores
during the current financial crisis are misguided, distortionary and unfair to the millions
of policyholders who impose few costs on the system. Insurance scoring is a proven risk
management tool. Property/casualty insurers’ adherence to the basic principles of risk
management has allowed them to weather the financial crisis far better than the banks,
whose inattention to risk was nothing short of a colossal failure of risk management so
profound that global financial chaos ensued. Legislation and/or regulation that would
remove or compromise the ability to thoroughly assess, quantify and price insurance risk
is tantamount to an endorsement of the disastrous risk management strategies employed
by many banks, whose willful ignorance of risk is at the very core of the current financial
crisis.
Thank you for the opportunity to appear at today’s hearing. I would be happy to answer
any questions you may have.
11
3.7%
0.8%
1.6%
2.5%
3.6%
3.1%
2.9%
0.1%
4.8%
4.8%
0.9%
2.8%
-0.5%
-2.1%
0.4%
1.6%
2.3%
2.7%
2.9%
3.1%
-5.1%
-6.3%
-0.2%
-8%
-6%
-4%
-2%
0%
2%
4%
6%
2000
2001
2002
2003
2004
2005
2006
07:1Q
07:2Q
07:3Q
07:4Q
08:1Q
08:2Q
08:3Q
08:4Q
09:1Q
09:2Q
09:3Q
09:4Q
10:1Q
10:2Q
10:3Q
10:4Q
Real GDP Growth*
*Yellow bars are Estimates/Forecasts from Blue Chip Economic Indicators.
Source: US De
p
artment of Commerce
,
Blue Economic Indicators 4/09
;
Insurance Information Institute.
Recession began in December 2007.
Economic toll of credit crunch, housing
slump, labor market contraction is growing
The Q4:2008 decline was
the steepest since the
Q1:1982 drop of 6.4%
Figure 1.
U.S. Unemployment Rate,
(
2007:Q1 to 2010:Q4F)*
4.5%
4.5%
4.6%
4.8%
4.9%
5.4%
6.1%
6.9%
8.1%
8.8%
9.3%
9.5%
9.6%
9.5%
9.4%
9.3%
4.0%
4.5%
5.0%
5.5%
6.0%
6.5%
7.0%
7.5%
8.0%
8.5%
9.0%
9.5%
10.0%
07:Q1 07:Q2 07:Q3 07:Q4 08:Q1 08:Q2 08:Q3 08:Q4 09:Q1 09:Q2 09:Q3 09:Q4 10:Q1 10:Q2 10:Q3 10:Q4
*
Blue bars are actual; Yellow bars are forecasts
Sources: US Bureau of Labor Statistics; Blue Chip Economic Indicators (4/09); Insurance Info. Inst.
Unemployment is
expected to peak near
10% in early 2010.
Figure 2.
12
Length of U.S. Business Cycles,
1929-Present*
43
13
8
11
10
8
10
11
16
6
16
88
18
50
80
37
45
39
24
106
36
58
12
92
120
73
0
10
20
30
40
50
60
70
80
90
100
110
120
Aug.
1929
May
1937
Feb.
1945
Nov.
1948
July
1953
Aug.
1957
Apr.
1960
Dec.
1969
Nov.
1973
Jan.
1980
Jul.
1981
Jul.
1990
Mar.
2001
Dec.
2007
Contraction Expansion Following
* As of May 2009, inclusive; **Post-WW II period through end of most recent expansion.
Sources: National Bureau of Economic Research; Insurance Information Institute.
Duration (Months)
Month
Recession
Started
Average Duration**
Recession = 10.4 Months
Expansion = 60.5 Months
Length of
expansions
greatly
exceeds
contractions
Figure 3.
Mortgage Debt Outstanding:
2004- 2008
1
$ Trillions
$9.35
$10.66
$12.10
$13.49
$14.71
$14.64
$14.74
$14.70
$14.57
$9.0
$10.0
$11.0
$12.0
$13.0
$14.0
$15.0
$16.0
03 04 05 06 07:Q4 08:Q1 08:Q2 08:Q3 08:Q4
1
End of period.
Sources: Board of Governors of the Federal Reserve,
http://www.federalreserve.gov/econresdata/releases/mortoutstand/
; Insurance Information Institute.
Mortgage debt
outstanding expanded
58% from year end 2003
through mid-2008, but is
now beginning to fall
4
Figure 4.
13
-4%
0%
4%
8%
12%
16%
2004:Q1
2004:Q2
2004:Q3
2004:Q4
2005:Q1
2005:Q2
2005:Q3
2005:Q4
2006:Q1
2006:Q2
2006:Q3
2006:Q4
2007:Q1
2007:Q2
2007:Q3
2007:Q4
2008:Q1
2008:Q2
2008:Q3
2008:Q4
Home Mortgage Consumer Credit
Households Are Now Rapidly
“Deleveraging” (Shedding Debt)
Percent Change in Debt Growth (Quarterly since 2004 at Annualized Rate)
Source: Federal Reserve Board, at http://www.federalreserve.gov/releases/z1/Current/z1r-2.pdf
Real Estate (Mortgage)
deleveraging
Consumer
deleveraging
As Americans begin to live more within
their means, they will strengthen their
household balance sheet and improve
their credit standing
Figure 5.
Credit Monitorin
g
Service Report:
Many Scores Rose in March 2009
Credit Score Increased
43%
Credit Score Decreased
27%
Credit Score Unchanged
30%
One consumer credit score
tracking and monitoring
service reported that more
if its customers saw their
credit score increase in
than decrease in March
2009. Those whose scores
remained unchanged had
the highest average score.
.
Source: Credit Karma U.S. Consumer Credit Score Climate Report (Press Release dated April 15, 2009) and reported in
Insurance Journal. Credit Karma is a San Francisco-based consumer credit score tracking and monitoring service.
Avg. Score
695
Avg. Score
665
Avg. Score
674
Methodology: Each month, the Credit Karma U.S. Consumer Credit
Score Climate Report compares the current credit scores of its 350,000
user base with previous scores pulled at least 30 days prior and no more
than 90 days prior to the stated month. This current report includes a
comparison of more than 30,000 Credit Karma user scores.
Figure 6.
Credit-Based Insurance Scores
As Currently Used by Personal Auto and Homeowners Insurers,
Insurance Scores Are Actuarially Sound,
Lead to More Accurate Risk Assessment,
Fairly Discriminate Between Risks,
And Work to the Advantage of a Majority of Insureds
Statement of Michael J. Miller, FCAS
Actuarial Consultant, EPIC Consulting, LLC
The Use of Credit-Based Insurance Scores
Public Hearing of NAIC Property and Casualty Insurance (C) Committee
and Market Regulation and Consumer Affairs (D) Committee
April 30, 2009
1
Introduction
My name is Michael J. Miller. I am a Fellow of the Casualty Actuarial Society and a Member
of the American Academy of Actuaries. My business address is 21253 N. 825 East Road,
Carlock, Illinois.
In order to offer a public actuarial opinion, the actuarial profession requires more of an
actuary than professional credentials. An actuary must also be experienced with the specific
subject matter and be in compliance with the profession’s continuing education requirements.
I fully comply with the actuarial profession’s experience and continuing education
requirements.
I have practiced as a professional actuary for over thirty years with a special emphasis in
ratemaking for auto and homeowners insurance. I co-authored in 2003 a major study
pertaining to credit-based insurance scores entitled “The Relationship of Credit-Based
Insurance Scores to Private Passenger Automobile Insurance Loss Propensity” (i.e., EPIC
Study). A more complete summary of my education, training, and experience is set forth in
the attached curriculum vitae (Exhibit I).
Overview
Accurate risk assessment and risk differentiation are essential when insurance is being
provided by multiple, competing insurers and the insureds are free to choose from among the
multiple providers. Reliable studies have consistently shown that credit-based insurance
scores enhance the accuracy of the risk assessment process. If the use of credit-based
insurance scores were banned the result would be insurance rates that are inadequate for
some insureds, excessive for other insureds, and unfairly discriminatory for all.
In July 2007 the Federal Trade Commission released a study of credit-based insurance
scores as used for personal automobile insurance (i.e., FTC Study). The FTC study
estimated that 59% of the national auto insured households benefit from reduced premiums
due to the use of credit-based insurance scores. According to the FTC data, credit-based
2
insurance scores benefit nearly one-half the Hispanic racial group of insureds, approximately
two-thirds of the Asian group, and over one-third of the African American group.
Although the percentage of insureds benefiting from the use of credit-based insurance
scores is lowest among African Americans, the number of African Americans that do benefit
is significant. Whether viewed as individual groups or viewed as a whole, it is clear that a
banning of credit-based insurance scores would work to the disadvantage of many minority
households throughout this country.
Concerns over the use of credit-based insurance scores arose several years ago before the
facts were fully known. We now know with certainty that credit-based insurance scores
enhance risk assessment and have a legitimate business purpose. We know that credit-
based insurance scores cannot be used to accurately predict an insured’s race or income.
We know that the use of credit-based insurance scores works to the advantage of a
significant number of racial minority households throughout this country.
What is not apparent is why, in the face of these facts, the use of credit-based insurance
scores continues to be a political issue.
Ratemaking and Estimating Loss Propensity
The essence of insurance is the transfer of risk. A commercial insurance enterprise does not
involve a sharing of losses with other insureds, unless the insurance policy includes a post-
assessment provision. An insurance consumer eliminates risk by choosing the certainty of
the insurance premium versus the uncertainty of suffering a severe financial loss.
An insurance premium is a combination of the insured’s expected loss, a provision for the
insurer’s expected operational/administrative expenses, and a provision for profit. An
insured’s expected loss is a function of the probability of a claim occurring (i.e., claim
frequency or likelihood) and the average cost of the claim once it occurs (i.e., claim severity).
For example, if an insured’s likelihood of an auto collision claim is 10% per year and the
average cost of a collision claim is expected to be $1,000, the insured’s expected loss is
$100 per year (i.e., 10% x $1,000). Another insured with a claim likelihood of 12% and an
3
expected claim cost of $1,000 would have an expected loss of $120 per year (i.e., 12% x
$1,000). Since the expected loss is part of the calculated rate, the insurance premium
charge for this second insured would be higher than the premium charge for the first insured
because the expected loss for the second insured is higher.
An insured’s expected loss is estimated based on a combination of several risk
characteristics, or risk factors. Each risk factor has been found to measure and predict at
least a portion of the total risk associated with each insured. For private passenger auto
insurance, where the car is garaged and principally operated has been statistically shown to
affect both the likelihood of claim occurrence and the cost of claims. Other important risk
factors that are statistically correlated to the risk of auto insurance claims include age,
gender, marital status, and driving record of the drivers; annual mileage and how the car is
used (i.e., pleasure, commuting, or business); and the make and model of the car.
For homeowners insurance, the risk factors that are commonly used to estimate the
expected loss include the estimated replacement cost of the house, the construction type of
the house, the geographical location, and the age of the utilities.
No single risk factor has been found that measures or predicts the total risk. Typically,
insurers rely on twenty or more risk factors to accurately estimate an insured’s likelihood of a
claim and the expected loss. All risk factors work in combination to measure and predict the
total risk.
The EPIC Study of 2003 which I co-authored showed, and the FTC Study of 2007 confirmed,
that credit-based insurance scores for personal auto insurance are strongly related to an
insured’s likelihood of claim occurrence and add significant accuracy to the risk assessment
process. In other words, credit-based insurance scores measure risk not previously
measured by other known rate factors. The strength of the statistical correlation is such that
a credit-based insurance score is among the most important risk factors used by insurers to
accurately estimate the probability of claim occurrence.
The EPIC Study found that a credit-based insurance score was among the top three most
important risk factors for each of the four major auto insurance coverages. No researcher
has yet been able to find an alternative risk factor that could replace a credit-based insurance
4
score as a predictor of claim likelihood without sacrificing a great deal of accuracy in the risk
assessment process.
Causation
Sometimes critics of the insurance industry complain that it is inappropriate to use a risk
factor, such as a credit-based insurance score, because it does not “cause” an insured to
have an auto accident or “cause” a homeowners insurance claim. If causation were a
standard for the use of any specific risk factor, there would be no risk factors that could be
used to predict and measure risk.
While understanding the causes of auto claim losses (e.g., inattention to driving, driving too
fast, following too closely, etc.) and understanding the causes of homeowners claim losses
(e.g., lightning strikes, wind storms, faulty wiring, ruptured washer supply hoses, etc.) may be
of interest when attempting to reduce losses, non-causal factors are the more practical and
powerful predictors of the probability of an insurance loss.
The classical example of a relationship to loss that is not a cause-and-effect relationship is a
home built in a river valley. Living in a river valley does not “cause” a flood. But there is a
predictive relationship between the risk of a flood loss and the construction of a home in a
flood plain. It would be foolish to presume there is no risk of a flood loss merely because the
location of the home does not “cause” the flood.
Many other examples of risk factors can be cited that do not cause accidents to occur.
Neither past traffic violations nor past accidents “cause” future insurance losses, but there is
a predictive relationship between past driving records and future losses. An age of the driver
is not the cause of an accident, but it is predictive of the likelihood of a future accident. No
predictive risk factor used in the risk assessment process can be said to actually cause an
auto accident.
Causality should not be the basis for allowing or disallowing the use of credit-based
insurance scores, just as it should not be the basis for allowing or disallowing all other risk
factors. The basis for allowing the use of any risk factor should be the ability of the risk factor
5
to significantly contribute to the accurate measurement of the propensity for insurance
losses.
It has long been a tenet of risk assessment that financial stability/responsibility was related to
risk for private passenger automobile insurance. However, the concepts of financial stability
and responsibility have been heretofore difficult to translate into objective, measurable risk
factors. Credit-based insurance scores offer, for the first time, the means of objectively
measuring the relationship between financial prudence and the propensity for insurance
losses.
While it would be inconsistent with sound actuarial principles to require credit-based
insurance scores to demonstrate a causal relationship, we could reasonably speculate that
there are psychological factors that likely affect our adversity to risk and how we manage our
personal lives. We could reasonably speculate that the results of these psychological
tendencies can be observed in many aspects of our personal lives, including our credit
history and insurance losses. Insurance scores may be providing an objective means of
measuring personal responsibility and its effect on insurance losses, even though we may
never fully understand the psychology involved.
Rather than speculate on what credit-based insurance scores are actually measuring, I
would prefer to rely on the statistics. I fully recognize that statistical correlations can be
spurious. Math statistical textbooks typically include warnings to math students that
statistical correlations can be spurious. However, in the case of insurance scores there have
been several studies published with analysts working independently, using different
databases. All of the studies indicate the same conclusion. Credit-based insurance scores
enhance the risk assessment process. It is highly unlikely that the statistically indicated
correlation between credit-based insurance scores and insurance loss propensity is a
spurious statistical relationship.
6
Unfairly Discriminatory Rates and Proxy Effect
Rate regulatory laws throughout the United States consistently require that insurance rates
not be unfairly discriminatory. This rate standard has a history in insurance literature and
rate regulation that goes back in time over 150 years.
Traditionally, insurance rates have been considered to be unfairly discriminatory if there are
premium differences that do not correspond to differences in expected losses and expenses,
or if there are differences in expected losses and expenses that are not reflected in premium
differences. Because credit-based insurance scores provide an important and accurate
measurement of risk, it would be unfairly discriminatory to charge insurance premiums that
ignored the differences in risk measured by these scores. Two insureds with significantly
different insurance scores represent a significantly different risk of loss and as such it would
be unfairly discriminatory to charge these two very different insureds the same premium.
In addition to studying the relationship of insurance scores and risk, the FTC also studied the
relationship between credit-based insurance scores and race, ethnicity, national origin, and
income. The FTC attempted to determine if differences in credit-based insurance scores
were correlated to differences in insurance risk or whether the scores were merely a proxy
for race or household income.
The FTC concluded:
a. Credit-based insurance scores are effective predictors of risk under automobile
policies.
b. Credit-based insurance scores appear to have little effect
as a proxy for membership
in racial and ethnic groups in decisions related to insurance (emphasis added).
The FTC’s use of the term “little effect” left the door open for the possibility that credit-based
insurance scores did have some proxy effect, no matter how small, with respect to race and
ethnicity. Based on its analysis the FTC estimated the proxy effect for African Americans to
be +1.1% and for Hispanics to be +0.7%.
7
The FTC measured this purported proxy effect by comparing the average predicted risk
derived from a model without controls for race to the average predicted risk for each racial
group derived from a model with controls for race. In order to have any confidence that the
small 1.1% and 0.7% proxy effects on risk are accurate and have any significance, the FTC
needed to precisely control its analyses for all known risk factors other than race.
Unfortunately, the FTC simply did not have a database that was refined enough to accurately
identify such a small proxy effect of 1.0% or less.
The FTC acknowledged “that the large differences in average risk on comprehensive
coverage for Hispanics and African Americans should be treated with some
caution, as the
geographic risk variable in the FTC database is not a very effective control for geographic
variation in risk on comprehensive coverage” (emphasis added). I was directly involved in
designing the database used by the FTC, exclusive of data concerning credit-based
insurance scores, race, and household income. I advised the FTC that the geographic risk
data were less than an ideal control for geographic variation in risk for the bodily injury
coverage, as well as for the comprehensive coverage.
The way that the FTC grouped the data by age/gender/marital status, by tenure, by mileage,
and by geography reduced the FTC’s ability to accurately control its statistical analysis for all
known risk factors. The FTC’s problems with the traffic violations data also limited its ability
to accurately measure a proxy effect. It is highly likely that the 1% and less proxy effect
which the FTC ascribes to race would have disappeared entirely had the FTC been able to
more accurately control the analysis for all known risk factors, especially the geographic risk
factor.
To support this contention I would draw attention to the calculation of the proxy effect for the
property damage liability coverage. The data used by the FTC to control for geographic risk
for the property damage liability coverage was not ideal, but it was better geographic data
than for any other coverage. Where the FTC could adequately control for geographic risk the
FTC found “very little difference in the impact of credit-based insurance scores on predicted
risk based on whether the model included controls for membership in a protected class”. In
fact, the FTC found some evidence that inclusion of race in its model may be having an
effect that was opposite a proxy effect for the property damage coverage (see FTC Study
page 68). I suspect that if the FTC had been able to control for geographic risk as accurately
8
for all coverages as it did for the property damage liability coverage, all hints of a proxy effect
would have disappeared, as was the case for the property damage coverage.
My primary criticism of the FTC Study is that readers were not properly warned as to the
limitations of the data. The database was sufficiently refined to allow for general conclusions
as to the ability of credit-based insurance scores to predict risk, both on an overall basis and
within racial and income groups. The database was not sufficiently refined to allow for the
measurement of a proxy effect that is as small as 1% or less.
In my opinion, the proper conclusion to be drawn from the FTC Study is that credit-based
insurance scores are not proxies for race or income. Knowing someone’s score provides no
information, or even the basis for an educated guess, as to their race or income. If a small
proxy effect does exist it is so small as to be unmeasurable by the FTC database. Surely the
very small proxy effect hypothesized by the FTC, but not statistically proven, cannot
invalidate an important risk factor that contributes significantly to the measurement of risk
and benefits everyone by making insurance coverage more readily available.
California Rate Regulation
Ratemaking regulations implemented in California subsequent to the passage in 1989 of
Proposition 103 have limited insurers’ abilities to implement accurate rate factors, thereby
creating hidden cross-subsidies within the insured population. In California these hidden
subsidies primarily benefit urban insureds and disadvantage rural insureds. California rate
regulations also result in the disallowance of legitimate and necessary business expenses
and limit the underwriting profit factors to amounts that are well below most insurers’ true
cost of equity capital.
California has imposed a ban on the use of credit-based insurance scores, apparently based
on the mistaken notion that it is “protecting” low-income insureds and racial minority
insureds. This ban actually creates significant and undesirable cross-subsidies. Many low-
income insureds with better than average insurance scores are currently being required to
subsidize relatively rich insureds with poor scores. Also, as shown by the FTC Study,
9
significant numbers of insureds within every racial minority group are being required to
subsidize the cost of insurance for those of all races with below-average scores.
This arbitrary ban in California on insurance scores potentially creates, in my opinion, an
undesirable change in the nature of competition. Rather than competing on price and
competing to write as much business as is financially prudent, insurers will likely tend to turn
to pre-screening marketing techniques and insurance will not be readily available for some
insureds with above-average risk.
Critics of the insurance industry often contend that any restrictions on the risk assessment
process, or arbitrary limits on rates, are justified as long as the insurance market has not
completely collapsed through insurer withdrawals or insolvencies. In my career, I have seen
rate regulation push the New Jersey personal insurance market towards collapse during the
1970’s and 1980’s. We may be seeing the same thing recurring now in the Florida
homeowners insurance market. In my opinion, there is little in the California “experiment”
with rate regulation that should be transported to other states. If all states copied California
there would be no states left to provide subsidization and create the capital necessary for this
industry to remain financially healthy for the benefit of every insured.
Federal Reserve Board Study
In August 2007 the Federal Reserve Board (i.e., FRB) released its study of the impact of
credit scoring on the availability and affordability of credit. There were both similarities and
striking differences in the FRB’s findings and the FTC’s findings with regard to credit-based
insurance scores.
The FRB found the mean credit score for Asians and non-Hispanic whites to be slightly
above average. The mean score for Hispanics was 38% of the average and for African
Americans it was 26% of the average. The FTC’s findings with respect to the median
insurance scores by racial group were strikingly similar to the FRB’s findings.
Both the FRB and the FTC conducted multi-variant analyses of other known risk factors. In
order to determine if the apparent differences in average scores by racial group were due to
10
race or due to other known risk factors, both the FRB and the FTC conducted a multi-variant
analysis within each racial group. When the race is the same for all participants in the study,
the FRB found that credit scores were predictive of credit risk and the FTC found that credit-
based insurance scores were predictive of auto insurance losses. These findings prove that
credit scores and credit-based insurance scores are not surrogates for race.
Unlike the FTC, the FRB recognized and discussed the fact that a practice applied uniformly
to all applicants may not have a precisely uniform impact on all sub-groups. The FRB
recognized that “courts and federal regulators of credit discrimination” have traditionally
accepted some degree of differential impact if there is a “sufficient business justification” for
the practice in question.
Unlike the FRB, the FTC mistakenly suggested that a small portion of the total risk being
measured may be ascribed to race. The portion of risk mistakenly ascribed to race was the
portion of the risk the FTC was unable to explain with other known risk factors. I have
previously discussed in this testimony how the FTC failed to adequately control for
geographic risk on three of the four major auto insurance coverages, and why what the FTC
labeled as a “proxy for race” was nothing more than unexplained geographic risk.
1
Exhibit I
CURRICULUM VITAE
NAME: Michael J. Miller
BUSINESS ADDRESS: 21253 N 825 East Road
Carlock, IL 61725
E-Mail: [email protected]
EDUCATION: ILLINOIS STATE UNIVERSITY
Bachelor of Science – 1968
Major – Mathematics
Minor – Accounting
CONTINUING Estimated study time exceeding 3,000 hours
EDUCATION: necessary for completion of 10 qualifying exams for
membership in Casualty Actuarial Society (CAS).
Participation as an attendee and on the faculty
of the CAS Loss Reserve Seminar, the CAS
Ratemaking Seminar, and other CAS educational
seminars on special topics, such as rate of return
and underwriting practices.
Meet all continuing education requirements of the
American Academy of Actuaries necessary to sign
a public actuarial opinion.
MEMBERSHIP IN Casualty Actuarial Society (CAS)
PROFESSIONAL Associate Member 1971
ORGANIZATIONS: Fellow 1981
American Academy of Actuaries (AAA) 1975
Conference of Consulting Actuaries 2002-2004
Fellow
International Actuarial Association
Midwestern Actuarial Forum
Chartered Life Underwriter (CLU)
2
PROFESSIONAL CAS Committee on Risk Classification,
ACTIVITIES: Member 1982-1984
Chairman 1983-1984
CAS Committee on Principles of Ratemaking
Member 1985-1987
Chairman 1991-1992
CAS Examination Consultant 1987-1990
CAS Long-Range Planning Committee 1993-1994
1997-2000
CAS Board of Directors 1992-1993
2001-2003
CAS Officer,
Vice President – Research and Development 1993-1996
CAS Task Force on Non-Traditional Practice Areas
Chairman 1998-2000
CAS/SOA Joint Task Force on Financial Engineers 1998-2001
AAA, Liaison Committee to the National
Association of Insurance Commissioners 1985-1988
Actuarial Education and Research Fund
Board of Directors 1994-1996
AAA, Casualty Practice Council 1990-1993
Property Casualty Committee of Actuarial
Standards Board, Member 1987-1993
Chairman of Ratemaking Subcommittee 1987-1988
Chairman of Property/Casualty Committee 1989-1993
Midwestern Actuarial Forum
Education Officer 1986-1987
President 1988
EMPLOYMENT State Farm Insurance 1967-1984
HISTORY: M. J. Miller and Company 1984
Tillinghast 1984-1993
Miller, Herbers, Lehmann, & Associates, Inc. 1994-2002
EPIC Consulting, LLC 2003-Present
3
PROFESSIONAL “Private Passenger Automobile Insurance
PUBLICATIONS: Ratemaking”, Proceedings of CAS, Volume LXVI.
“Review – Risk Classification Standards by
Walters”, Proceedings of CAS, Volume LXVIII.
“A History of the Rating and Regulation of
Personal Car Insurance in the United States”,
The Institute of Actuaries of Australia, February, 1990.
“An Evaluation of Surplus Allocation Methods
Underlying Risk Based Capital Applications”,
CAS Discussion Paper Program, Volume I, 1992.
“How to Successfully Manage the Pricing Decision
Process”, CAS Discussion Paper Program, 1993.
“Building a Public Access PC-Based DFA Model”,
CAS Forum, Summer 1997, Volume 2.
“Auto Choice: Whose Fault Is It Anyway”, Contingencies,
January/February 1998
“Actuarial Implications of Texas Tort Reform”, CAS Forum,
Spring 1998.
“The Relationship of Credit-Based Insurance Scores to Private
Passenger Automobile Insurance Loss Propensity”, June 2003.
“Disparate Impact and Unfairly Discriminatory Insurance Rates”,
CAS Call Paper Program, February 2009.
PRESENTATIONS: Faculty member on National Association of Insurance
Commissioners’ orientation program for new insurance
commissioners, 1987-1994.
Faculty member on National Association of Independent
Insurers’ seminars on ratemaking and loss reserving.
“Key Provision in Rate Filings”, Society of State Filers.
Numerous presentations at educational seminars and meetings
conducted by the Casualty Actuarial Society on topics including
ratemaking, loss reserving, underwriting, risk classification
and rate of return.
EXPERT TESTIMONY: Rate Regulatory Hearings in Alberta, California, Florida, Georgia,
Louisiana, Maryland, Massachusetts, Michigan, Mississippi,
New Brunswick, New Jersey, New York, North Carolina, Ohio,
Oklahoma, Ontario, Pennsylvania, Texas, Vermont, West Virginia,
and Wyoming.
Courts in Alabama, California, Florida, Minnesota, Mississippi,
New Hampshire, Pennsylvania.
Birny Birnbaum
Center for Economic Justice
NAIC Hearing on Insurance Credit Scoring
April 30, 2009
Actuarial Considerations
The Role of Risk Classifications in the Insurance System
Risk classifications are any factor used by an insurer to segment the population for
purposes of determining whether to offer insurance, what terms and products to offer and
what price to offer, whether that is called underwriting, tier placement or rating.
What is role of risk classification? Why not one average rate for everyone -- why not
average rates for everyone?
1. Protect insurer financial condition by preventing or limiting adverse selection
2. Promote loss prevention / loss mitigation
3. Fairness in pricing -- group consumers of similar risk for purposes of assigning
premium
4. Fairness in pricing -- rates based on characteristics society deems fair
Some risk classification is essential to prevent adverse selection, to provide incentives for
loss prevention and meet basic societal standards of fairness and equity..
Revolution in Risk Classification
In recent years, risk classification has become more and more detailed, with more and
more rate levels, more rating factors and more categories within rating factors. As risk
classification becomes more and more detailed, the spread of prices increase as the cost
for the most desirable policyholder goes down and the cost for the most undesirable
policyholder goes up.
This has profound implication for the affordability of insurance because, inevitably, those
consumers least able to afford insurance are the least desirable consumers for insurers
and the ones facing higher and higher prices because of ultra risk classification.
Risk classification taken to the extreme is the end of insurance – a pay-as-you-go system
in which those who pose no risk pay little or nothing and those who have a claim pay the
cost of the claim.
NAIC Hearing Insurance Credit Scoring
Birny Birnbaum, Center for Economic Justice
Actuarial Considerations
April 30, 2009
Page 2
Insurance credit scoring was the beginning of this revolution in risk classification because
it was a massive database on a huge number of consumers that lent itself to the data
mining necessary for refined risk classifications. That data mining / risk classification
process has moved on to other consumer databases and other questionable risk
classifications: education, occupation, prior liability limits, household composition,
property-specific catastrophe and geographic rating, policy inquiries and surely others we
do not know about.
Correlation Necessary, But Not Sufficient to Justify a Risk Classification
The fact that a characteristic of the consumer, vehicle or property is associated with a
difference in expenses or expected losses is a necessary, but not sufficient, justification
for use as a risk classification.
For actuaries, the definition of fair or equitable is simply a difference in expected costs:
Rates within a risk classification system would be considered equitable if differences in
rates reflect material differences in expected cost for risk characteristics. In the context of
rates, the word fair is often used in place of the word equitable.
1
While this may be a sufficient definition of fair for insurers and actuaries, it is not
sufficient for meeting public policy goals of insurance.
Just because insurers can find a characteristic that is correlated with expenses or expected
losses does not mean that characteristic should or must be used.
Insurers themselves ignore or downplay risk factors demonstrably related to expected
losses for their own business purposes – miles driven is a prime example – and yet the
insurance system has not collapsed.
Insurers argue that failure to use insurance scoring will result in “low risk” consumers
subsidizing “high risk” consumers.
The purpose of insurance is to pool risks – those with claims are subsidized by those
without claims. The purpose of insurance is not to allocate costs to those who generate
the costs, but to spread the risk and the costs across a pool of consumers. This is what
makes it insurance and not a pay-as-you-go system.
1
ASOP 12, 2005 Edition, Section 3.2.1. The original ASOP on Risk Classification, issued in 1989, Section
2.4, defined “Equitable or Fair—Appropriately reflecting differences among the costs of identifiable risk
characteristics. The two terms are used interchangeably in this standard.”
NAIC Hearing Insurance Credit Scoring
Birny Birnbaum, Center for Economic Justice
Actuarial Considerations
April 30, 2009
Page 3
Certain characteristics are not permitted – race, religion, national origin and in some
states, marital status, credit information or being an elected official. They are not
permitted because society does not believe these are proper bases for charging different
insurance premiums – regardless of whether these characteristics are correlated with
claims. We don’t allow race-based life insurance premiums even though there is
evidence that African Americans have shorter life expectancies that White Americans.
While we will argue that insurance scoring violates public policy and should be banned
for that reason, it is clear that insurers’ use of insurance scoring is not necessary , one, to
protect insurer financial condition by preventing adverse selection, and two, to create a
fair risk classification system.
It is not necessary because sufficient risk classification exists to protect insurer financial
condition and prevent adverse selection. Insurance scoring not only does not prevent
adverse selection – most consumers do not know that their credit information is used for
underwriting or rating insurance – insurance scoring lends itself to adverse selection
because it invites consumers to manipulate their credit score. The evidence of this is that
are no problems associated with the absence of credit scoring to be found in those states
which ban credit scoring. In fact, insurers tout the new Massachusetts auto insurance
regulatory regime – Progressive and GEICO have recently entered the market – despite a
ban on insurance scoring. There are no problems in the California auto or Maryland
homeowners markets attributable to these states’ prohibition on insurance scoring.
Insurance scoring is not necessary to create a fair insurance system. A ban on insurance
scoring does not prohibit insurers from using risk classifications that consumers
understand and respond to for purpose of loss mitigation, like driving record, anti-theft
devices, type of vehicle, catastrophe-resistant construction and many others.
Further, insurance scoring should be banned because it undermines the vital loss
mitigation role of insurance. Insurance scoring provides no economic incentives for
changing risky behavior; only economic signals to manipulate a credit report.
But beyond this, insurance scoring should be prohibited on purely actuarial grounds as
unfairly discriminatory in the actuarial sense and in violation of actuarial standards of
practice.
Insurance Scoring is Unfairly Discriminatory within Traditional Actuarial
Standards and Should Be Prohibited by Regulators Using Existing Regulatory
Authority
First, it is important to state that actuarial standards are generally developed by actuaries
who work for insurance companies and the effect of the standards is broaden the
acceptable practices of actuaries rather than limit them.
NAIC Hearing Insurance Credit Scoring
Birny Birnbaum, Center for Economic Justice
Actuarial Considerations
April 30, 2009
Page 4
Yet, even the actuaries, within the Actuarial Standard Of Practice (ASOP) 12 on Risk
Classification, acknowledge the obvious – that a risk classification must be objectively
and specifically identified:
3.2.3 Objectivity
—The actuary should select risk characteristics that are capable
of being objectively determined. A risk characteristic is objectively determinable
if it is based on readily verifiable observable facts that cannot be easily
manipulated. For example, a risk classification of “blindness” is not objective,
whereas a risk classification of “vision corrected to no better than 20/100” is
objective.
There is also a document called the “Risk Classification Statement of Principles,” which
contains much the same guidance as the ASOP, but also includes a section on
controllability.
Controllability refers to the ability of a risk to control its own characteristics as
used in the risk classification system. While controllability is in many cases a
desirable quality for a characteristic in a risk classification system to have,
because of its close association with an effort to reduce hazards and the resulting
general acceptability by the public, it can easily be associated with undesirable
qualities, such as manipulation, impracticality and irrelevance to predictability of
future costs.
Clearly, a risk classification that can be manipulated by the consumer is not objective or
specifically identifiable. Further, if, for example, there are five categories of a particular
risk classification, a consumer should be identified with one, and only one, category. If
the categories for miles driven are 0 to 5,000; 5,001 to 7,500; 7,501 to 10,000; 10,001 to
12,500 and greater than 12,500, a consumer is eligible for one category only.
The risk classification categories must be mutually exclusive else the risk classification
will be unfairly discriminatory in the actuarial sense – it will cause consumers of similar
expect risk to be treated differently or cause consumers of different expected risk to be
treated the same.
For example, a risk classification based on hair – length, color, thickness – would not be
actuarially sound because a consumer could manipulate his or hair – by cutting it,
coloring it, treating it, having hair transplants, putting on a wig. And the hair could
change over time – brown one day, grey a few weeks later.
NAIC Hearing Insurance Credit Scoring
Birny Birnbaum, Center for Economic Justice
Actuarial Considerations
April 30, 2009
Page 5
Insurance scoring violates actuarial standards of practice because the risk classification is
not objectively and specifically determinable and because it is subject to manipulation.
Insurance scoring is unfairly discriminatory for the following specific reasons and should
be prohibited by regulators under existing statutory authority.
Not Objective
1. Differences across credit bureaus
2. Differences within a credit bureau due to lender choices
3. Changes in definitions of credit report items – bankruptcy law change
4. Public policy initiatives changing credit scores – moratorium on foreclosures
5. Lack of information – 25% of reports contain insufficient information for scoring,
clearly that 25% of population have a variety of risk characteristics
6. Timing of report – balance to limits varies by time of the month
7. Decisions of lenders – not reporting limits, changing limits
Manipulation
1. Invitations/Solicitations for Manipulation
2. Piggy-Back on another consumer
3. Shift balances from one car to multiple cards
Penalize Consumer for Rational Behavior
1. Shop around for best rates
2. Cancel a card when lender acts unfairly
3. Get a card to get 10% first visit discount
Impact of Economic Conditions / Model Versions
1. Miles driven: fewer miles driven means less exposure regardless of economic
conditions
2. Delinquency: means something different in 2008 than in 2004 – FICO has updated
its credit scoring model (FICO 08) to address the fact that its scoring model did not
work well in predicting defaults and subprime crisis.
Multiple Models
1. Modelers produce multiple auto and homeowners models: Preferred Auto Minimum
Limits; Preferred Auto Greater Than Minimum Limits, Standard Auto Minimum
Limits, Standard Auto Greater Than Minimum Limits, Non-Standard Auto.
2. Consumer outcomes can vary based on which model the consumer is channeled into.
NAIC Hearing Insurance Credit Scoring
Birny Birnbaum, Center for Economic Justice
Actuarial Considerations
April 30, 2009
Page 6
Accident Frequencies Decline as Credit Scores Worsen
There is strong evidence that insurance scoring itself is not a predictor of risk or
insurance claims, but, rather, that insurance scoring is a proxy for some other factor or
factors that are truly related to claim experience.
If a risk classification is truly related to claims, we expect to see that relationship hold
over time and as the incidence of the risk classification in the population increases or
decreases. For example, we know that fewer miles driven means, on average, less
exposure to accidents and claims, while more miles driven means more exposure. If
there is a significant increase or decrease in miles driven, we expect to see fewer claims.
We know that youthful drivers are more likely to be in auto accidents that drivers with
greater experience. This insight has led to graduated licensing programs and a resulting
reduction in youthful driver accidents. But, if the percentage of youthful drivers in the
population were, for example, to double from 10% to 20% with a couple of years, there is
no doubt that the overall population frequency of accidents would increase.
With insurance scoring, we do not see the relationship between credit scores and claims
hold over time. Over the past two years, credit scores have suffered because of:
• an increase in loan delinquencies
• an increase in mortgage defaults
• an increase in foreclosures
• an increase in bankruptcies
• an increase in debt to limits ratios because of lenders slashing credit
limits and unemployment
• an inability to tap home equity due to negative equity and tightened
lending standards
All these economic results impact credit scores – payment history, public records,
balance to limits ratios.
If credit scoring were truly related to the likelihood of filing a claim, then during this
unprecedented period of financial stress on consumers, we would expect to see an
increase in claims. Stated differently, just as we would expect an increase in claims if the
percentage of youthful drivers in the population increased rapidly, we would expect an
increase in claims if the percentage of consumers in the population with poor credit
scores increased.
Yet, during this period of worsening credit scores, auto claim frequency has declined.
The Insurance Services Office reports the following changes in claim frequency from
first quarter 2006 through fourth quarter 2008:
NAIC Hearing Insurance Credit Scoring
Birny Birnbaum, Center for Economic Justice
Actuarial Considerations
April 30, 2009
Page 7
Bodily Injury Total Limits -10%
Property Damage Liability -5%
Personal Injury Protecton -15%
Collision -2%
Comprehensive -13%
The Missouri, Texas and FTC Studies
The Missouri Department of Insurance Study
The Missouri Department of Insurance released a study that specifically examined the
impact of insurance credit scoring on the availability of insurance coverage in poor and
minority communities. This was the first independent study based on detailed insurance
credit scoring data using rigorous statistical analysis. The Department collected credit
score data aggregated at the ZIP Code level from 12 insurers for the study period of 1999
to 2001. For each Missouri ZIP Code, the Department obtained:
• Mean credit score
• The number of exposures for each of five equal credit score intervals
The Department then utilized a variety of multi-variate statistical techniques to isolate the
relationship of income and race to insurance credit scoring, independent of other factors.
The study found:
• The insurance credit-scoring system produces significantly worse scores for
residents of high-minority ZIP Codes. The average credit score rank in “all
minority” areas stood at 18.4 (of a possible 100) compared to 57.3 in “no minority”
neighborhoods – a gap of 38.9 points. This study also examined the percentage of
minority and white policyholders in the lower three quintiles of credit score ranges;
minorities were overrepresented in this worst credit score group by 26.2 percentage
points.
• The insurance credit-scoring systems produces [sic] significantly worse scores for
residents of low-income ZIP Code. The gap in average credit scores between
communities with $10,953 and $25,924 in per capita income (representing the
poorest and wealthiest 5 percent of communities) was 12.8 percentiles. Policyholders
in low-income communities were overrepresented in the worst credit score group by
7.4 percentage points compared to higher income neighborhoods.
NAIC Hearing Insurance Credit Scoring
Birny Birnbaum, Center for Economic Justice
Actuarial Considerations
April 30, 2009
Page 8
• The relationship between minority concentration in a ZIP Code and credit scores
remained after eliminating a broad array of socioeconomic variables, such as
income, educational attainment, marital status and unemployment rates, as
possible causes. Indeed, minority concentration proved to be the single most reliable
predictor of credit scores.
• Minority and low-income individuals were significantly more likely to have worse
credit scores than wealthier individuals and non-minorities. The average gap
between minorities and non-minorities with poor scores was 28.9 percentage points.
The gap between individuals whose family income was below the statewide median
versus those with family incomes above the median was 29.2 percentage points.
Based upon the results of this study, the former Governor of Missouri has called for a ban
on insurance credit scoring.
The Texas Department of Insurance Preliminary Report
The Texas Department of Insurance (TDI) reviewed over 2 million policyholder records
and obtained policyholder-specific information on race. The TDI report, issued in the
beginning of January 2005, states unequivocally that insurance credit scoring
discriminates against minority consumers:
The individual policyholder data shows a consistent pattern of differences in
credit scores among the different racial/ethnic groups. The average credit scores
for Whites and Asians are better than those for Blacks and Hispanics. In addition,
Blacks and Hispanics tend to be over-represented in the worse credit score
categories and under-represented in the better credit score categories.
2
The TDI study confirms and validates the Missouri Department of Insurance (MDI)
study. Insurers complained about the Missouri study because it inferred socio-economic
characteristics from ZIP Codes to average credit scores. But the MDI methodology is
well accepted in the field of fair lending analysis. The TDI study not only confirms the
MDI study results – it validates the MDI methodology.
2
Texas Department of Insurance, “Report to the 79
th
Legislature: Use of Credit Information in Texas,”
December 30, 2004, page 3.
NAIC Hearing Insurance Credit Scoring
Birny Birnbaum, Center for Economic Justice
Actuarial Considerations
April 30, 2009
Page 9
The FTC Study
Mandated by Congress, the Federal Trade Commission conducted a study of insurance
credit scoring for auto insurance. The study was flawed, biased and unreliable for many
reasons, some of which are listed below. But even this flawed report on insurance
scoring – despite relying upon data hand-picked by the insurance industry – found
insurance scores were worse on average for African-Americans and Hispanics and that
insurance scoring was a proxy for race. And had the FTC actually used an independent
and comprehensive set of insurance data, the measured racial discrimination would have
been much greater.
The FTC used only data on policies secretly selected by insurers. No data on applications
that did not result in policies were obtained or analyzed. Consequently, consumers who
were priced out of the market for the handful of insurers included in the study because of
insurance scoring did not get counted or analyzed. It is certain that this population was
disproportionately minority.
The FTC analysis of insurance scoring is deeply flawed and the report is unresponsive to
its Congressional mandate. The problems include:
1. The failure to obtain a comprehensive and independent data set for analysis
and the reliance upon a data set hand-picked by the insurance industry. The
insurance industry effectively controlled the study by dictating the data that
would be used in the study.
2. No substantive analysis of the impact of insurance scoring on the availability
and affordability of insurance products as requested by Congress. Because of
its reliance on industry-selected data, the FTC performed no analysis of how
consumers actually fared from insurers’ use of credit scoring.
3. Regurgitating insurer claims about credit scoring despite evidence that
contradicts these claims. The FTC ignored evidence indicating that the
correlation between insurance scores and claims was a spurious correlation –
that insurance scoring was a proxy for some other factor actually related to
claims.
4. The failure to analyze the "blaming-the-victim" strategy used by insurers to
justify insurance scoring -- the bogus claim that people who manage their
finances well are likely to manage their risks well and that's why credit
scoring works. The fact is that, by the credit modelers own admission, fully
20% of the population is unscorable with tradition credit reports because of
little or no information in the files. These folks are disproportionately low
income and minority consumers who get charged higher rates through no fault
NAIC Hearing Insurance Credit Scoring
Birny Birnbaum, Center for Economic Justice
Actuarial Considerations
April 30, 2009
Page 10
of their own. And even a cursory examination of actual scoring models
reveals that most of the factors determining an insurance score have nothing to
do with whether a consumer pays her bill on time, but with factors related to
socio-economic status. Yet, the FTC report dutifully repeats this desperate
rationalization for insurance scoring with no critical analysis.
5. The failure to examine any alternatives to insurance scoring that are predictive
of claims but are not based on any consumer credit information. The FTC
ignored research indicating that insurers could eliminate the use of credit
information but obtain the same ability to predict claims with advanced
modeling and data mining of traditional rating factors. Consequently, the
FTC ignored an obvious alternative to insurance scoring that could reduce the
impact on low income and minority consumers.
Comments by J. Robert Hunter, FCAS, MAAA
1
on
“Some of the Reasons Why Credit Scoring is Actuarially Unsound”
IamBobHunter,DirectorofInsurancefortheConsumerFederationofAmerica.
Actuarieswho,forthemostpart,workdirectlyorindirectlyforinsurance
companiesdevelopedtheactuarialstandardspublishedbytheactuarial
associations.Whilemostofthepeoplewhoparticipateonactuarialcommitteesare
finepeople,theyhaveapointofviewthatisverymuchreflectiveoftheirtraining
andtheiremployers’pointofview.Fewofthedevelopersofthestandardshave
experiencethinkingaboutinsuranceissuesfromtheconsumerpointofview.Thus,
thestandardshaveaninsurer‐biasandminimizerestrictionsonfreedomofaction
ofboththeactuaryandtheiremployers.
Becauseofthis“flexibility”Ihavebeeninmanyrateandotherpublichearings
wheretheactuaries,allclaimingtofullyadheretothestandards,comeoutwith
wildlydifferentrecommendations.
Regulatorsshoulddeveloptheirownimpartialstandardstodefine“actuarial
soundness.”
Butevenwiththisbiastowardallowingeverythingtopassmuster,CreditScoringis
ActuariallyUnsoundforSeveralReasons.
Here are a few of the reasons:
• Credit scores are subject to manipulation
, for example by services promising vast
improvements in a person’s score
• Credit scores are not based on a plausible (logical) relationship to risk and is thus
obscure and irrelevant to the insurance provided.
• Credit scores are not objective
because of how scores vary between credit
bureaus, how lender decisions impact a score, how definitions of such key items
as “bankruptcy” vary over time and other reasons.
• Credit scores are not supportive of the hazard reduction incentives
of a sound
class system; indeed, it undermines such incentives
1
Mr. Hunter is Director of Insurance for Consumer Federation of America and formerly served as Texas
Insurance Commissioner and Administrator of the Federal Insurance Administration.
• Credit scores are a proxy for income and race, which actuaries know are
prohibited criteria that should not be used, even in a back-door fashion.
Toexpandabit:
Creditscoringisnotobjectivebecause:
1. Therearesharpdifferencesforthesameinsuredacrossthethreecredit
bureaus.CFA’sstudyof500,000creditreportsforindividualsfoundthat
consistentresults(within20points)forallthreerepositoriesonlyhappened
21percentofthetime.Errorswerewidespread,with22percentofpeople
gettingtoolowascoreandthesamepercentage,22percent,receivingtoo
highascore,whichrepresentsa44percenterrorrate.Thisresearch
involvedonlyonepricebreakregardingthescoreneededtoobtainthebest
creditterms,notthe50ratetiersthatsomeinsurersuse.Obviouslyitis
likelythatalargenumberofinsuranceapplicantsarenotproperlyratedfor
insuranceusingcreditscore.
2. Therearedifferenceswithinacreditbureauduetolenderchoicesofhowto
reportandiftoreportinformation.
3. Therearechangesindefinitionsofkeycreditreportitems–bankruptcylaw
changeisanexample.
4. Therearepublicpolicyinitiativeschangingcreditscores–forexample,a
moratoriumonforeclosures
5. Thereisalackofinformationformanypeople–25%ofreportscontain
insufficientinformationforscoring,clearlythat25%ofpopulationhavea
varietyofriskcharacteristicsandarenotthesamefromariskstandpoint.
6. Thetimingofreportcanchangetheresult–thebalancetolimitsitemvaries
bytimeofthemonth
7. Decisionsoflenderscanimpacttheresult–notreportinghighlimitstonot
alertcompetitorsthatapersonisagoodcreditriskisanexample.
Creditscorescanbemanipulated.Hereisacoupleofwaysthathappens:
1. Aconsumerreceivesasolicitationformanipulationfromaservice
offeringtoraiseyourscoreby100pointsin24hours.
2. OneconsumerPiggy‐Backsonanotherconsumer
3. Theconsumershiftsbalancesfromonecardtomultiplecards
Creditscorescanpenalizeaconsumerforrationalbehavior,forexample:
1. Whenaconsumershopsaroundforbestrates
2.
Whenaconsumercancelsacardwhenlenderactsunfairly
In my testimony, I comment on several excerpts from the American Academy of
Actuaries “Risk Classification Statement of Principles,” showing some of the reasons that
credit scoring is actuarially unsound, even using a document developed by industry-
oriented actuaries.
The Statement of Principles says an actuarially sound class system groups “risks with
similar risk characteristics” together based on “relevant” factors. The statement goes on
to say that “Risk classification characteristics should be neither obscure nor irrelevant to
the insurance provided…” and that there must be a “plausible relationship between the
characteristics of a class and the hazard insured against.”
The problem is that insurers cannot tell us what it is about a credit score that is linked
with risk. They have merely a correlation to lean on, not a logical thesis underpinning the
correlation. This is data mining at its worst and, because of the lack of any underlying
rationale; the use of credit scoring in insurance is obscure to the insurance provided and
therefore actuarially unsound.
Some actuaries have said that a thesis is not required because actuarial principles state
that a cause and effect relationship is not required. While this is true, the principles also
say that a “plausible relationship between the characteristics of a class and the hazard
insured against.”—a logical underpinning for the use of the information -- is required.
Here is the key quote from the Causation Section of the Principles:
Often causality is not used in its rigorous sense of cause and effect but in
a general sense, implying the existence of a plausible relationship
between the characteristics of a class and the hazard insured against…
Risk classification characteristics should be neither obscure nor
irrelevant to the insurance provided; but they need not always exhibit a
cause and effect relationship.
2
Credit scoring is at best obscure relative to auto and home insurance, and probably is
downright irrelevant.
Since there is no plausible logical basis underlying credit scoring, since it can be
manipulated and since it is not objective the classification violates actuarial principles.
TherestofthistestimonyshowsselectedexcerptsformtheStatementofPrinciples
alongwithcommentsastowhycreditscoringviolatesthequotedmaterial.
EXCERPTS FROM “RISK CLASSIFICATION STATEMENT OF PRINCIPLES”
of the American Academy of Actuaries Committee on Risk Classification
3
2
Risk Classification Statement of Principles, American Academy of Actuaries Committee on Risk
Classification, at http://actuarialstandardsboard.org/pdf/appendices/risk.pdf.
3
Risk Classification Statement of Principles, American Academy of Actuaries Committee on Risk
Classification, at http://actuarialstandardsboard.org/pdf/appendices/risk.pdf.
Statement from the “Summary” Section of the Statement of Principles
The grouping of risks with similar risk characteristics for the purpose of setting
prices is a fundamental precept of any workable private, voluntary insurance
system. This process, called risk classification, is necessary to maintain a financially
sound and equitable system.
The following basic principles should be present in any sound risk classification
system in order to achieve the above purposes:
* The system should reflect expected cost differences.
* The system should distinguish among risks on the basis of relevant cost-related
factors.
* The system should be applied objectively.
* The system should be practical and cost-effective.
* The system should be acceptable to the public.
CFAComment
Thereisnoreasontobelievethataperson’screditscoreisa“relevant”factorin
measuringrisk.Nordoestheuseofcreditscoresplace“risksofsimilarrisk
characteristics”intoeachclassslot.Groupingpeoplebycreditscoreisarbitraryand
notbaseduponanylogicalconnectionbetweencreditscoreandinsurancerisk.I
willdiscussthelackofalogicalbasisor“thesis”forcreditscoringatsomelength
under“Causation”below.Creditscoresarenot“relevantcostrelatedfactors.”
Creditscoringisnotobjectivesince,asIdiscussbelow,creditscoresareerror‐laden
andsubjecttomanipulation.
Thereisseriousquestionabouttheacceptabilitytothepublicoftheuseof
insurancescoringasdiscussedbelow.Therashofstatelegislativeattemptstoban
orotherwisecontrolthepracticemeasurestheconcernofthepublic.
Statementfromthe“ConsiderationsinDesigningaRiskClassificationSystem”
SectionoftheStatementofPrinciples
Program Design
1. Degree of Choice Available to the Buyer
The design of a risk classification system is affected by the degree to which the
insurance program is compulsory or voluntary. For programs which are largely
or entirely compulsory and where there is no voluntary choice among
competing institutions, broad classifications are sometimes used, the extreme
being a single class.
CFAComment
Creditscoringisusedparticularlyinhomeandautoinsurance,whicharenot
voluntarypurchasesforconsumersgivenlenderandstatepurchaserequirements.
Thiscreatesmanytiersofprices,sometimesmorethan50tiers,resultinginvery
narrowclasses,withsomeslotshavingfeworevennopeopleintheclass.
Program Design – Cont’d
5. Absence of Ambiguity
The definition of classes should be clear and objective. Once a factual
assessment of an individual risk has been made, no ambiguity should
exist concerning the class to which that risk belongs. The classes
should be collectively exhaustive and mutually exclusive.
6. Manipulation
The system should minimize the ability to manipulate or misrepresent a
risk’s characteristics so as to affect the class to which it is assigned.
CFAComment
Credit scoring is not clear and objective and, thus, ambiguous. Scores vary
between the three repositories, for instance, so an insurance price could depend on which
credit repository the insurer chooses. Errors in the credit score abound, as CFA’s
research into 500,000 credit scores showed. Consistent results (within 20 points) for all
three repositories only happened 21 percent of the time in our research. Errors were
widespread, with 22 percent of credit reports surveyed receiving a score that was lower
than they should have received and the same percentage, 22 percent, receiving too high a
score, which represents a 44 percent error rate. This research was involved only one
price break regarding the score needed to obtain the best credit terms, not the 50 rate tiers
that some insurers use. Obviously it is likely that a large number of insurance applicants
are not properly rated using credit scores.
Astomanipulation,justtype“ServicetoImproveCreditScore”intoGoogleandyou
getover27millionresults.Someservicespromisesuchresultsasthesepromises:
Don’tbesurprisedifyousave27%onyourautoinsurancealone!
(RepairYourBadCredit.com)
Increaseyourcreditscore61pts.in30days?(YourCreditAttorney.com)
LegallyRaiseCreditScore100pts.in30days.(ecreditattorney.com)
HowIraisedmycreditscore40pts.in24hrs.andsaved$8,000!
(thebestever/credit)
Statement from the “Hazard Reduction Incentives” Section of the Statement of
Principles
Risk classification systems can be designed to provide incentive for insureds to
act to reduce expected losses and thus operate to reduce the overall costs of
insurance in total. For example, recognizing sprinklers for classifying risks for
fire insurance coverages may encourage their installation and thereby reduce
expected losses. Or reduced life insurance prices for non-smokers may
encourage people not to smoke, thus reducing the hazard of premature death
caused by diseases linked to smoking.
Such incentives are desirable, but not necessary, features of a risk classification
system. Although worth pursuing, it must be recognized there are limits to
which a risk classification system can be extended in an attempt to solve
society’s problems and still serve the necessary and useful purposes for which
such a system is designed.
CFAComment
Useofcreditscoringnotonlydoesnotadvancethegoalofhazardreduction,it
activelyunderminesit.Creditscoringhasamajorimpactonprice,oftenmore
impactthanclasseswithaclearhazardreductionincentive,suchasdrivingrecord
ormilesdriven.Consumersdonotunderstandwhatcreditscoringhastodowith
theirabilitytodrivewellorbeasafehomeowner.(Infact,neitherdothedesigners
andusersofcreditscoringininsurance)Whenconsumersrealizethattheirgood
drivingdoesnotmeanasmuchascreditscoring,itfrustratesthemandundermines
safetyefforts.
Statement from the “Public Acceptability” Section of the Statement of Principles
Any risk classification system must recognize the values of the society in which
it is to operate. This is a particularly difficult principle to apply in practice,
because social values:
* are difficult to ascertain;
* vary among segments of the society; and
* change over time.
The following are some major public acceptability considerations affecting risk
classification systems:
They should not differentiate unfairly among risks.
* They should be based upon clearly relevant data.
* They should respect personal privacy.
*They should be structured so that the risks tend to identify naturally with their
classification.
Laws, regulations and public opinion all constrain risk classification systems
within broad social acceptability guidelines. Legislative and regulatory
restrictions on risk classification systems must balance a desire for increased
public acceptability with potential economic side effects of adverse selection or
market dislocation.
CFAComment
Creditscoresfailmiserablyinmeetingthecited“publicacceptability
considerations.”Creditscoresdifferentiateunfairlybetweenrisksbecauseoferrors
placingatleast50%ofapplicantsinthewrongtierandbecausericherpeopleare
lesslikelytobeincreditdifficultyand,iftheyare,aremorelikelytobeableto
affordaservicetomanipulatetheresultsintheirfavor.Further,creditscoreisa
proxyforincomeandrace,asmanystudies,citedintheNAICdraftreport,make
clear.
Creditscoresasaclassarenotbasedonrelevantdata.Acreditscorehasnothingto
dowithinsurancerisk.Acreditscorehastodowithfinancialhistoryandfortunate
orunfortunatecircumstances.Aspecificscorehasnothingtodowithdriving
capacityorabilitytobesafeinahome.
Usingcreditinformationtopriceinsuranceviolatesprivacy.Theintentseemsmore
tofindrichpeoplethansafepeople.
Itislaughabletothinkthatcreditscoresare“structuredsothatriskstendto
identifynaturallywiththeirclassification.”IfIhaveascoreof600,whatdoesthat
sayaboutmyidentificationwithanotherpersonwitha600score?DoIfeela
kinshipwiththe600scorepeople?Thinkofthemyriadwaysonecanbuildsucha
score.Howdoesthatnumbercreateanidentity?
Thestatementrequiresactuariestofollowthelaw.Allstatelawsdisapproveof
classesthatareunfair.Creditscoringisunfairforinsurancepurposesand
questionableevenforsomecreditpurposes.Peoplewithacertaincreditscoremay
beinthecategorybecauseofvastlydifferentreasons.Considerthesefacts:
• Aperson’screditreportcanvarydramaticallyamongthethreemajor
creditbureaus,soacreditscorecanvarysignificantlydependingupon
whichbureauprovidedtheinsurerwithinformation.
• Acreditscorecanvarydependingonwhattimeinthemonthyourcredit
reportwasordered.
• Acreditscoredependsonthetypeofcredityouhave,meaningthata
personcanhavealowscoreevenwithaperfectpaymentrecord.Acredit
cardwithsomecompanies,obtainingaloanfromaconsumerfinance
company,orhavinganinstallmentplanfromacardealer,mayleadtoa
lowerscoreregardlessofyourpaymentrecord.
• Acreditscoredependsonthepresenceofloaninformation,soaperson
willreceivealowerscoreforpayingincash.
• Alowerscorecanoccurifapersondoesnotborrowmuchoruseslenders
thatdon’treporttocreditbureaus.
• Becausetheratiooftheamountofdebtrelativetoacreditcardlimit,a
consumerwhousesoneofherfourcreditcardstomaximizefrequent
fliermilesgetsalowerscorethananotherconsumerwhochargesthe
sameamountbutdoesitoverallfourcards.
Theseodditiesputpeoplewithdifferentcreditrisksintothesamecreditscore
categories,makingcreditscoringquestionableevenforsomecreditpurposes.
Relyingonthisinformationtomakedecisionsaboutgrantingorratinginsurance
coveragecompoundstheunfairness.
Statement from the “Causality” Section of the Statement of Principles
Scientists seek to infer some cause and effect relationship in natural phenomena,
in order to attempt to understand and to predict. It is philosophically satisfying
to some when data exhibit such a cause and effect relationship.
Risk classification systems provide a framework of information which can be
used to understand and predict future insurance costs. If a cause and effect
relationship can be established, this tends to boost confidence that such
information is useful in predicting the future and will produce some stability
of results. Thus classification characteristics may be more acceptable to the
public if there is a demonstrable cause and effect relationship between the risk
characteristics and expected costs.
However, in insurance it is often impossible to prove statistically any postulated
cause and effect relationship. Causality cannot, therefore, be made a
requirement for risk classification systems.
Often causality is not used in its rigorous sense of cause and effect but in a
general sense, implying the existence of a plausible relationship between the
characteristics of a class and the hazard insured against. Living in a river valley
would not seem to cause a flood insurance claim, but it does bear a reasonable
relationship to the hazard insured against and thus would be a reasonable basis
for classification.
Risk classification characteristics should be neither obscure nor irrelevant to the
insurance provided; but they need not always exhibit a cause and effect
relationship.
CFAComment
Thismaybethemostimportantreasonwhytheuseofcreditscoringisactuarially
unsound.Icalltheinsurerargumentthe“CausationMyth”–thataskingfora
“plausiblerelationshipbetweentheclassandthehazardinsuredagainst”isnot
requiredbecauseaskingforsucharelationshipisakintoaskingforacauseand
effecttest.
Obviously,theActuarialStandardsdiscriminatebetweenaplausiblerelationship
andcauseandeffect.TheStandardsthusrequirethattherebeaplausible
relationshipandthattherelationshipbe“neitherobscurenorirrelevanttothe
insuranceprovided.”
The problem is that insurers cannot tell us what it is about a credit score that is linked
with risk. I have asked the proponents of the use of credit scoring to explain to the world
why a person who suffered a decline in credit as a result of being in Hurricane Katrina or
lost her job because of outsourcing or lost his job in the current economic downturn is
suddenly a worse auto or home insurance risk? They do not have a credible response or
they guess that maybe it is in the human genome. Often they say the person with a poor
credit score might be sloppy and that this carries over into driving or housekeeping. But
they only guess, they have no real logical, plausible basis for use of credit score in
insurance. What they have is merely a correlation to lean on, not a logical thesis
underpinning the correlation. This is data mining at its worst, which means that the use
of credit scoring in insurance is actuarially unsound.
Unlike insurance classifications that were in use before credit scoring was adopted, this
classifier is not based on an appropriate thesis, confirmed by a statistical analysis. In fact,
there is no legitimate thesis for the use of credit scoring. There is only an alleged
correlation based on proprietary information not open to public scrutiny.
4
However, a
correlation in search of an appropriate thesis raises serious questions about the
classification that is being used.
4
This is another difference from all previous classes where the data are public and part of rate filings
made with insurance departments. Previously, an insurer would propound a thesis and test it with the data.
If a thesis was confirmed, the insurer would file for a new class with the commissioner showing the thesis
and the data in the rate filing. An example was the use of accidents and tickets. The thesis was that people
with more accidents and tickets would be worse drivers in the future because their historic driving record
indicated less care in driving. The thesis was confirmed by data that can be viewed in its’ entirety in rate
filings.
The lack of a thesis means that credit scoring violates actuarial principles. Some
actuaries say that a thesis is not required because actuarial principles state that a cause
and effect relationship is not required. Although this is true, the principles, which were
developed by a group of excellent but mostly industry-employed actuaries and therefore
holding an overwhelmingly industry-oriented point of view, also say that a thesis -- a
logical underpinning for the use of the information -- is required.
Let me repeat the key part of the “Causation” Section of the Principles:
Often causality is not used in its rigorous sense of cause and effect but in
a general sense, implying the existence of a plausible relationship
between the characteristics of a class and the hazard insured against…
Risk classification characteristics should be neither obscure nor
irrelevant to the insurance provided; but they need not always exhibit a
cause and effect relationship.
5
Credit scoring is at best obscure relative to auto and home insurance, and
probably is downright irrelevant. Since there is no clear relationship, no thesis,
underlying credit scoring, the classification violates actuarial principles.
6
Statement from the “Controllability” Section of the Statement of Principles
Controllability refers to the ability of a risk to control its own characteristics as
used in the risk classification system. While controllability is in many cases a
desirable quality for a characteristic in a risk classification system to have,
because of its close association with an effort to reduce hazards and the
resulting general acceptability by the public, it can easily be associated with
undesirable qualities, such as manipulation, impracticality and irrelevance to
predictability of future costs.
CFAComment
Wehavealreadypointedouttheeasymanipulationpossibleintheuseofcredit
scoresininsuranceandtheirrelevanceoftheclass.So,itisclearthatthe
downsideofcontrollabilityexistsincreditscoring.
Theupside,theincentiveofaclasstoreducehazardiswhollymissingintheuse
ofcreditscores.Indeed,asnotedabove,itsuseunderminessafetyincentivesby
makingdrivingrecord,milesdriven,alarmsystems,deadboltsandothersafety‐
relatedclasseslessimportantinthedevelopmentofthefinalprice.
5
Risk Classification Statement of Principles, American Academy of Actuaries Committee on Risk
Classification, at http://actuarialstandardsboard.org/pdf/appendices/risk.pdf.
6
There are other actuarial principles that credit scoring violates as well, including the fact that it is not
socially acceptable, is subject to manipulation (there are firms that offer, for a fee, to sharply improve your
score) and is ambiguous.
STATEMENT
OF
AMERICAN INSURANCE ASSOCIATION
ON
CREDIT-BASED INSURANCE SCORING
NAIC HEARING
APRIL 30, 2009
David F. Snyder
Vice President & Associate General Counsel
Public Policy
American Insurance Association
2101 L Street, NW Suite 400
Washington, DC 20036
STATEMENT OF AMERICAN INSURANCE ASSOCIATION
ON CREDIT-BASED INSURANCE SCORING
NAIC HEARING
APRIL 30, 2009
Personal lines of insurance are performing very well by objective measures, whether you are
a consumer, company or producer. Prices are largely stable, even down in many states.
Companies are well capitalized and aggressively marketing their products. Residual markets
have shrunk to historic lows. In most areas there are dozens of companies offering personal
insurance through a wide variety of distribution channels, including independent and captive
agents, the internet and telephone.
This favorable personal lines experience for all concerned has resulted from insurers pricing
insurance based on risk, instead of ignoring it, a major cause of the financial turmoil among
lenders. Credit-based insurance scoring (CBIS) has played a key role in maintaining this risk
based pricing and in producing the favorable – for all parties – competitive personal lines
market conditions we see today. As compared to the millions of annual personal lines
underwriting and rating transactions, CBIS complaints are scant. Over-regulating, or worse
yet banning, insurance scoring, would disrupt the property and casualty insurance market in
the US, severing, as it would the link between risk and pricing of personal insurance and
eliminating a cost/effective tool that has enabled competition.
Today’s Personal Lines Market Is Performing Well By Every Measure.
Automobile Insurance.
Voluntary insurance has kept up with consumer demand and the residual markets have
dropped, all good signs of a healthy competitive market. From 1995-2005, the total number of
new cars insured increased 32%, the voluntary car years increased 36% while the residual
market car years dropped 60% and the residual market as a percentage of the total market
declined 70%.
From 1994 to 2008, the auto insurance CPI increase of 49.1% is only slightly above food and
beverages, electricity and all items. It was significantly below energy (126.3%), medical care
(72.5%) professional services (61.5%) and housing (49.4%). Auto insurance costs actually
declined as a percentage of personal income from 1995 to 2006, a long term trend that we
expect continued through 2008.
Finally, according to a widely used measure of market concentration, the Herfindahl-
Hirschman Index (HHI) where a “not concentrated” market is a rating under 1000, the auto
insurance market is quite competitive at 651, with 326 insurers writing in 2007.
Homeowners Insurance
The performance of homeowners insurance is still by and large quite favorable for
consumers. It increased marginally as a percentage of family income from .81% to 1.09%.
However, this small increase can be explained by increases in catastrophe prone areas and
other factors such as increases in insured values. Despite recent real estate declines,
housing prices and insured values are still significantly higher than a decade ago. Even with
the marginal shift of less than two-tenths of one percent, renters and household insurance
increased, from 1999-2008, at a rate far lower than energy, medical care, professional
services, and for “all consumer items” measured by the CPI. The property insurance residual
markets in many states are less than 1%, but the overall averages are skewed by a few
catastrophe-prone states. In fact, 4 states have 82% of the nationwide FAIR Plan exposures.
Homeowners insurance is also quite competitive. Using the HHI, homeowners scores 759
(again, anything under 1000 is “not concentrated”). Nationally there were 369 companies
writing this business.
Competition Made Possible By CBIS Helps Promote Availability and Affordability.
The emergence of CBIS, an objective rating and underwriting tool, has enhanced both
availability and affordability. Many government studies demonstrate that the factor is a good
predictor of risk and has assisted with affordability. The percentages range from the FTC’s
estimate that 59% of policyholders save as a result of CBIS use, to much higher percentages
for some companies.
Beyond affordability, the existence of a highly cost effective tool has allowed companies to
continue to write coverage and to increase their writings. This has improved availability. For
some companies, this means they can write virtually every risk with confidence that they have
more accurately identified and priced for risk. The resulting competition helps put pressure
on lowering prices and offers consumers more choices.
Credible Evidence That Widespread Harm To Consumers Resulting From CBIS Use In
The Current Economic Conditions Is Lacking—Indeed There Is Evidence That Such
Harm Is Not Occurring.
One of the questions the NAIC is asked as it framed this hearing was whether the current
economic conditions have caused widespread consumer difficulties due to insurers’ use of
CBIS. Fortunately, it does not appear that they have.
Credit scores do not seem to be declining en masse despite the current down economy. Fair
Isaac Corp. (now known formally as FICO) has reported that in their recent studies, CBIS
“have remained virtually the same for the general population” and “more and more
consumers appear to be realizing the value of prudent financial and credit management
practices.”
1
Additionally, FICO found in an analysis of impact on consumer scores due to
lenders’ decisions to decrease some customers credit limits that “[T]he median FICO score
for the national population did not change between April 2008 and October 2008 (based on
Equifax data alone, the national median FICO score remained 713).”
2
Experian’s “National
Score Index” report from September 2008 showed that 58 percent of Americans have credit
scores above 700 and the national average is 680. A “good” credit scores is considered
anything over 700.
3
.
The vast majority of states, NCOIL and the NAIC have all acted responsibly in balancing the
market value of CBIS with the need to assure the factor is not over-used. The industry, as
1
“Fair Isaac Credit-Based Insurance Scores” message document, January 2009
2
“Study: How Credit Line Decreases Can Affect FICO® Scores”; see: http://www.fico.com/en/Company/News/Pages/study-
findings.aspx for more information
3
See: http://www.nationalscoreindex.com/ for more information
well, has used it responsibly. Today, this combination of factors has resulted in a very low
level of complaints that belies the charges of critics. In most states that we know, they
amount to a few dozen compared to millions of business transactions using CBIS and state
and federal regulatory systems that require upfront disclosures and adverse action notices, to
encourage the filing of complaints.
There are several reasons for the lack of complaints. The first is that the evidence is that
credit scores are not deteriorating as speculated. Representatives from that industry will
share their findings. Next, the insurance scores contain other factors that would tend to
dampen the effect of lowered credit scores, if that were happening. In addition, most states
have a version of the NCOIL model law with sole basis restrictions and restrictions on the use
of certain information. In addition, some states have “extraordinary life circumstances”
language that encourages individual reviews. Finally, insurers maintain review systems that
allow agents and their policyholders to reconsider cases upon a foreclosure or loss of a job,
for example. Attached to this statement is a case in point of how one multi-line insurer uses
credit scoring and prevents complaints. See Exhibit 1.
All of these factors combine, we believe, to explain why the system is working despite the
broader economic concerns outside the insurance context. Under these circumstances,
banning or over-regulating CBIS is not only not called for but such a move would actually
inconvenience and harm the majority of all policyholders, including people of all ethnic
backgrounds and income levels.
CBIS Is Subject to Extensive Federal and State Regulation.
The federal Fair Credit Reporting Act, as amended, expressly allows insurers to use credit
information. That use, however, is subject to many federal regulatory provisions, including
that adverse action notices be provided as required by law. In addition, the sources of credit
information insurers use are heavily regulated.
States have added specific laws relating to CBIS to their pre-existing insurance statutes and
regulations. Generally, the new laws follow the NCOIL model which requires upfront
disclosures and adverse action notices, prohibits the use of certain information, requires
prompt remedy in case of incorrect information and provides sole basis restrictions.
There are established anti-discrimination protections that apply to CBIS use, with well
understood legal standards. No court has found CBIS to be unfairly discriminatory. This is
the appropriate legal and actuarial standard, as indicated in an exhibit to this testimony.
CBIS have been found to be predictive of risk across different demographic groups. Even if
average scores were to differ as well, the predictive nature remains and “disproportionate
impact” is not a standard under any law for any rating factor. See Exhibit 2.
Companies Are Taking Proactive Steps To Prevent Problems.
Based on public statements, insurers have in place various mechanisms for themselves and
for their agents to address customers’ unique or extraordinary circumstances that merit
review. Some insurers may do this in states as mandated. Others may extend this option
more broadly. See Exhibit 1, a profile of one such company.
Companies also have the ability to adjust rating tiers so as to take into account over-all
changes in the economy. This would be an additional safety valve, while still maintaining the
comparative value of CBIS.
Finally, insurers assist the public by making information available on CBIS. We believe this
helps prevent problems, as well. And when fully informed, the public has accepted the
validity of CBIS. See Exhibit 3. Attached as Exhibit 4 are some examples of public
information that AIA has made available in English and Spanish and to agents.
Government And Private Studies Have Consistently Shown That CBIS Improves Risk
Assessment and Most People Benefit From Its Use.
In recent years, there have been many public and private studies of CBIS. One of the largest
and most sophisticated, is the 2007 Federal Trade Commission report that made the
following findings:
• CBIS helps assess risk;
• CBIS may improve availability;
• Ethnicity is not used by insurers;
• CBIS does not serve as a proxy for race; and
• The majority of policyholders benefit from its use through lower costs.
These findings are consistent many other public and private reports. See Exhibit 5 for the
highlights of these studies.
Conclusion
The hearing notice indicates that it will focus on three areas: (1) definition of what constitutes
CBIS; (2) evaluation of how insurers use CBIS; and (3) discussion of how current economic
conditions have affected policyholder premiums related to CBIS. Over the years, AIA has
submitted detailed information to the NAIC on the first two items; AIA is ready to assist the
NAIC and individual insurance Commissioners further on this topic.
There is no evidence to support claims that there is widespread harm to insurance
consumers as a result of CBIS, even in today’s poor economic conditions. Instead, the
evidence is to the contrary: most people continue to benefit from the use of CBIS. There are
very few CBIS complaints, even though the regulatory systems encourage them, because of
responsible business practices and existing regulation. On the other hand, banning or over-
regulating CBIS may disrupt and weaken markets and harm far more consumers than it
helps.
EXHIBIT 1
COMPANY PROFILE
Commissioner Holland presented a question at the NAIC Spring Meeting - what are insurers doing
with respect to credit-based insurance scores (CBIS) in light of current economic challenges? Given
antitrust concerns and practical considerations, AIA tapped one member company to get an up close
look at its efforts.
BACKGROUND
Lines of Business
This insurer uses CBIS for auto and homeowners business.
Duration
It has used credit information in many states for over 10 years.
DATA REVIEW
Consider Whether CBIS Have Been Changing
In light of the current economic climate, the insurer has been reviewing its personal lines business to
see if there have been notable changes.
This insurer has not noticed any significant downward trend for its book of business.
This insurer is in the process of pulling an archive study to compare and understand score
distributions.
Given the press on foreclosures, it dug into its database to investigate whether there were changes in
scores and loss history in high foreclosure areas. Its preliminary findings show that deterioration has
not occurred.
Consider Impact of CBIS Ban
This insurer pulls sample states and looks at one of its programs to gauge the possible rating impact
if it were to be required to remove CBIS. Current information shows that the following disruption could
occur:
COMBINED
PL
PERSONAL AUTO PERSONAL PROPERTY
Growth in PL
policies from
2005 to present
Overall impact -
% getting rate
increase
Drivers 60 +
have increase
increase %
Drivers owning
home to get
increase
Families w/ teen
drivers to get
increase
Overall impact -
% getting rate
increase
Customers 60 +
average
increase %
Customers w/
renter’s ins to
get increase
Customers w/
loss in past 3 yrs
to get increase
and average %
Illinois
(C)
24% 62% 83% /
10%
67% 64% 60% 79% /
13%
71% 63% /
5%
Maryland
(D)
n/a n/a n/a n/a n/a
Wisconsin
(D)
4% 64% 82% /
8%
68% 64% 61% 83% /
17%
69% 64% /
5%
NOTE: We may update the chart with more information, as we receive it.
This insurer also considers that eliminating CBIS could impact its underwriting and eligibility. With the
advent of CBIS, the insurer has expanded its eligibility base—writing more risks in underserved areas
countrywide, regardless of where one lived. This insurer is better able to identify and provide the
most appropriate price for each risk. Without the use of CBIS, those benefits will be gone.
CONSUMER-ORIENTED EFFORTS
Notice - Expanded Reasons
Some states and the NCOIL Model require that insurers provide credit-related reasons for taking an
adverse action. A few states require that these reasons be more expansive.
This insurer has opted to use expanded reasons countrywide.
Extraordinary Life Circumstances
Some states require an insurer to offer to reconsider an applicant who has experienced certain
extraordinary life circumstances.
This insurer has opted to make its extraordinary life circumstances procedures available
countrywide. Indeed, thousands of people have benefitted from their procedures.
This insurer’s list goes beyond the state-enumerated items to consider additional hardship
situations.
This insurer’s agents are aware of this procedure to use the company’s Insurance Score
Helpline. Information about the Helpline is available on their intranet.
This insurer’s adverse action notices (and their consumer report notice) includes an 800
number for consumers to access the Insurance Score Helpline directly.
This insurer has reviewed whether their Insurance Score Helpline has experienced a recent
increase in volume. It has not noticed much recent change. In fact, fewer than ½ of 1% of
applicants and policyholders use call the Helpline.
Consumer Complaints
This insurer has tracked credit complaints – those directly to the company and those via the insurance
departments – since the late 1990s. In those years, it has gotten 69 complaints.
Education
This insurer informs its agents of the availability of consumer brochures.
This insurer has information about their use of CBIS available on their website.
EXHIBIT 2
UNFAIR DISCRIMINATION AS ACTUARIAL STANDARD
State insurance laws, and indeed the principles underpinning property and casualty insurance pricing,
rely on actuarial science to determine rates that most accurately measure loss potential. Actuarial
science accomplishes this task by finding relationships between factors and risk of loss and then
allocating costs accordingly. This is the essence of risk-based pricing. Importantly, to disregard the
predictive value of a factor (1) ignores actuarial support; (2) results in better risks subsidizing worse
risks; and (3) moves closer to a one-size-fits-all approach in direct conflict with risk classification
standards.
Pricing programs of most insurers depend on making distinctions based upon a number of different
factors. All things being equal, the one who reflects a worse risk based on this difference will pay
more. To explain, the process of risk classification involves segmenting groups of individuals
expected to have similar costs. The use of more segments makes for a more granular approach in
which actuaries can more finely hone review of an individual in order to more accurately create class
plans and measure risk potential. When there are a greater number of risk levels and pricing
variations, insureds are placed with others with a more similar risk profile, which results in a fairer
price and insurers are better able to offer coverage to people they might have otherwise declined.
Most insurers’ pricing or risk classification programs depend on making distinctions based upon
several factors (or rating variables). Common homeowners insurance factors include claim history of
applicant, construction material(s), distance from fire station, dog/breed of dog owned, fire
suppression devices, home-based business presence and type, lead paint potential (constructed pre-
1978), loss history of property, roofing material, trampoline use, slab versus basement, security
system. Common personal automobile insurance factors include age, coverage limits desired,
deductibles selected, driving record/at-fault crashes, gender, marital status, miles driven, territory,
vehicle age, vehicle make, and vehicle model. Credit-based insurance scores, like these other
factors, are predictive of loss. Neither race nor ethnicity is ever collected or considered by property
and casualty insurers.
The insurer is typically required have experience justifying its rates and in some states it must supply
this information to state insurance regulators for approval. Restricting rates, when contrary to
actuarial indications, violates the prohibition against rates that are “excessive, inadequate,
unreasonable or unfairly discriminatory.” The definition of “unfairly discriminatory” is tied to accurately
measuring risk, meaning that rates must be cost-based and treat policyholders with equal risks
equally. Consider laws that state that a rate is “unfairly discriminatory” if it “(A) is not based on sound
actuarial principles; (B) does not bear a reasonable relationship to the expected loss and expense
experience among risks; or (C) is based in whole or in part on the race, creed, color, ethnicity, or
national origin of the policyholder or an insured.” To dismiss for political or personal reasons the
predictive value of a valid factor is to ignore actuarial science, which then risks violating state
prohibitions against insurance rates that are "unfairly discriminatory."
To come full circle in our description of the background of the regulatory context, the “unfairly
discriminatory” is the very foundation for insurance regulation. It consumes the field in areas where a
state legislature does not otherwise deem a particular factor to be “unfairly discriminatory” via a public
policy mandate.
EXHIBIT 3
CONSUMERS HAVE SPOKEN – OREGON
During the November 2006 elections, Oregon voters were asked to consider a statewide ballot
initiative (Measure 42) that would have banned insurer use of credit. The measure was defeated with
citizens voting more than 2-1 (65.6% to 34.4%) against it, rejecting “mass subsidization.”
That fall, a study was commissioned to examine the potential impact on consumers if the ballot
measure was successful and the results spoke volumes about the consumer benefits of credit-based
insurance scoring. The study indicated that nearly 60 percent of personal auto policyholders paid
lower rates than they would if credit information was not used and that many insurers were writing
policies that they would not have otherwise were it not for access to credit information.
Oregon voters understood the harm Measure 42 would have caused – higher insurance rates for 60
to 70 percent of residents – and illustrates the voting public’s support for insurance pricing that
accurately reflects individual risk.
AIA’s Ken Gibson, vice president, Western Region, summed it up well at the time saying: “voters said
yes to personal responsibility, yes to risk-based pricing and no to mass subsidization.”
EXHIBIT 4
EXAMPLES OF CONSUMER INFORMATION
AIA has consumer brochures available to the public – in both English and Spanish - in hard copy and
on its website. Applicable URLs follow:
http://www.aiadc.org/AIAdotNET/docHandler.aspx?DocID=290558
http://www.aiadc.org/AIAdotNET/docHandler.aspx?DocID=290559
EXHIBIT 5
CONCLUSIONS FROM MAJOR CREDIT-BASED INSURANCE SCORING STUDIES
• “…91% of consumers either received a discount for credit or it had no effect on their
premium” and “for those policies in which credit played some role in determining the final
premium, those receiving a decrease outnumbered those who received an increase by 3.44
to 1.”
Source
: “Use and Impact of Credit in Personal Lines Insurance Premiums Pursuant to Ark.
Code Ann. §23-67-415”; A report to the Legislative Council and the Senate and House
Committees on Insurance & Commerce of the Arkansas General Assembly by the Arkansas
Insurance Dept. July 2008. The Arkansas Insurance Dept. examined approximately 2 million
auto and over 620,000 homeowners policies. Arkansas enacted the National Conference of
Insurance Legislators Model Act on Credit in 2003.
• “Credit-based insurance scores are effective predictors of risk under automobile policies.
They are predictive of the number of claims consumers file and the total cost of those
claims.” and “Scores also may make the process of granting and pricing insurance quicker
and cheaper, cost savings that many be passed on to consumers in the form of lower
premiums.” Also, when scoring is used “…more consumers (59%) would be predicted to
have a decrease in their premiums than an increase (41%).”
Source
: “Credit-based Insurance Scores: Impacts on Consumers of Automobile Insurance,” A
Report to Congress by the Federal Trade Commission, July 2007. The FTC examined more
than two million insurance policies.
• “A survey of Oregon insurers indicates that nearly 60 percent of personal auto
policyholders…pay lower rates than they would if credit information was not used. In
addition, many insurers report writing policies that they would not have written had they
not had access to credit information.”
Source:
“The Use of Credit Information by Insurers,” ECONorthwest, October 2006. This
study was commissioned during the November 2006 elections when Oregon voters were
asked to consider a statewide ballot initiative (Measure 42) that would have banned insurer
use of credit. The measure was defeated with citizens voting more than 2-1 (65.6% to 34.4%)
against it, rejecting “mass subsidization.”
• “These results [impact of using credit information] corroborate the insurance industry’s
contention that the majority of policyholders benefit from the use of credit scoring.”
Source:
“Report on the Use of Consumer Credit and Loss Underwriting Systems,” Nevada
Dept. of Business & Industry, Division of Insurance, July 2005. Insurers representing 60% of
the auto and homeowners market were surveyed for this report.
• As part of the Michigan insurance industry’s successful legal efforts to stop a regulatory
ban on credit, multiple companies reported in lawsuit filings that a ban would produce
premium increases up to 68% for both auto and homeowner policies, with individual rates
rising hundreds of dollars.
Source:
In the case of Insurance Institute of Mich., et. al. v Commissioner of the Office of
Financial and Insurance Services, (2005) Case #05-156-CZ, Barry County (MI) Circuit Court.
There the Judge issued a clear and definitive opinion saying in part credit “clearly shows an
actual
effect on losses and expenses” (Judge’s emphasis). The case is now on appeal
(#262385).
• “For both personal auto liability and homeowners, credit score was related to claim
experience even after considering other commonly used rating variables. This means that
credit score provides insurers with additional predictive information distinct from other
rating variables. By using credit score, insurers can better classify and rate risks based on
differences in claim experience.” Also, “[C]redit scoring…is not unfairly
discriminatory…because credit scoring is not based on race, nor is it a precise indicator of
one’s race.”
Source
: “Use of Credit Information by Insurers in Texas: The Multivariate Analysis,”
Supplemental Report to the 79
th
Legislature by Texas Department of Insurance (TDI), January
2005. The study analyzed scores and rating factors for over two million auto and homeowners
insurance policies in Texas.
• “…the lowest range of insurance scores produce indicated pure premiums 33% above
average and the highest range of insurance scores produce indicated pure premiums 19%
below average.”; and “…insurance scores significantly increase the accuracy of the risk
assessment process.”
Source
: “The Relationship of Credit-Based Insurance Scores to Private Passenger Automobile
Insurance Loss Propensity,” EPIC Actuaries, LLC, June 2003. The EPIC study reviewed more
than 2.7 million auto policies.
• “The correlation between credit score and relative loss ratio is .95, which is extremely high
and statistically significant. The lower a named insured’s credit score, the higher the
probability that the insured will incur losses on an automobile insurance policy, and the
higher the expected loss on the policy.”
Source:
“A Statistical Analysis of the Relationship Between Credit History and Insurance Losses,”
University of Texas Bureau of Business Research at the McCombs School of Business, March 2003.
ALEX M. HAGELI
MANAGER, PERSONAL LINES
April 30, 2009
The Honorable Michael McRaith
Chair, The Property and Casualty Insurance (C) Committee
National Association of Insurance Commissioners
2301 McGee Street
Kansas City, MO 64108
The Honorable Kim Holland
Chair, The Market Regulation and Consumer Affairs (D) Committee
National Association of Insurance Commissioners
2301 McGee Street
Kansas City, MO 64108
RE: The Use of Credit-Based Insurance Scores
Dear Director McRaith and Commissioner Holland:
Thank you for this opportunity to comment on insurers’ use of credit-based insurance scores. My
name is Alex Hageli and I represent the Property Casualty Insurers Association of America (PCI).
PCI is a national property casualty trade association comprised of more than 1,000 member
companies, representing the broadest cross-section of insurers of any national trade association.
PCI members write 39.6 percent of all personal lines insurance sold in the United States.
While PCI does not endorse the use of any particular rating factor, we do support the right of
insurers to use actuarially justified rating factors. As such, we believe they should have the ability to
use credit-based insurance scores.
The federal Fair Credit Reporting Act first authorized insurers to consider credit information nearly
40 years ago. Within the past 15 years, however, the use of credit information in insurance has
grown substantially as insurers continue to perfect its use and appreciate its accuracy. Credit-based
insurance scoring, alternatively referred to simply as insurance scoring, is an objective and accurate
method for assessing the likelihood of insurance losses. Insurers that consider credit information in
their underwriting and pricing decisions do so for only one reason – insurance scoring allows them
to rate and price business with a greater degree of accuracy and certainty. Sound underwriting and
rating, in turn, allows insurers to write more business – a direct benefit for consumers.
Honorable Michael McRaith and Kim Holland
The Use of Credit-Based Insurance Scores
April 30, 2009
2
It is important to understand how insurers use credit information and to note that there are
significant differences between the credit scores used by lenders and the credit-based insurance
scores used by many insurers. Although both are derived from information found on credit reports,
the information is measured differently. Insurers use credit information in developing insurance
scores to predict the likelihood of future insurance loss. Credit-based insurance scores provide an
objective measurement of how one manages the risk of credit. Lending institutions, on the other
hand, use credit scores to determine the availability, amount and price of credit products offered to
the consumer. Lending institutions use credit to determine the likelihood of repayment. The most
significant difference between insurers and lending institutions is that insurers never consider
income. Insurers measure “how,” not “how much.”
In addition to income level, one’s address, ethnicity, religion, gender, familial status, nationality,
age and marital status are also not considered within a credit score calculation. Further, there is no
reliable evidence that points to insurance scoring resulting in higher insurance rates for any specific
class of individual, or that higher scores correlate with higher incomes. In fact, Federal Housing
Administration Commissioner Brian Montgomery declared in speech last year that the
administration’s data, if anything, tended to show that families with lower incomes actually have
higher credit scores.
A 2003 study by EPIC Actuaries (now part of Tillinghast), the largest and most comprehensive
study ever undertaken on the connection between credit history and insurance risk, found that a
consumer's credit-based insurance score is unquestionably correlated to that consumer's propensity
for auto insurance loss. Even more significantly, the study found that insurance scores are
consistently among the most important rating variables used by insurers. The EPIC researchers used
a multivariate analysis technique to determine indicated risk factors. After fully accounting for all
overlap and relationship with other risk factors, such as age/gender, territory, model year, driving
record and coverage limit – credit was found to clearly be an independent and significant tool for
predicting insurance loss. The propensity for loss was found to decrease as the insurance score
increases. For example, after adjusting for other variables, individuals with the lowest insurance
scores were found to incur 33 percent higher losses than average, while those with the highest
scores incurred 19 percent lower losses than average.
Every serious and reputable actuarial study on the issue, including a 2007 study by the Federal
Trade Commission, has reached the same conclusion: there is a very high correlation between
insurance scores and the likelihood of filing insurance claims. Without the ability to consider credit,
many insurers would be less aggressive in their marketing, and far more cautious in accepting new
business. Thus, consumers would quickly have fewer choices in the marketplace.
That consumers do in fact enjoy more choices in the marketplace is borne out by the massive
double-digit percentage declines in the population of state residual markets over the past 10 years.
While no definitive study of this phenomenon is available, it is the general consensus of the industry
that the decrease is directly attributable to the increased accuracy afforded by the use of insurance
scoring.
Not only do credit-based insurance scores allow insurers to offer more coverage, it also allows them
to offer coverage at. lower rates. The majority of consumers have good credit-based insurance
Honorable Michael McRaith and Kim Holland
The Use of Credit-Based Insurance Scores
April 30, 2009
3
scores and benefit accordingly – with rates refined to reduce disproportionate subsidies of higher
risk individuals. An annual survey issued by the Arkansas Insurance Department consistently finds
approximately 32 percent of policyholders enjoy a decrease in premium while approximately nine
percent of policyholders pay more because of insurance scoring (the remainder being otherwise
unaffected), a ratio of 3.44 to 1.
Credit-based insurance scoring is an effective tool for insurers - and a fair one for consumers. To
protect competition and consumer choice, it is imperative that insurers be permitted to fully price
risks using nondiscriminatory and statistically valid tools available to them.
PCI appreciates the opportunity to provide our comments on this bill, and would be happy to
address any questions you may have on this subject.
Sincerely,
Alex M. Hageli
Testimony of Neil Alldredge
On Behalf of the National Association of Mutual Insurance Companies
Joint NAIC C/D Committee Hearing on Credit-Based Insurance Scores
April 30, 2009
Good afternoon Director McRaith and Commissioner Holland, I’m Neil Alldredge, Vice
President – State & Policy Affairs for the National Association of Mutual Insurance
Companies (NAMIC). NAMIC represents 1,300 member companies that underwrite over
40 percent of the insurance market in the United States.
Before I begin, I will note that NAMIC has submitted several documents for the record.
In particular, I draw your attention to our Issue Analysis public policy paper that
examines the issue of disparate impact, unfair discrimination and the use of insurance
scoring. We have also submitted an Issue Brief titled: Credit-Based Insurance Scoring,
Separating Facts from Fallacies. This policy briefing does a good job (in 4 pages)
summarizing most of the points you have heard today.
Introduction
Credit-based insurance scores have been used by insurance company underwriters and
actuaries for nearly two decades to more accurately assess risk and price coverage for
automobile and homeowners’ insurance policies.
The use of insurance scores encourages competition and enables insurers to offer
coverage to more consumers at a fairer price. Furthermore, consumers benefit from
insurance scoring because it keeps the insurance marketplace competitive, resulting in
lower prices, better service, and more product choices.
Insurance scores provide an objective, fair, and consistent tool that insurers use with
other information to better predict the likelihood of future claims and the cost of those
claims. During the late 1990s, lawmakers and regulators in several states began enacting
laws and regulations that established procedures for insurers to follow in using an
individual’s credit information. In 2002, the National Conference of Insurance
Legislators (NCOIL) created a “Model Act Regarding Use of Credit Information in
Personal Insurance,” which became the basis for additional legislation in other states.
Today, 47 states have laws or regulations pertaining to credit-based insurance scoring.
In spite of an apparent consensus on this issue, some public officials and advocacy
groups have continued to press for further restrictions on the use of insurance scores, or
to prohibit the practice entirely. We believe this course of action is not warranted and
would be harmful to the vast majority of policyholders.
Studies
The focus of my testimony today is to review the various studies conducted on insurance
scoring. To date, 17 industry, state or federal agency studies have been conducted.
Typically these studies have examined the correlation between credit-based insurance
scores and the propensity for insured losses and/or the impact credit-based insurance
scores have on low-income or minority populations. It is certainly the most studied rating
and underwriting variable currently used by insurers.
These studies all share some common findings, primarily that credit-based insurance
scores are predictive of loss, that the majority of consumers benefit from the practice and
that credit-based insurance scores are not a proxy for race or income.
The focus of our testimony today is on the two most comprehensive studies conducted to
date, the Texas Department of Insurance study of 2005 and the study conducted by the
Federal Trade Commission released in 2007. I will also review the survey conducted by
the Arkansas Department of Insurance.
Texas
The Texas Department of Insurance (TDI) released the main body of the report in
December 2004, and issued a supplemental report in January 2005. The TDI study was
based on data obtained from six leading insurers for approximately 2 million policies. Of
these, approximately 1.2 million were for personal auto insurance and 800,000 were for
homeowners insurance. The personal auto policies covered roughly 2.5 million vehicles.
The TDI study was unusual both because of the size of its database, and because it
included individual information on race and ethnicity. That information was missing from
other studies because insurers do not collect information concerning the race or ethnicity
of their policyholders. The TDI, however, was able to draw on the resources of the Texas
Department of Public Safety and the Texas Office of the Secretary of State. Based on
data supplied by those agencies, the TDI was able to classify individual policyholders as
white, black, Asian, and Hispanic.
The December report concluded that “there appears to be a strong relationship between
credit scores and claims experience on an aggregate basis,” but cautioned that “credit
scores, to some extent, may be reflective of other risk characteristics associated with
claims.” The report explained that the department would need to perform a multivariate
analysis to determine whether credit scoring enables an insurer to predict losses more
accurately than it could by relying solely on more traditional underwriting variables. The
report also found that some minority populations were over-represented in the lower
score categories but that no unfair discrimination was detected.
A month later, the department released its supplemental report containing the multivariate
analysis. It found that “for both personal auto liability and homeowners, credit score was
related to claim experience even after considering other commonly used rating variables.
This means that credit score provides insurers with additional predictive information
distinct from other rating variables. By using credit score, insurers can better classify and
rate risks based on differences in claim experience.”
This finding so surprised then Texas Insurance Commissioner Jose Montemayor that he
felt obliged to acknowledge, in a letter to Governor Rick Perry, that his “initial suspicions
were that while there may be a correlation to risk, credit scoring’s value in pricing and
underwriting risk was superficial, supported by the strength of other risk variables.” The
study, however, “did not support those initial suspicions.” Moreover, credit scoring “is
not unfairly discriminatory as defined in current law because credit scoring is not based
on race, nor is it a precise indicator of one’s race.” A copy of the letter from
Commissioner Montemayor is included in this testimony.
Federal Trade Commission
When the federal Fair Credit Reporting Act with reauthorized in 2003 there was language
inserted directing the Federal Trade Commission (FTC) to study the impact insurance
scores had on the availability and affordability of insurance. The authorizing statute also
directed the FTC to examine whether insurance scores had a disparate impact on
protected classes. The study was made public in early 2007.
The FTC found that 59 percent of consumers benefitted from the use of credit-based
insurance scores and that scores were correlated to and predictive of loss.
The FTC stated that credit-based insurance scoring provides benefits to consumers,
including rates that are more accurate, effectively reducing subsidies which also allow
insurers to offer insurance to higher risk drivers who otherwise may not be able to obtain
coverage. The FTC also said that credit-based insurance scoring may reduce the cost of
granting and pricing insurance with the cost savings passed along to consumers in the
form of lower premiums.
The FTC found that “credit-based insurance scores appear to have little effect as a
‘proxy’ for membership in racial and ethnic groups in decisions related to insurance.”
The study noted that there was a range of credit-based insurance scores and losses within
every group studied and that insurance scores are predictive within racial groups. In other
words, every racial group has individuals with low scores and high scores and within
those groups insurance scores are predictive of loss.
Arkansas
In 2005, the Arkansas Department of Insurance began an annual survey of the effect of
the state’s insurance scoring law, which is based on the NCOIL model, on insurance
consumers. The 2007 survey concluded that of 3,026,092 personal lines policies written
or renewed in that year, 32 percent of customers received a discount, 9 percent received
an increase, and the remaining 59 percent of consumers saw a neutral impact due to
insurer use of insurance scores. In other words, 91 percent of personal lines customers
either received a discount for credit or it had no impact on premium. For policies where
credit played some role in determining the final premium, those receiving a decrease
outnumbered those receiving an increase by a ratio of 3.44 to 1. These results were
virtually identical to findings of the 2005 and 2006 survey. All the surveys are available
on the Arkansas Department of Insurance website.
Conclusion
Several other studies have been conducted by different state insurance departments over
the last 10 years (Virginia, Washington and Alaska, for example), all these studies have
two clear findings – that credit-based insurance scores are predictive of loss and that the
vast majority of consumers benefit from the tool. There have been no academic studies
that include insurance loss or rate information that have found that credit-based insurance
scores are either predictive of, or a proxy for, race or income.
We also believe a common sense examination of insurance markets is revealing. It would
stand to reason that if credit-based insurance scores have a negative effect on availability
or affordability that residual market mechanisms and consumer complaints would be
skyrocketing. Neither phenomenon is occurring. In nearly every state in the country the
personal lines markets are competitive, vibrant and healthy. Consumers have choices and
all the serious research indicates that a vast majority benefit from the use of this valuable
tool.
National Association of Insurance Commissioners
Property and Casualty Insurance (C) Committee &
Market Regulation and Consumer Affairs (D) Committee
Public hearing on Credit-Based Insurance Scores
Testimony of Charles Neeson, MAAA, ACAS
Senior Executive
Westfield Insurance
April 30, 2009
Thank you for the opportunity to speak today on Credit Based
Insurance Scores (insurance scores). My name is Charles Neeson
and I am a Senior Executive with Westfield Insurance. I am also a
member of the American Academy of Actuaries and an Associate in
the Casualty Actuarial Society. Westfield Insurance is a multi line
regional insurance company, writing both personal lines and
commercial lines insurance. We have been in business since 1848
and are proud to offer our products exclusively through professional
local independent agents. Partly because of our tag line, “Sharing
Knowledge, Building Trust” I am here today.
My prepared testimony will focus on three areas; the history and use
of insurance scores at Westfield Insurance, how insurance scores
benefit the majority of Westfield customers, and the stability of
insurance scores for Westfield Insurance customers.
Insurance is an incredibly competitive business. One way an
insurance company, such as Westfield, can distinguish itself from its
competitors is to find better ways to price its business. That means
improving the accuracy of premium as an estimate of expected future
loss. When insurers are able to properly underwrite risks, consumers
benefit with lower rates and more choices.
Prior to our use of insurance scoring, Westfield utilized traditional
classification variables like vehicle use and vehicle performance to
help estimate future risk of loss. In 1999, we conducted research on
insurance scoring, to determine whether it would benefit the company
and our customers. In analyzing the relationship between credit
information and our loss data, we found a strong correlation. We also
found it did not replace traditional classification variables, but worked
well with them.
(Appendix 1 shows the relationship of several classification variables,
both traditional and credit, to loss.)
Based upon that research, Westfield Insurance began using
insurance scores in 2000 as part of its pricing of automobile and
homeowners’ insurance. Used in conjunction with more traditional
rating factors such as vehicle performance, age, territory and prior
claims, credit-based insurance scoring allowed Westfield to more
accurately price its products, improve its competitive position and
write more business.
(Appendix 2 shows how insurance scoring and prior claims can work
together.)
Today, about 90% of Westfield auto-home customers either benefit or
have no impact from insurance scoring, indeed, approximately 75%
pay less.
Westfield’s introduced insurance scoring in 2000. As part of the roll
out, we provided training to help our agents effectively communicate
with customers. The training included information about credit
reports and insurance scoring. We followed up at several later dates
in group meetings and seminars, plus we developed literature and job
aids.
We also tracked how our customers accepted our new pricing that
included credit based insurance scoring. We were pleased to find
that not only did new business increase, but we also saw an
improvement in retention on policies renewed with insurance scoring.
(Appendix 3 shows how Westfield Insurance customers benefited and
how their retention improved with insurance scoring.)
I cannot say insurance scoring has been without its unique
challenges. For example, during the early days of its use, some
agents spoke up about the additional effort needed to quote multiple
companies. Each year though, I have heard fewer and fewer issues
raised.
More recently, some parties worry poor current economic conditions
may cause a steep decline in credit based insurance scores. Data
shows these worries to be unfounded.
All of the recent score stability studies done by credit bureaus show
no decline in FICO credit based insurance scores. In fact, they show
the trend to be flat or slightly improving. Westfield Insurance data
supports this conclusion as well.
(See Appendix 4 showing the stability of insurance score pricing tier.)
Considering the troubling economic times we are in, it is good to
know insurance scoring continues to benefit so many people through
lower insurance premiums.
Thank you for allowing me to testify before you today. I would be
happy to address any questions you may have on this subject.
WRITTEN STATEMENT OF WESLEY BISSETT
SENIOR COUNSEL, GOVERNMENT AFFAIRS
INDEPENDENT INSURANCE AGENTS & BROKERS OF AMERICA
BEFORE THE
NATIONAL ASSOCIATION OF INSURANCE COMMISSIONERS
PUBLIC HEARING ON CREDIT-BASED INSURANCE SCORES
APRIL 30, 2009
On behalf of the Independent Insurance Agents and Brokers of America (IIABA), the nation’s
oldest and largest association of insurance producers, I am privileged to offer the association’s
outlook on the manner in which the insurance industry uses credit information. IIABA
represents a network of more than 300,000 agents, brokers, and employees nationwide, and
our members provide insurance products to and serve the insurance needs of millions of
American consumers. The independent agent community brings a unique perspective to this
issue, largely because we work with our insurance company partners while remaining sensitive
to and focused on the needs and concerns of consumers.
Background and Historical Context
This is not a new issue for the association, and we appreciate having the opportunity to briefly
discuss the issue today. For more than 12 years, the Big “I” has worked with insurers on
business issues and concerns related to the use of credit information, and, throughout that time,
we have also been incredibly active in the public policy arena.
In preparing for Thursday’s hearing, I was reminded that the NAIC’s Market Conduct and
Consumer Affairs Committee held a similar public hearing examining the industry’s use of credit
histories and credit scoring back in December 2001. Much has changed since that time. While
financial services providers had long used credit data and credit history to evaluate loan
applications, determine creditworthiness, and predict the likelihood of default and delinquency,
the insurance industry was only beginning to widely utilize similar data to predict future losses
and claim costs and to determine prices in a more accurate manner. One study at the time
found that approximately one-half of the 100 largest personal automobile insurers in the country
had only begun utilizing credit histories and scores since 1998, so the use of such information
was relatively new and novel.
The somewhat sudden increase in the use of credit information produced what might
generously be descried as “growing pains,” and many of these were self-inflicted by the insurer
community at the time. Some carriers did a poor job of educating agents and the public about
the use of credit data, and there were documented abuses and poor business practices that
generated concern among those in the agent community. The exclusive reliance on and
inflexible use credit information by some carriers, the lack of transparency and meaningful
disclosure to consumers, the negative effects on those with no credit history, and similar
questions and problems created skepticism and doubts among many.
During this time, the use of credit information was a regular topic of discussion among IIABA’s
leadership structure and public policy-related committees, and agents were voicing many of the
same concerns that they were hearing from their clients. The often contentious debate and
controversy that existed, however, was largely alleviated by three important developments:
• First, personal lines insurers began to do a much better job educating consumers,
agents, and other stakeholders about the use of credit information, and this had the
effect of increasing awareness and understanding.
• Second, many carriers reevaluated the manner in which credit data was being utilized
and ultimately implemented more reasonable business practices that considered the
interests of consumers and addressed many of the legitimate criticisms that were being
made.
• Third, due in large part to the efforts of the NAIC and state officials (and with the strong
support of IIABA), comprehensive and effective regulation and meaningful restrictions
and limitations on the use of credit information were implemented in nearly every state.
The statutory enactments and business practice reforms put in place since the early part of this
decade have dramatically improved the manner in which credit information is utilized in the
underwriting and rating process and have dramatically reduced the level of consternation that
previously existed. While limited problems and unfortunate anomalies may still arise on a
periodic basis, IIABA rarely receives complaints today from its members about the use of credit
information. This is a dramatic change from the state of affairs eight years ago and reflects the
positive changes and reforms outlined above.
Additional Observations
IIABA supports the use of underwriting and rating tools that foster enhanced competition and
the fair and accurate pricing of risk and recognizes that consumer credit information is a
powerfully predictive tool when used appropriately. The effectiveness of utilizing credit
information has become increasingly apparent and widely accepted, even to those who were
previously critical of its use, and agents can attest to the fact that it enables insurers to more
accurately predict losses and the severity of future claims. The increased use of credit-based
insurance scores has enhanced competition as companies have become more confident with
the accuracy of their underwriting and rating tools, and, as a result, many agents are now able
to find coverage (and prices) for clients in instances where such options were unavailable in the
past.
At the same time, however, independent agents and brokers believe credit-based insurance
scores must be used in sensible, responsible, and consumer-friendly ways – and IIABA has
supported and helped implement a meaningful series of consumer protections at the state level.
Most states have now enacted restrictions that limit when and how credit information and scores
may be used in the insurance arena. These safeguards, for example, require additional
underwriting factors to be taken into consideration when evaluating whether to underwrite, deny,
cancel, or non-renew a policy; protect those with little or no credit history; impose helpful
disclosure requirements; and restrict the use of certain types of factors or credit information.
These and other critical measures have proven to be highly successful.
IIABA believes that effective regulation of the use of credit information and credit-based
insurance scores by the insurance industry helps ensure that this powerful tool is used in a
reasonable and proper manner. State policymakers have enacted comprehensive legislation
that strikes the appropriate balance between the concerns of consumers and the needs of the
industry and are considering additional steps. Insurance agents and brokers believe credit-
based insurance scores are an effective, objectively verified, and fair risk measurement tool,
and IIABA strongly opposes any efforts to ban the use of this information or unnecessarily
restrict its use.
Conclusion
While it is unclear what, if any, future discussion and deliberation concerning these issues may
occur at the NAIC level, IIABA looks forward to assisting you in any manner that we can. We
have a strong interest in this subject matter, considerable experience with legislative action in
this area, and insight as to what is happening on the ground floor of the marketplace. Thank
you again for the opportunity to appear before your committees.
TESTIMONY OF
J. ROBERT HUNTER,
DIRECTOR OF INSURANCE,
CONSUMER FEDERATION OF AMERICA
BEFORE
SUBCOMMITTEE ON OVERSIGHT AND INVESTIGATIONS
OF THE
COMMITTEE ON FINANCIAL SERVICES
OF THE
UNITED STATES HOUSE OF REPRESENTATIVES
REGARDING
THE IMPACT OF CREDIT-BASED INSURANCE SCORING ON
THE AVAILABILITY AND AFFORDABILITY OF INSURANCE
MAY 21, 2008
1
Good morning Mr. Chairman and members of the Subcommittee. Thank you for inviting
me here today to discuss the impact of credit-based scoring on the availability and affordability
of insurance. And thank you for all you are doing for the many consumers of insurance who are
being harmed by the use of credit scoring today. My name is Bob Hunter and I am the Director
of Insurance for the Consumer Federation of America (CFA). CFA is a non-profit association of
300 organizations that, since 1968, has sought to advance the consumer interest through research,
advocacy and education. I am a former Federal Insurance Administrator under Presidents Ford
and Carter and have also served as Texas Insurance Commissioner. I am also an actuary, a
Fellow of the Casualty Actuarial Society and a member of the American Academy of Actuaries.
I am testifying on behalf of CFA and the Center for Economic Justice.
1
At your last hearing on this subject, testimony was delivered by Birny Birnbaum, the
Executive Director of the Center for Economic Justice. A statement on insurance credit scoring
was also submitted by CFA, Consumers Union, National Council of LaRaza, National Consumer
Law Center, and National Fair Housing Alliance. Today, I will touch on a number of the
concerns raised in the testimony and statement, which are attached.
KEY FINDINGS
Insurance scoring occurs when insurers use consumer credit information to determine
whether a person is eligible for coverage, which company affiliate will offer the coverage, the
“rate tier” at that company in which the person will be placed and, finally, the premium the
consumer will pay. Insurance scoring is used by nearly all insurers and has grown to become
one of the most important factors in determining a consumer’s automobile and homeowners
insurance premium. Insurance scoring is typically done through the use of a computer model
that converts information in a consumer’s credit report into a score, or numerical value.
Many organizations have called for a prohibition on insurers’ use of consumer credit
information for underwriting and ratings. These groups include not only consumer
organizations, but civil rights groups, several associations representing insurance agents and
some insurers. The case for such a prohibition is strong. There is more than enough information
currently available to justify such a prohibition. A closer look at insurance scoring reveals that
the practice has the following serious flaws:
• Undermines core functions of the insurance system by decreasing insurance
availability and affordability, and undermining the critical role of insurance in
encouraging loss prevention;
• Has an adverse, disparate impact on low income and minority consumers and is
discriminatory;
• Is based on credit reports that often have erroneous or incomplete information;
1
Center for Economic Justice is a Texas-based non-profit organization that advocates on behalf of low income and
minority consumers on insurance, credit and utility issues.
2
• Is inherently unfair and penalizes consumers who are the victims of economic,
medical or natural catastrophes;
• Penalizes consumers because of the business decisions of lenders.
The insurance industry maintains that there are a variety of benefits from their use of
credit scoring. Upon examination, these assertions are illusory and contradicted by the available
evidence. Ultimately, however, all of the insurer arguments for insurance scoring come down to
a single point: insurance scoring is predictive of the likelihood that a consumer will have a claim
and consumers will benefit if insurers are able to price more accurately.
The problem with this contention is that insurers cannot tell us what it is about a credit
score that is linked with risk. If you ask proponents of the use of credit scoring to explain to a
person who suffered a decline in credit as a result of being in Hurricane Katrina, or lost her job
because of outsourcing, or lost his job in the current economic downturn, why these events that
they had no control over made them a worse auto or home insurance risk, they have no response..
Unlike insurance classifications that were in use before credit scoring was adopted, credit
scoring is not based on an appropriate thesis, confirmed by a statistical analysis. In fact, there is
no legitimate thesis for the use of credit scoring. There is only an alleged correlation based on
proprietary information not open to public scrutiny.
2
However, a correlation in search of an
appropriate thesis raises serious questions about the classification that is being used.
The lack of a thesis means that credit scoring violates actuarial principles. Some
actuaries say that a thesis is not required because actuarial principles state that a cause and effect
relationship is not required. Although this is true, the principles, which were developed by a
group of mostly industry-employed actuaries with an overwhelming industry bias, also say that a
thesis -- a logical underpinning for the use of the information -- is required. Here is what the
principles say, in relevant part, on this subject:
Classification characteristics may be more acceptable to the public if there is a
demonstrable cause and effect relationship between the risk characteristic and
expected costs. However, in insurance it is often impossible to prove
statistically any postulated cause and effect relationship. Causality cannot,
therefore, be made a requirement for risk classification systems.
Often causality is not used in its rigorous sense of cause and effect but in a
general sense, implying the existence of a plausible relationship between the
characteristics of a class and the hazard insured against. Living in a river valley
would not seem to cause a flood claim, but it does bear a reasonable relationship
2
This is another difference from all previous classes where the data is public and part of rate filings made with
insurance departments. Previously, an insurer would propound a thesis and test it with the data. If a thesis was
confirmed, the insurer would file for a new class with the commissioner showing the thesis and the data in the rate
filing. An example was the use of accidents and tickets. The thesis was that people with more accidents and tickets
would be worse drivers in the future because their historic driving record indicated less care in driving. The thesis
was confirmed by data that can be viewed in its’ entirety in rate filings.
3
to the hazard insured against and thus would be a reasonable basis for
classification.
Risk classification characteristics should be neither obscure nor irrelevant to the
insurance provided; but they need not always exhibit a cause and effect
relationship.
3
Credit scoring is at best obscure relative to auto and home insurance, if not downright
irrelevant. Since there is no clear relationship, no thesis, underlying credit scoring, the
classification violates actuarial principles.
4
Some in the industry appear to believe that a correlation between the classification and
the risk of loss is all you need to create a class, despite the principles. Taken to its logical
extreme, this point-of-view would indicate that race should be used if a correlation existed.
Obviously, this is wrong from a public policy perspective. The fact that credit scoring triggers
the indirect use of race for insurance underwriting and rating purposes makes it no more socially
acceptable. Policymakers need to control the use of such illegitimate classes. Congress should
do so since the insurance industry lobby is too strong to overcome in many states.
In fact, there is strong evidence that insurance scoring itself is not a predictor of risk or
insurance claims, but, rather, that insurance scoring is a proxy for other factors that are related to
claims experience, such as the income, miles driven, or geographic location of the consumer. In
particular, insurance scoring is a proxy for race and income. Two independent studies by the
Texas and Missouri Departments of Insurance found a strong relationship between insurance
scores and race and income.
5
The Missouri study found the single most predictive factor of an
insurance score was race.
Even the recent substandard report of the Federal Trade Commission (FTC) on the use of
automobile insurance scores, despite relying upon data hand-picked by the insurance industry,
found insurance scores were worse on average for African-Americans and Hispanics and that
insurance scoring was a proxy for race. Had the FTC actually used an independent and
comprehensive set of insurance data, the measured negative racial impact would likely have been
much greater.
6
Although the FTC report discounts its own findings and plays down the
possibility of racial discrimination, the strong evidence of an adverse, disparate racial impact
from insurance scoring justifies a prohibition on its use. Insurers should not be permitted to use
a proxy for race when the direct use of race itself for underwriting or rating is prohibited.
3
Risk Classification Statement of Principles, American Academy of Actuaries Committee on Risk Classification, at
http://actuarialstandardsboard.org/pdf/appendices/risk.pdf.
4
There are other actuarial principles that credit scoring violates as well, including the fact that it is not socially
acceptable, is subject to manipulation (there are firms that offer, for a fee, to sharply improve your score), and is
ambiguous.
5
Texas Department of Insurance, “Report to the 79
th
Legislature: Use of Credit Information in Texas,” December
30, 2004, page 3. “Insurance-Based Credit Scores: Impact on Minority and Low Income Populations in Missouri,”
State of Missouri Department of Insurance, January 2004.
6
Credit-Based Insurance Scores: Impacts on Consumers of Automobile Insurance,” Federal Trade Commission,
July 2007, at http://www.ftc.gov/os/2007/07/P044804FACTA_Report_Credit-Based_Insurance_Scores.pdf
.
4
In fact, I would strongly encourage the Subcommittee to continue to critically evaluate
the FTC credit scoring analysis of automobile insurance scoring, which is deeply flawed and
unresponsive to its Congressional mandate. The problems with the report include the failure of
the FTC to obtain a comprehensive and independent data set for analysis and the agency’s
reliance upon a data set hand-picked by the insurance industry. The report also lacks any
substantive analysis of the impact of insurance scoring on the availability and affordability of
insurance products as requested by Congress, ignores evidence indicating that the correlation
between insurance scores and claims is spurious, and fails to analyze the false claim that the use
of insurance scoring is legitimate because people who manage their finances well are likely to
manage other risks well.
7
The FTC passed a resolution on May 16 that could lead to a better data collection process
for the home insurance scoring study that is now underway. However, given the serious flaws
detailed above with the automobile insurance report, we continue to have significant concerns
about the FTC’s ability and willingness to conduct a thorough, unbiased review of the impact
credit scoring on those who purchase home insurance.
Insurers also claim that competition would be harmed and that the availability of
insurance would be curtailed if credit scoring was banned. This is a false claim. I need only to
point to California, where credit scoring is banned from use in auto insurance. In CFA’s recent
in-depth study of auto insurance regulation,
8
we found that the state had the best system of
regulation in the nation. In particular, California is a leader in protecting consumers from
abusive class systems. Rate increases in California were the lowest in the nation over the period
we studied. More importantly, despite claims by insurers that a credit scoring ban would harm
competition, California had the fourth most competitive automobile insurance market. Further,
the number of Californians who were required to receive insurance for the state’s high-cost
assigned risk plan was very low; only 0.1 percent of the state’s automobiles were insured in the
plan. The California system proves that robust competition and insurance availability can occur
without the use of credit scoring.
LEGISLATION BEFORE THE COMMITTEE
H.R. 5633 -- Gutierrez
CFA very much appreciates the efforts of the sponsors of this bill to curb the inappropriate
use of insurance scoring. We support the legislation’s goal to ban insurance credit scoring if the
use of consumer credit information for insurance underwriting or rating discriminates on the
7
The fact is that, by the credit modelers own admission, fully 20 percent of the population is unscorable with
traditional credit reports because of little or no information in the files. These individuals are disproportionately low
income and minority consumers who get charged higher rates through no fault of their own. Even a cursory
examination of actual scoring models reveals that many of the factors determining an insurance score have nothing
to do with whether a consumer pays his or her bill on time, but with factors related to socio-economic status. Yet,
the FTC report dutifully repeats this rationalization for insurance scoring with no critical analysis.
8
State Automobile Insurance Regulation: A National Quality Assessment and In-depth Review of California’s
Uniquely Effective Regulatory System, April 24, 2008 at
http://www.consumerfed.org/topics.cfm?section=Finance&Topic=Insurance&SubTopic=Insurance%20Regulation
.
5
basis of race or ethnicity. However, as written, we fear that the legislation will not achieve the
desired goal:
• The bill could serve to legitimize insurers’ use of credit-based insurance scoring so long
as the use of the scoring methodology was not found to be discriminatory.
• The bill establishes the FTC as the arbiter of determining racial discrimination, although
the agency has virtually no track record or enforcement experience in this area. In fact,
the FTC study demonstrated a severe bias against consumers in favor of insurers
regarding insurance scoring. We do not trust the FTC to fairly make impartial findings
relative to credit scoring. To give just one example of the agency’s bias, Congress asked
the FTC to study the impact of insurance scoring on the availability and affordability of
automobile insurance. Instead of getting data on applications for coverage that resulted in
policies being issued or rejected from a large number of insurers serving all parts of the
market, the FTC relied upon data handpicked by the industry from a few companies for
only the policies they issued. Thus, the FTC had no ability to determine whether
insurance scoring resulted in large numbers of consumers being denied coverage, priced
out of the market, or charged higher premiums. Yet, despite this obvious limitation, the
FTC concluded that credit scoring was a benefit to the majority of consumers. The data
problem was brought to the FTC's attention early on, yet despite offers of assistance from
state insurance regulators and a period of three years to do the study, the FTC was
apparently satisfied to let insurers exercise undue influence over the study through their
control of the data.
• The bill lacks an objective standard for identifying racial discrimination, again giving
broad discretion to the FTC. As written, the proxy effect language does not clearly and
adequately incorporate the legal concept of disparate impact. Under the bill, the FTC
could find some statistical correlation to race and income and some proxy effect, but
determine that this effect is not substantive and conclude that no discrimination or proxy
effect exists. The bill should prohibit BOTH systems that incorporate racial proxies and
those that have unlawful disparate impacts.
• To make determinations of discrimination and proxy effect, Congress should vest
authority with agencies that have the experience and jurisdiction to regulate insurance
and enforce anti-discrimination laws. State insurance departments and the National
Association of Insurance Commissioners, who are already authorized to collect the
necessary data and take corrective regulatory action, should be allowed to make these
determinations. If any federal agency is given authority to make these determinations,
the U.S. Department of Justice, not just the FTC, should also be provided with
jurisdiction.
• The bill makes no provisions for a private right of action. If the FTC has the final say,
there is no recourse for anyone who wants to challenge the racially discriminatory use of
credit in insurance. This would be a significant problem for civil rights groups and
individual consumers who wish to challenge this practice in the future.
6
• The bill is unclear about what types of state insurance regulation are or are not pre-
empted. Although the bill strives to not pre-empt stricter state laws on insurance scoring,
the legislation vests authority with federal agency -- the task of identifying and stopping
unfair discrimination – that has traditionally been the role of states.
• The bill does not provide timely assistance for the millions of consumers who are facing
higher auto and homeowners insurance rates now because their credit scores have been
negatively affected by abusive and reckless lending practices.
We believe it would be simpler to ban the use of consumer credit information for insurance.
In the near term, we would encourage you to consider legislation to at least impose a temporary
"freeze" on the use of this information by insurers during the current mortgage crisis.
HR 6062 – Waters
CFA supports the bill but we seek clarification on one aspect of the bill.
Since the bill declares that some type of reports, such as motor vehicle records, Comprehensive
Loss Underwriting Exchange (CLUE), and medical history records are not consumer reports for
purposes of the section, is there any chance that, the way bill is written, it could be interpreted as
eliminating adverse action notification for insurers' use of non-credit consumer reports? It
should be clarified if there is any chance of such an interpretation.
CONCLUSION
Credit scoring is harmful to consumers, particularly low income and minority consumers.
Millions of consumers are threatened with foreclosures and a variety of financial stresses
resulting from the sub prime lending crisis, the resulting credit crunch, and the loss of jobs in the
current weak economy. It is clearly unfair for millions of consumers to experience higher auto
and homeowners’ insurance rates because of reckless and abusive practices by lenders or
because of conflicts between lenders and bondholders, which are preventing foreclosure
assistance. As part of the package of assistance to consumers in financial distress, a ban, or, in
the short term, a moratorium on insurance scoring should be enacted.
Credit scoring also undermines the very foundation of a sound insurance system, which involves
the use of broad, risk-spreading classes tied to risk factors understandable by consumers that
promote loss prevention.
It is time to ban the use of these unfair classes. It is time to pass H.R. 6062.
7
ATTACHMENT 1
Written Testimony Before the
Subcommittee on Oversight and Investigations
Financial Services Committee
U.S. House of Representatives
October 2, 2007
The undersigned civil rights and consumer organizations applaud Chairman Watt and members
of the Subcommittee on Oversight and Investigations for holding this hearing on Credit-Based
Insurance Scores: Are They Fair? This statement is intended to supplement the written
testimony submitted by the Center for Economic Justice and the National Council of La Raza.
Unknown to most consumers, insurers’ use of consumer credit information has spread to almost
all insurers and is one of the most important factors in determining how much a consumer pays
8
for auto or homeowners insurance. Insurance companies use credit scores – three digit numbers
generated using a consumer’s credit report – in insurance underwriting and rate setting. This
practice creates wide racial disparities as previous studies have found. Nevertheless, much of the
insurance industry relies on credit scoring because it is allegedly predictive in forecasting which
consumers will have higher loss ratios. Yet the industry has not been able to provide credible
explanation as to why there is a correlation between credit scores and loss ratios.
For these reasons, we echo the call of many organizations and public officials for a prohibition
on insurance scoring and insurers’ use of consumer credit information for underwriting and
ratings purposes.
Before the introduction of the credit scoring systems, the insurance industry had used other
unsupported standards and stereotypes with a racial proxy effect. After the major companies
were sued for fair housing violations and were forced to eliminate these practices, the industry
introduced a new practice – credit-based insurance scoring – that consumer and civil rights
groups see as re-introducing unfair racial and ethnic impacts into the pricing of insurance.
Previous studies by the Missouri and Texas Departments of Insurance have found that insurance
scoring discriminates against low income and minority consumers because of the racial and
economic disparities inherent in scoring. The Missouri study concluded that a consumer’s race
was the single most predictive factor determining a consumer’s insurance score and,
consequently, the consumer’s insurance premium.
We were pleased that Congress, through the inclusion of Section 215 of the Fair and Accurate
Credit Transactions Act of 2003, directed the Federal Trade Commission in conjunction with the
Federal Reserve Board to study the impact of credit scoring on the availability and affordability
of credit and insurance and to determine whether credit scoring was truly related to insurance
losses or simply a proxy for race, income or other factors. The FTC conducted the insurance
scoring component of this research.
Unfortunately, we find that the FTC study is fatally flawed in key areas and is not responsive to
the Congressional mandate contained in the FACT Act. Most critically, instead of requiring the
submission of comprehensive policy data by a large number of insurers, the FTC allowed the
insurance industry to self-select the data for analysis. Thus the industry was unnecessarily
afforded an opportunity to control the outcome of the study.
Even so, the FTC study found that insurance scores were worse on average for African
Americans and Latino consumers, although this finding is downplayed in the report. The study
also confirms that despite the growing reliance on credit-based insurance scores, there was no
evidence to prove a causal connection between a consumer’s score and auto insurance losses.
Without the need to demonstrate such a connection, insurers could use any consumer
characteristic, such as hair color, to price insurance products.
The FTC report acknowledges that the alleged correlation between risk and credit-based
insurance scores might be explained by other factors. Instead of pursuing these other factors, the
FTC employed subjective and pejorative racial stereotypes to try to support the alleged link
9
between credit-based insurance scores and legitimate risk. Thus the FTC report mimics the
insurance industry blaming-the-victim rationalization of claiming credit history is related to
responsibility and risk management. A look at the actual scoring models shows that socio-
economic factors have more impact on the score than loan payment history and that an insurance
credit score has little to do with personal responsibility and everything to do with economic and
racial status.
In short, there is ample evidence to justify banning credit-based insurance scores. Moreover,
given the biased and flawed nature of the FTC study on scoring for auto insurance, the
undersigned organization encourages Congress to consider assigning responsibility to conduct
the homeowners scoring study to another agency, such as the U.S. General Accountability
Office, which could then work in conjunction with state insurance regulators who have the
necessary authority to obtain the desired data set from the insurance industry.
###
Center for Economic Justice is a Texas-based non-profit organization that advocates on behalf
of low income and minority consumers on insurance, credit and utility issues
Consumer Federation of America is a nonprofit association of some 300 pro-consumer groups,
with a combined membership of 50 million people. CFA was founded in 1968 to advance
consumers' interests through advocacy and education. www.consumerfed.org
National Consumer Law Center is a non-profit organization specializing in consumer issues on
behalf of low-income people. NCLC recently released Credit Scoring and Insurance: Costing
Consumers Billions and Perpetuating the Economic Racial Divide, available at
www.consumerlaw.org.
National Council of La Raza is a private, nonprofit, nonpartisan organization established in
1968 to reduce poverty and discrimination and improve opportunities for the nation’s Hispanics.
As the largest national Latino civil rights and advocacy organization, NCLR serves all Hispanic
nationality-groups in all regions of the country through a network of more than 300 affiliate
community-based organizations.
National Fair Housing Alliance is a consortium of more than 220 private, non-profit fair
housing organizations, state and local civil rights groups, and individuals from 37 states and the
District of Columbia. Headquartered in Washington, DC and founded in 1988, NFHA, through
comprehensive education, advocacy and enforcement programs, provides equal access to
housing for millions of people.
Consumers Union of U.S., Inc. Consumers Union (CU) is an expert, independent, nonprofit
organization, whose mission is to work for a fair, just, and safe marketplace for all consumers.
CU publishes Consumer Reports and ConsumerReports.org in addition to two newsletters,
Consumer Reports on Health and Consumer Reports Money Adviser with combined
subscriptions of more than 7 million. Consumers Union also has more than 500,000 online
activists who help work to change legislation and the marketplace in favor of the consumer
10
interest and several public education Web sites. Since its founding in 1936, Consumers Union
has never taken any advertising or freebies of any kind. The organization generates more than
$160 million in revenue and a staff of more than 500 work at either CU's 50 state-of-the-art labs
in Yonkers, NY; its 327-acre auto test facility in East Haddam, CT.; or the three advocacy
offices in Washington DC, Austin, TX, and San Francisco, CA.
11
ATTACHMENT 2
Testimony Before The
House Financial Service Committee
Subcommittee on Oversight and Investigations
Credit-Based Insurance Scores: Are They Fair?
October 2, 2007
Birny Birnbaum
Executive Director
Center for Economic Justice
1701 A South Second Street
Austin, TX 78704
(512) 912-1327
1. Introduction
Chairman Watt, Ranking Member Miller and Members of the Committee:
Thank you for the opportunity to discuss insurers’ use of consumer credit information for auto
and homeowners insurance. My name is Birny Birnbaum and I am the Executive Director of the
Center for Economic Justice, an Austin, Texas-based non-profit that advocates on behalf of
consumers on insurance, credit and utility matters.
I have been working on insurance credit scoring issues since 1991 as both an insurance regulator
– Chief Economist and Associate Commissioner for Policy and Research at the Texas
Department of Insurance – and as a consumer advocate. I have testified about insurance credit
scoring before legislatures and administrative agencies, including insurance departments and
public utility commissions, and provided expert testimony in litigation related to insurance credit
scoring. I received my formal training in economics from the Massachusetts Institute of
Technology and have been accepted as an expert on both economic and actuarial matters related
to auto and homeowners insurance rates and risk classification.
12
2. Summary of Testimony
Insurance scoring is the use by insurance companies of consumer credit information to determine
whether a consumer is eligible for coverage, the types and amount of coverage offered to a
consumer and the premium charged to the consumer. The use of insurance scoring has grown to
become one of the most important factors in determining a consumer’s auto and homeowner’s
insurance premium and is used by almost all insurers. Insurance scoring is typically done
through the use of computer model that converts information in a consumer’s credit report into a
score, or numerical value, which is then used as an underwriting or rating factor.
Many organizations have called for a prohibition on insurance scoring and insurers’ use of
consumer credit information for underwriting and rating. These groups include not only
consumer organizations, but civil rights groups, insurance agents’ groups and some insurers.
The case for such a prohibition is strong – there is more than enough information currently
available to justify such a prohibition. A closer look at insurance scoring reveals that the
practice
• Undermines core functions of insurance system by worsening insurance availability
and affordability and undermining the critical role of insurance in encouraging loss
prevention;
• Discriminates against low income and minority consumers;
• Is arbitrary and unrelated to how well a consumer "manages" her finances;
• Is inherently unfair and penalizes consumers who are the victims of economic or
medical or natural catastrophes;
• Penalizes consumers because of the business decisions of lenders.
The insurance industry claims a variety of benefits from their use of credit scoring. Upon
examination, these claims are illusory and contradicted by the available evidence. Ultimately,
however, all of the insurer arguments for insurance scoring come down to a single claim:
insurance scoring is predictive of the likelihood of a consumer having a claim and consumes
benefit if insurers are able to price more accurately.
There is, however, strong evidence that insurance scoring itself is not a predictor of risk or
insurance claims, but, rather, that insurance scoring is a proxy for some other factor or factors
that are truly related to claim experience. In particular, insurance scoring is a proxy for race and
income. Two independent studies by the Texas and Missouri Departments of Insurance found a
strong relationship between insurance scores and race and income. The Missouri study found the
single most predictive factor of an insurance score was race. Even the recent flawed and biased
FTC report on insurance scoring – despite relying upon data hand-picked by the insurance
industry – found insurance scores were worse on average for African-Americans and Hispanics
and that insurance scoring was a proxy for race. And had the FTC actually used an independent
13
and comprehensive set of insurance data, the measured racial discrimination would have been
much greater. Although the FTC report discounts its own findings and plays down the
importance of racial discrimination, the finding of racial discrimination from insurance scoring
justifies a prohibition. Insurers should not be permitted to use a proxy for race when the direct
use of race itself for underwriting or rating is prohibited.
The FTC analysis of insurance scoring is deeply flawed and the report is unresponsive to its
Congressional mandate. The problems include:
1. The failure to obtain a comprehensive and independent data set for analysis and the
reliance upon a data set hand-picked by the insurance industry. The insurance
industry effectively controlled the study by dictating the data that would be used in
the study.
2. No substantive analysis of the impact of insurance scoring on the availability and
affordability of insurance products as requested by Congress. Because of its reliance
on industry-selected data, the FTC performed no analysis of how consumers actually
fared from insurers’ use of credit scoring.
3. Regurgitating insurer claims about credit scoring despite evidence that contradicts
these claims. The FTC ignored evidence indicating that the correlation between
insurance scores and claims was a spurious correlation – that insurance scoring was a
proxy for some other factor actually related to claims.
4. The failure to analyze the "blaming-the-victim" strategy used by insurers to justify
insurance scoring -- the bogus claim that people who manage their finances well are
likely to manage their risks well and that's why credit scoring works. The fact is that,
by the credit modelers own admission, fully 20% of the population is unscorable with
tradition credit reports because of little or no information in the files. These folks are
disproportionately low income and minority consumers who get charged higher rates
through no fault of their own. And even a cursory examination of actual scoring
models reveals that most of the factors determining an insurance score have nothing
to do with whether a consumer pays her bill on time, but with factors related to socio-
economic status. Yet, the FTC report dutifully repeats this desperate rationalization
for insurance scoring with no critical analysis.
14
5. The failure to examine any alternatives to insurance scoring that are predictive of
claims but are not based on any consumer credit information. The FTC ignored
research indicating that insurers could eliminate the use of credit information but
obtain the same ability to predict claims with advanced modeling and data mining of
traditional rating factors. Consequently, the FTC ignored an obvious alternative to
insurance scoring that could reduce the impact on low income and minority
consumers.
There is no need for further study of insurance scoring to justify its prohibition. The problems
with insurance scoring are well documented and the alleged benefits claimed by insurers are
illusory. However, if Congress does want additional study, it has become clear that the FTC
should not be doing that analysis. The FTC has not only revealed a strong bias toward the
insurance industry in the July report on auto insurance, but has indicated it remains willing to
allow the insurance industry to control the data for an analysis of insurance scoring for
homeowners insurance. Congress should turn to the Government Accountability Office and state
insurance regulators for any additional research on insurance scoring. The active involvement of
state insurance regulators is particularly important for two reasons. First, state insurance
regulators have authority to obtain data from insurance companies and the use of a
comprehensive and independent data set is crucial to an unbiased analysis. Second, insurance
scoring is primarily regulated by the states. State insurance regulators should be the most
knowledgeable about how insurance scoring is used and how it impacts the availability and
affordability of insurance.
The remainder of my testimony expands upon these points.
3. Insurance Credit Scoring is an Unfair Practice
Insurance credit scoring is the practice by insurers of using consumers’ credit information for
underwriting, tier placement, rating and/or payment plan eligibility. The problems with
insurance scoring are so great that the practice should be prohibited. Insurance scoring should be
prohibited because it:
• is inherently unfair;
• has a disproportionate impact on consumers in poor and minority communities;
• penalizes consumers for rational behavior and sound financial management practices;
• penalizes consumers for lenders’ business decisions unrelated to payment history;
• is an arbitrary practice; and
• undermines the basic insurance mechanism and public policy goals for insurance.
There is widespread opposition to insurance credit scoring among consumers and insurance
agents. There are hundreds of agents who want to come forward and tell why they are opposed
to insurance credit scoring, why insurance credit scoring has worsened insurance availability and
how insurance credit scoring has a disproportionate impact on poor and minority consumers.
But they can’t tell their stories because of their fear of reprisal by the insurance companies they
represent. To hear from these agents, the agents must be given protection against these reprisals.
To give you a sense of who these agents are, the following agent organizations have come out
15
against insurance credit scoring – National Association of State Farm Agents, National
Association of Professional Allstate Agents and the United Farmers Agents Association.
Insurance Scoring is Inherently Unfair
You’ve just been laid off from your job. Or your daughter has a major medical problem that
your health insurance (if you have any) doesn’t fully cover. Or you’ve just gotten a divorce.
These three life events account for 87% of family bankruptcies.
9
To “help” you out in this
stressful time, your insurance company will raise your homeowners and auto insurance rates
because of insurance credit scoring.
The disagreements about insurance credit scoring really boil down to what “fair” means. For
insurers, “fair” means that an insurer can produce some kind of data showing a statistical
relationship between credit scores and insurance losses. For consumer groups, such a statistical
relationship is a necessary, but not sufficient, definition of fair insurance practices. Fair rating
factors must also not penalize consumers for rational behavior, for factors outside of their control
and for arbitrary practices of insurers and lenders. Fair means that consumers who are the
victims of some economic or medical catastrophe are not penalized because they were unlucky
enough to lose their jobs, have a family member get sick or get divorced.
When it comes to the real world understanding of fair, insurance credit scoring is terribly unfair.
• Because your credit score depends on having the “right” kind of information in your
credit report, you can have a perfect credit history and still get a bad credit score.
Contrary to insurer credit scoring myths, your credit score has nothing to do with
your “financial responsibility.”
• Because your credit report can vary dramatically among the three major credit
bureaus, your credit score can vary from good to bad depending upon which bureau
provided your insurer with information.
• Because your credit score is based on many things other than how timely you pay
your loans, you score can vary dramatically depending on what time in the month
your credit report was ordered.
• Because your credit score depends on what type of credit you have, you can get a low
score even if you have a perfect payment record. If you have a credit card with a tire
company, a loan from a consumer finance company like Household or Beneficial, or
have an installment sales contract from a used car dealer, you get a lower score
regardless of whether you pay on time. But if you have a gas station credit card, you
score is higher!
• Because your credit score depends on the presence of loan information, you get a
lower score if you pay in cash or don’t borrow much or if you use lenders that don’t
9
2001 Consumer Bankruptcy Project, cited on page 81 of The Two Income Trap, Elizabeth Warren and Amelia
Tyagi.
16
report to credit bureaus. Many younger consumers were penalized with higher rates
due to so-called “thin” credit files because the Sallie Mae – the student loan lender to
millions – decided it would only report payment history to one of the three major
credit bureaus.
• Because your credit score depends on the ratio of your debt to your credit card limit, a
consumer who uses one credit card to maximize frequent flier miles gets a lower
score than another consumer who charges the same amount but does it on three or
four cards.
Insurance Scoring Penalizes Victims or Economic or Medical Catastrophes
Insurance credit scoring is inherently unfair because it penalizes consumers who are the victims
of economic or medical catastrophes, such as job loss, divorce, dread disease or terrorist attack.
For example, in the aftermath of the September 11 attack, hundreds of thousands of people
working in the travel-related industry lost their jobs. Out of this group, thousands had to increase
borrowing to offset loss of income or loss of health insurance. Many filed for bankruptcy. In the
aftermath of Hurricane Katrina, hundreds of thousands of consumers were displaced and placed
in financial stress. It is unfair for insurance companies to further penalize these victims by
raising their homeowners and auto insurance rates.
One of the myths perpetrated by insurers to rationalize the use of insurance credit scoring to
legislators is the myth of the immoral debtor. Insurers argue that good credit scores reflect the
financial responsibility of consumers. And they ask why should financially responsible
consumers subsidize the rates of consumers who are not financially responsible? As explained
further below, this argument fails because a good credit history does not equate to a good credit
score. Stated differently, an insurance score is simply not a measure of financial responsibility.
Regarding the “immoral debtor,” data on the causes of bankruptcies reveal that the
overwhelming majority of bankruptcies result from job loss, medical problems and divorce.
Fully 87% of bankruptcies for families with children arise from these three reasons. And the
remaining 13% includes reasons such as natural disaster or crime victim.
10
In their recent book, The Two Income Trap, Elizabeth Warren and Amelia Tyagi study the
growth, composition and causes of bankruptcy. They were astonished to find that the number of
women filing for bankruptcy grew from 69,000 in 1981 to nearly 500,000 by 1999. As they
researched the causes of this phenomenon, they documented the fact that financial strain on
families – particularly families with children – resulted from dramatic increases in the cost of
housing, health care and schooling combined with deregulation of interest rates for loans and
business decisions made by lenders for easy credit. They found that married couples with
children are more than twice as likely to file for divorce than couples without children and that a
divorced woman raising a child is nearly three times more likely to file for divorce than a single
woman without a child. They concluded that “having a child is the single best predictor that a
10
2001 Consumer Bankruptcy Project, cited on page 81 of The Two Income Trap, Elizabeth
Warren and Amelia Tyagi.
17
woman will end up in financial collapse.” Their research shows that the insurer rationalization
for insurance credit scoring – “financial responsibility” – is indeed a myth refuted by the facts.
A Good Credit History Does NOT Equal a Good Credit Score
Insurance credit scoring is inherently unfair because a good credit history does not equal a good
credit score or favorable insurance treatment. This occurs because insurance credit scores are
based not just on bankruptcies and delinquencies, but also on other factors unrelated to credit
management. For example, credit scores are often based on the type of credit (consumer finance
loans are less favorable than bank loans), the number of credit cards (there is a magic number
that is optimal, even if the consumer only uses the retail store cards once to get the first time 10%
purchase discount), length of time credit has been established (which is another way of charging
younger people more), length of time since last account opened (which penalizes families that
have just moved or refinanced their mortgage) and the number of inquiries (which penalizes
consumers who shop around for the best rate – behavior that should be rewarded and not
punished with higher insurance rates.) While the insurance industry offers a rationale for each of
these factors, the fact is that insurance credit scoring casts too wide a net and penalizes people
engaged in behavior we would all consider good financial management.
18
Insurance Credit Scoring Produces Arbitrary Results
Insurance credit scoring is unfairly discriminatory and violates actuarial standards for risk
classification because it is an arbitrary process. For example, your score can vary from very bad
(“high risk”) to very good (“low risk”) depending on which credit reporting agency provides the
credit information to the insurer because a consumer’s information varies among the big three
bureaus. A representative from ChoicePoint admitted this in a hearing before the Georgia
Insurance Commissioner in 2001. The author recently ordered my three-bureau credit report and
found different inquiries in each of the three bureaus – not one single inquiry was reported by
more than one bureau.
Insurance credit scoring is arbitrary because a score can change dramatically over a short time
frame for no apparent reason. The author’ auto credit score in November 2002 (obtained from
www.choicetrust.com
) was very low – around the 17
th
percentile. In May 2003, the author’s
score was in the 82
nd
percentile. In six months (or perhaps a shorter period), the author’s score
went from very high risk to very low risk. No other insurance risk factor is so arbitrary.
Consumers Penalized for Lenders’ Business Decisions
Over the course of the 1990’s consumer debt grew dramatically as lenders made credit more
easily available to many consumers. The number of credit card solicitations grew from 1 billion
to 5 billion annually. Lenders moved to low- or no-down payment mortgages. Although lenders
are certainly free to make business decisions about loaning money, consumers should not be
penalized with higher homeowners or auto insurance premiums because of those decisions.
To illustrate the problem, Fannie Mae recently began requiring a 10% down payment for 30 year
mortgages on manufactured homes. Previously, consumers could get a loan with no money
down. In defending the proposal, Deborah Tretler, vice president of single family homes for
Fannie Mae, stated, "We don't serve borrowers well when it is easy for a borrower to get into a
home under very flexible terms, only to have them lose their home, their credit ruined and their
homeownership dreams turned into a nightmare.”
11
Warren and Tyagi, in The Two-Income Trap, explain how lenders make lots of money off of
problem borrowers through higher interest rates and substantial penalty fees.
It is not only lenders’ lending decisions that make insurance scoring unfair, it is also lenders’
reporting
decisions to credit bureaus. In some cases, lenders report only partial information
about loans to credit bureaus. For example, some major credit card vendors do not report card
limits, to prevent competitors from learning about their customers. But by failing to report credit
limits, the insurance credit scoring models often use the current balance as the limit – with the
result that the consumer appears to be maxing out his or her credit line. Which, in turn, lowers
the insurance score.
11
“Mortgage regulations could stop some would-be homeowners,” by Genaro C. Armas of the
Associated Press in the September 12, 2003 issue of the Austin American-Statesman.
19
In another example, Sallie Mae, the nation’s largest lender for student loans with millions and
millions of borrowers, has decided to report loan information to only one of the three major
credit bureaus – again, to protect its customer list. If a consumer who has a good student loan
payment history seeks auto insurance and the insurer happens to use a credit bureau that Sallie
Mae has not
reported to, the consumer gets a lower score than he or she should because a lack of
information penalizes a consumer in an insurance score.
In yet another example, journalist Ken Harney explains how some lenders refuse to report the
credit limits on credit cards and other loans to credit bureaus. Absent this information, the credit
bureaus report the current debt balance as the credit limit. This harms consumers because a
factor in credit scores is the ratio of current debt to credit limits. Harney cites a consumer who
was charged a much higher rate than she would have been had the lenders reported her credit
limits:
That extra expense would not have been caused by anything she did wrong, but rather by
what the card company did without her knowledge: keep her good credit behavior a
secret from potential competitors by withholding her credit limit and highest balance,
thereby decreasing her credit score. Credit card companies sometimes try to hide their
best customers' identities from other lenders trolling the credit bureaus' vast databases to
prescreen targets for card offers. Typically the trollers ask the bureaus for lists of
cardholders with higher scores, and avoid those with marginal or lower scores.
12
These examples of how lenders’ business decisions can dramatically affect an insurance
consumer’s insurance score further illustrate the arbitrary and unfair nature of insurance credit
scoring.
Most recently, the explosion in subprime lending included thousands of instances of
inappropriate loans to consumers – loans the consumer would clearly be unable to afford even if
housing prices continued to grow and interest rates remained low. There were instances of
abusive sales practices. Again, the question arises, why should these consumers suffer higher
auto and homeowners insurance rates because of the business decisions and practices of lenders?
12
Ken Harney, “2 Missing Numbers Can Doom a Loan,” Washington Post, 1/1/05, page F1.
See also Kenneth Harney, “Credit Card Limits Often Unreported,” Washington Post, 12/25/05,
page F1.
20
Insurance Credit Scoring Penalizes Consumers in Poor and Minority Communities
In addition to being arbitrary, insurance credit scoring also has a systematic bias against
consumers in poor and minority communities, described further below. It is important to state
clearly that the claim that insurance credit scoring has a disproportionate impact on
consumers in poor and minority communities is NOT an argument that poor people are poor
financial managers. The two arguments are unrelated because good financial management /
good credit history does NOT equate to a good insurance credit score. It is the structure of
insurance credit scoring models – and not the financial management habits of low-income
consumers – that creates the bias against consumers in poor and minority communities.
Further, it is unclear how anyone who has actually examined the factors and structure of
insurance credit scoring models could legitimately assert that the claim of systematic bias against
consumers in poor and minority communities is a critique of the financial management habits of
low-income consumers.
Insurance Credit Scoring: 21
st
Century Redlining and the End of Insurance
There are two main reasons CEJ works on insurance issues, particularly as they impact low
income and minority consumers. First, insurance is the mechanism that consumers and
businesses use to protect their assets in the aftermath of a catastrophic event – whether that’s a
fire, an auto accident, a natural disaster, theft. Insurance enables consumers and businesses to
preserve and to build assets, wealth and financial security. Insurance is essential for individual
and community economic development. And low income consumers should have the same
access to these essential financial tools as more affluent consumers. The history of insurance
redlining, however, is a story of less access, inferior products and higher prices for low income
and minority consumers.
Second, insurance is the primary mechanism for loss prevention – insurance provides economic
incentives for less risky behavior and economic disincentives for more risky behavior. Or at
least, that is what insurance pricing should do. Insurance pricing should be based on factors that
are under the control of the consumer and which make a difference in the likelihood of an auto
accident or homeowners’ claim. Insurance is the primary tool to encourage behavioral changes
that actually reduce accidents, human suffering and property damage.
Insurance credit scoring undermines these public policy goals in at least two ways. First, even if
insurance credit scoring did what it’s purported to do – charge higher rates for consumers with a
poor credit history – it is inherently unfair and undermines the basic purpose of insurance which
is to protect consumers’ assets in catastrophic times. Consider that 87% of families who file for
bankruptcy do so because of one of three reasons – job loss, divorce, catastrophic illness. So
even if insurance credit scoring is working as its proponents claim, the practice penalizes those
consumers who are victims of an economic catastrophe with, at best, higher rates, and at worst,
the elimination of coverage in the time of greatest need.
Second, the use of insurance credit scoring undermines the other core purpose of insurance by
giving more and more weight in the rating process to factors outside of the consumer’s control
and which provide no economic incentive for loss prevention. Insurance credit scoring
undermines the loss prevention capacity of insurance because it is unrelated to behavioral
21
changes that reduce the likelihood of an accident or damage from an event. When you know that
insurance rates will go up by 25% if you get a speeding ticket or an at-fault accident, that
knowledge affects your behavior. When you get a discount for putting on hail-resistant shingles
on your home or installing an anti-theft device in your vehicle, the consumer is in a position to
take positive action to not only affect the likelihood of an accident or claim, but also in a position
to lower his or her premium. And these types of discounts provide a benefit to some consumers
without raising the rates for other consumers – you can give someone a 40% discount for a hail
resistant roof and pay for that discount with lower expected losses – so a discount for one does
not mean a rate increase for another. With insurance credit scoring, it’s less than a zero sum
game – since is there no reduction in losses, any discounts for some consumers must be paid for
by rate increases for other consumers and insurance credit scoring adds costs to the system.
4. The Impact of Insurance Credit Scoring on Poor and Minority Consumers
Despite insurers’ claims to the contrary, it is clear that insurer underwriting and rating practices
now emphasize a consumer’s economic status rather than their driving record.
4.1 Prior Bodily Injury Limits
For example, several insurers now charge higher rates to consumers because of their prior
liability limits. If your previous policy was a basic limits policy, you will be charged more than if
your previous policy was, say, 50,000/100,000 limits. The use of prior liability limits by insurers
to determine assignment to a rating tier clearly penalizes low income consumers because of their
income. Given that insurers are completely willing to use underwriting and rating factors that
penalize consumers because of economic status, it should be no surprise that insurance credit
scoring has a disproportionate impact on consumers in low-income and minority communities.
4.2 Insurance Credit Scoring Penalizes Consumers in Low-Income and Minority
Communities
Despite insurer protests, there is no ample evidence that insurance credit scoring penalizes
consumers in low-income and minority communities.
22
4.2.1 Fair Isaac Admission
On the issue of insurance credit scoring versus income and race, the Executive Vice President of
Fair, Isaac and Company, Peter McCorkell, admitted that insurance credit scoring has a disparate
impact based upon race and income:
Doesn’t scoring result in higher reject rates for certain minorities than for whites?
Again, the short answer is, “Yes,” but it is the wrong question. The question ought to be:
“Does credit scoring produce an accurate assessment of credit risk regardless of race,
national origin, etc.?” Studies conducted by Fair, Isaac, and Company, Inc. (discussed in
more detail below) strongly suggest that scoring is both fair and effective in assessing the
credit risk of lower-income and/or minority applicants. Unfortunately, income, property,
education, and employment are not equally distributed by race/national origin in the
United States. Since all of these factors influence a borrower’s ability to meet financial
obligations, it is unreasonable to expect an objective assessment of credit risk to result in
equal acceptance and rejection rates across socioeconomic or race/national origin lines.
By definition, low-income borrowers are economically disadvantaged, so one would not
expect their score distributions to mirror those of higher-income borrowers.
13
4.2.2 Freddie Mac Study
In its 1999 National Consumer Credit Survey, Freddie Mac found:
Having a poor credit record is a relatively common problem in today’s
society. Using the combined results from the CCS (i.e., African-
Americans, Hispanics and Whites) we estimate that:
30% of these groups have "bad" credit records
13% of these groups have "indeterminate" credit records
57% of these groups have "good" credit records
Credit problems persist across income groups. We estimate that:
36 % of consumers with incomes under $25,000 had "bad" credit records
33 % of consumers with incomes of $25,000 to $44,999 had "bad" credit records
25 % of consumers with incomes of $45,000 to $64,999 had "bad" credit records
22 % of consumers with incomes of $65,000 and $75,000 had "bad" credit
records
Minority borrowers are more likely than white borrowers to experience
credit problems. For African-Americans we estimate that:
48% of African Americans have "bad" credit records
13
Page 15, Fall 2000 Issue of Profitwise, a publication of the Federal Reserve Bank of Chicago.
23
16% of African Americans have "indeterminate" credit records
36% of African Americans have "good" credit records
For Hispanics we estimate that:
34% of Hispanics have "bad" credit records
15% of Hispanics have "indeterminate" credit records
51% of Hispanics have "good" credit records
For Whites, in contrast, we estimate that:
27% of Whites have "bad" credit records
12% of Whites have "indeterminate" credit records
61% of Whites have "good" credit records
It is unclear how the quality of credit histories can vary by income and race, but the
insurance industry still maintains insurance credit scoring has no disparate impact based
upon income and race.
4.2.3 Data from the Survey of Consumer Finances
Statistics the Survey of Consumer Finances, reported in the 2000 Statistical Abstract of the
United States reveal that credit characteristics vary not only by age and income, but also over
time within age and income segments. Table 792 – Financial Assets Held by Families by Type
of Asset: 1992 to 1998 shows the ownership of any financial assets varies dramatically by age
and income. The ownership of financial assets is related to the ability of a family to withstand an
economic or medical catastrophe.
Table 796 – Ratios of Debt Payments to Family Incomes: 1992 to 1998 shows higher ratios of
debt payments to family income and much higher ratios of families with payments 60 or more
days due for younger and lower income families. The table also shows how these ratios – both
of which figure prominently in insurance credit scores – vary over time.
Table 817 – Usage of General Purpose Credit Cards by Families: 1992 to 1998 shows that
younger and poorer families are much less likely to pay off credit card balances each month and
far more likely to hardly ever pay off the balance than older or more affluent families. Again,
these characteristics – which vary by age and income – figure prominently in insurance credit
scores.
4.2.4 The University of Texas Study
Further evidence of the disproportionate impact of insurance credit scoring on poor and minority
consumers comes from the report prepared by the University of Texas Bureau of Business
Research on the relationship between insurance credit scoring and insurance losses. The authors’
analysis of the correlation between insurance credit scoring and insurance losses is unreliable – it
relies upon a simple loss ratio methodology that the NAIC insurance credit scoring working
24
group rejected in 1996 as “misleading and counterproductive.” However, the report does reveal
other important findings.
The authors found that average and median credit scores were much higher in the standard
market than in the nonstandard (so-called “high risk”) market. But the scores were taken from
policies issued in 1998 – before the insurers were using credit history to underwrite consumers in
the standard and nonstandard markets. Consequently, if credit history was unrelated to
underwriting risk factors used by insurers, we would expect average scores to be similar in the
standard and nonstandard markets. The fact that the scores were so different between the two
markets means that insurers were already using some underwriting factor or factors to
distinguish risk of consumers that is correlated to credit.
In addition to showing that credit scores are a proxy for other risk factors used by insurers, the
difference in credit scores between the standard and nonstandard markets also indicates that
credit scores are correlated to race and income of consumers. Just as low credit scores are more
prevalent in the nonstandard market, the likelihood of being denied coverage in the standard
market and ending up in a high-cost county mutual grows dramatically as the neighborhood
becomes less affluent and less white.
25
Standard Auto Insurance Market Rejection Rates in Texas versus Race and Income
1996 1996
Average of Average of
Automobile Non-Anglo Median 1996
Rejection Population Household Number of
Rate Percentage Income ZIP Codes
0.0% to 5.2% 4.7% $22,414 1
5.3% to 10.4% 12.1% $44,042 74
10.5% to 15.6% 13.6% $30,565 317
15.7% to 20.8% 20.7% $24,871 413
20.9% to 26.0% 29.4% $24,523 280
26.1% to 31.1% 43.0% $23,456 142
31.2% to 36.3% 54.6% $21,549 79
36.4% to 41.5% 68.5% $19,954 65
41.6% to 46.7% 82.7% $17,682 45
46.8% to 51.9% 83.7% $16,441 38
Over 51.9% 92.3% $14,015 26
4.2.5 Factors Used in Insurance Credit Scoring Models are Biased Against Consumers in Low-
Income and Minority Communities
A review of the factors contained in insurance scoring models – and the information missing
from consumer credit reports and scoring models – further documents the disproportionate
impact of insurance credit scoring against poor and minority consumers.
Reason codes for insurance models from ChoicePoint include factors that systematically
discriminate against consumers in poor and minority communities. In the ChoicePoint models, a
consumer's score is affected by the type of credit and/or the type of lender -- regardless of
whether the consumer is current on the payments. A consumer who gets a loan from a consumer
finance company gets a lower score than a consumer who gets a loan from a bank – even if the
consumer has a perfect payment record. A consumer who has a credit card from a tire store --
such as Goodyear -- gets a lower score just for having that account. A consumer who buys a car
through an installment sales contract gets a lower score -- even if the payment record is perfect.
Clearly, consumers in less affluent neighborhoods are far more likely to use these types of credit
mechanisms than consumers in more affluent communities.
The fact is that the financial institutions in poor and minority communities are different from
those in more affluent white communities. And this difference results in a systematic bias in
insurance credit scoring models. As a further example, consider payday lenders, check cashing
lenders and rent-to-own businesses – which target poor consumers. Even if a consumer was able
to pay the extraordinarily high interest rates from these businesses, it would not help the
consumer’s insurance score – because these institutions do not report to credit bureaus. And the
26
absence of information in a credit report is a credit score negative. Consequently, consumers
who pay in cash or who use financial institutions that do not report to a credit reporting agency
are penalized with lower scores. Finally, consider a consumer who demonstrates financial
responsibility by paying all her utility bills on time for decades. This actual financial
responsibility is not rewarded in insurance credit scoring models because these payments do not
appear in credit reports.
4.2.6 The Missouri Department of Insurance Study
A few weeks ago, the Missouri Department of Insurance released a study that specifically
examined the impact of insurance credit scoring on the availability of insurance coverage in poor
and minority communities. This is the first independent study based on detailed insurance credit
scoring data using rigorous statistical analysis. The Department collected credit score data
aggregated at the ZIP Code level from 12 insurers for the study period of 1999 to 2001. For each
Missouri ZIP Code, the Department obtained:
• Mean credit score
• The number of exposures for each of five equal credit score intervals
The Department then utilized a variety of multi-variate statistical techniques to isolate the
relationship of income and race to insurance credit scoring, independent of other factors. The
study found:
• The insurance credit-scoring system produces significantly worse scores for residents of
high-minority ZIP Codes. The average credit score rank in “all minority” areas stood at 18.4
(of a possible 100) compared to 57.3 in “no minority” neighborhoods – a gap of 38.9 points.
This study also examined the percentage of minority and white policyholders in the lower
three quintiles of credit score ranges; minorities were overrepresented in this worst credit
score group by 26.2 percentage points.
• The insurance credit-scoring systems produces [sic] significantly worse scores for residents
of low-income ZIP Code. The gap in average credit scores between communities with
$10,953 and $25,924 in per capita income (representing the poorest and wealthiest 5 percent
of communities) was 12.8 percentiles. Policyholders in low-income communities were
overrepresented in the worst credit score group by 7.4 percentage points compared to higher
income neighborhoods.
• The relationship between minority concentration in a ZIP Code and credit scores
remained after eliminating a broad array of socioeconomic variables, such as income,
educational attainment, marital status and unemployment rates, as possible causes. Indeed,
minority concentration proved to be the single most reliable predictor of credit scores.
• Minority and low-income individuals were significantly more likely to have worse credit
scores than wealthier individuals and non-minorities. The average gap between minorities
and non-minorities with poor scores was 28.9 percentage points. The gap between
27
individuals whose family income was below the statewide median versus those with family
incomes above the median was 29.2 percentage points.
Based upon the results of this study, the former Governor of Missouri has called for a ban on
insurance credit scoring.
4.2.7 The Texas Department of Insurance Preliminary Report
The Texas Department of Insurance (TDI) reviewed over 2 million policyholder records and
obtained policyholder-specific information on race. The TDI report, issued in the beginning of
January 2005, states unequivocally that insurance credit scoring discriminates against minority
consumers:
The individual policyholder data shows a consistent pattern of differences in credit
scores among the different racial/ethnic groups. The average credit scores for Whites and
Asians are better than those for Blacks and Hispanics. In addition, Blacks and Hispanics
tend to be over-represented in the worse credit score categories and under-represented in
the better credit score categories.
14
The TDI study confirms and validates the Missouri Department of Insurance (MDI) study.
Insurers complained about the Missouri study because it inferred socio-economic characteristics
from ZIP Codes to average credit scores. But the MDI methodology is well accepted in the field
of fair lending analysis. The TDI study not only confirms the MDI study results – it validates the
MDI methodology.
4.2.8 Traditional Credit Reports Penalize Low Income and Minority Consumers
CEJ and other consumer groups have long argued that traditional credit reports penalize low
income and minority consumers because the absence of credit information – so-called “thin
files” – results in higher premiums. In the past year, the credit report and credit scoring industry
has admitted this bias against consumers. Several vendors are now developing “non-traditional”
credit reports, which include information not contained in traditional credit reports, such as rent
and utility payments and activity related to non-traditional loans. Fair, Isaac, the original
developer of lending and insurance credit scoring models claims that 50 million Americans are
unscorable using traditional credit information because of thin files.
15
First American, a provider
of credit information, claims its non-traditional credit reports will benefit minority and low-
income families
16
, indicating that traditional credit reports harm these consumers. Insurers have
always used traditional credit reports and penalized consumers with thin files and such practices
have resulted in disproportionately higher premiums for low-income and minority consumers as
well as some seniors.
14
Texas Department of Insurance, “Report to the 79
th
Legislature: Use of Credit Information in
Texas,” December 30, 2004, page 3.
15
“Giving Credit Where Credit’s Due,” Kenneth Harney, Washington Post, November 11,
2006, Page F1
16
http://www.credco.com/Anthem/default.htm
28
4.3 Conclusion
In conclusion, the problems with insurance credit scoring are apparent and even acknowledged
by the industry, as evidenced by their “compromise” proposal (the NCOIL model) with a variety
of purported restrictions and regulatory oversight. But what are the great benefits to consumers
that warrant the use of this problematic factor and intense regulatory resources? Ultimately,
there are none. Moreover, all the benefits alleged by the insurance industry come down to one
claim – the purported statistical relationship between credit scores and loss ratios. And while a
definitive statistical relationship is a necessary justification for the use of certain information as
an underwriting or rating factor, such a statistical relationship can not be sufficient justification.
If it were, then race would be a legitimate rating factor. But lawmakers across the country have
decided that race is not a legitimate basis for underwriting for rating insurance. If race can not
be used directly by insurers, then insurers should not be permitted to use race indirectly through
insurance credit scoring.
5. False Industry Claims About Insurance Scoring
The insurance industry, at one time or another, has claimed insurance scoring is the cause of
untold benefits for consumers and has denied any problems or consumer harm resulting from
insurance credit scoring. Simply stated, the insurance industry has no credibility when it comes
to insurance credit scoring. For example, in 1999, at the same time the industry was denying
state insurance regulators the data necessary to evaluate the impact of insurance scoring on low
income and minority consumers, the American Insurance Association issued a report claiming a
study by one of its member companies (Hartford) had shown “that credit score is not
significantly related with income. . .”
17
The insurance industry also claimed no relationship
between insurance score and race.
18
Once insurance regulators obtained the data necessary to
perform an independent study, the industry claims were proven false. The Texas and Missouri
Departments of Insurance both found that insurance scoring has a disproportionately negative
impact on low income and minority consumers, as discussed above.
The insurance industry continues to make false claims about the benefits of insurance scoring.
Just this week, the industry media organization, the Insurance Information Institute, claimed
insurance scoring was responsible for auto insurance rate reductions. As shown below, this
claim is incorrect. In fact, insurance scoring has been responsible for excessive auto insurance
rates.
Industry Claim 1: Insurance Scoring Is an Accurate Predictor of Claims, Promotes
Competition and the Availability of Affordability of Insurance
17
Statement of the American Insurance Association on the Lack of Correlation Between
Income and Credit Score, March 1999, page 1
18
See testimony of Progressive Insurance before the Florida Task Force on the Use of Credit
Reports in Underwriting Automobile and Homeowners Insurance, 2001-02.
29
Insurance scores can help make insurance more affordable.
Insurers have found that using insurance scores as a factor in the underwriting process
helps them to more accurately price policies and actually write more policies. In some
cases, consumers pay less for insurance. This information helps insurance companies
determine a fair premium for each consumer that is related to their potential for filing a
claim.
Insurance scoring can help increase the availability of insurance.
Many consumers, who might otherwise have less access to or have been denied coverage
for a variety of reasons, are able to find coverage because insurance companies use credit
history to underwrite policies.
Insurance scoring promotes competition.
Facts:
Insurance scoring decreases insurance availability by raising rates for those consumers for whom
price increases make a difference in the ability to purchase insurance – low income consumers.
Objective measures indicate that insurance scoring has decreased competition and worsened
insurance availability and affordability.
Insurers claim that insurance credit scoring allows more accurate pricing. If this were the case,
we would expect some consumers to pay more and some to pay less while the ratio of claims
paid to premiums collected to remain constant. In fact, insurance scoring has led to lower loss
ratios and higher profits for insurers. In addition, measures of uninsured motorists by the
industry’s own research organization indicate more uninsured motorists – direct refutation of the
claim that insurance credit scoring promotes greater insurance availability and affordability
Excessive Rates and Profitability:
Private Passenger Automobile Loss Ratios, Countrywide
2000 71.2%
2001 72.7%
2002 67.5%
2003 62.8%
2004 58.6%
2005 60.1%
2006 57.9%
The report Credit Scoring And Insurance: Costing Consumers Billions And Perpetuating The
Economic Racial Divide analyzes auto insurer profitability over the period in which insurers
started using insurance scoring more intensively. The report found over $55 billion in excessive
auto insurance premiums for the three years 2004 through 2006.
30
As the profitability data show, any recent reduction in auto insurance rates has not been caused
by insurance scoring. In fact, auto insurance rates are too high and the absence of competition to
drive rates to reasonable levels is attributable to insurance scoring. Consider the comments of Ed
Liddy, then-CEO of Allstate to investment analysts in 2005:
Tiered pricing helps us attract higher lifetime value customers who buy more products
and stay with us for a longer period of time. That’s Nirvana for an insurance company.
That drives growth on both the top and bottom line.
This year, we’ve expanded from 7 basic price levels to 384 potential price levels in our
auto business.
Tiered pricing has several very good, very positive effects on our business. It enables us
to attract really high quality customers to our book of business.
Make no mistake about it, the economics of insurance are driven largely by retention
levels. It is a huge advantage. And our retentions are as high as they have ever been.
The key, of course, is if 23% or 20% of the American public shops, some will shop every
six months in order to save a buck on a six-month auto policy. That’s not exactly the
kind of customer that we want. So, the key is to use our drawing mechanisms and our
tiered pricing to find out of that 20% or 23%, to find those that are unhappy with their
current carrier, are likely to stay with us longer, likely to buy multiple products and that’s
where tiered pricing and a good advertising campaign comes in.
It (tiered pricing) has raised the profitability of the industry.
19
As made clear by Ed Liddy’s comments, insurance scoring is used to predict consumer
profitability, which is not the same as predicting risk of loss.
Uninsured Motorists
According to a recent Insurance Research Council (IRC) study, the estimated percentage of
uninsured motorists increased nationally from 12.7 percent in 1999 to 14.6 percent in
2004. (Uninsured Motorists, 2006 Edition) These data directly refute industry claims that
insurance scoring promotes insurance availability and affordability.
Residual Market
According to data from the Auto Insurance Plan Service Office, an organization that operates or
assists in the operation of assigned risk plans across the country, the number of vehicles insured
through assigned risk plans grew by about 70% from 217,200 in 2000 to 368, 831 in 2003 not
including the New York assigned risk plan and 100% from 433,242 to 864,074 including New
19
Partial Transcript of Presentation to Edward M. Liddy, Chairman and CEO, The Allstate Corporation
Twenty-First Annual Strategic Decisions Conference, Sanford C. Bernstein & Co., June 2, 2005.
31
York.
20
These data directly refute industry claims that insurance scoring promotes insurance
availability and affordability.
No Evidence of Consumer Harm in States Where Insurance Scoring is Banned
In addition, there is no evidence that insurers have restricted their writings in states that ban
insurance credit scoring. In California, insurance credit scoring is not permitted for private
passenger automobile insurance, yet there are many insurers offering insurance and, in 2003, the
percentage of vehicles insured through the involuntary market (assigned risk plan) was 0.3% or
3 out of every 1,000 vehicles insured. In contrast, in 2003 in New York, where insurers use
insurance credit scoring, the assigned risk share of the market is 5.5% or 18 times higher than in
California
Insurance Credit Scoring is Part of a Trend to Rating Based on Economic Status
The insurance industry has long targeted low income and minority communities with high-cost
auto and home insurance products, in the same manner that predatory lenders targeted low-
income and minority communities with subprime and predatory loans. A recent risk
classification filing in Texas provides a tier matrix based on the following factors, showing that
economic status has greater weight in determining a consumer’s premium that driving record or
miles driven.:
• Prior insurer
• Prior liability limits
• Previous non-standard insurance
• Lapse status
• College education
• Occupation
• Age of vehicle
• Multi-car policy
• Years with current employer
• Home ownership
• Not-at-fault accidents
• Credit score
Some Evidence Refutes the Alleged Relationship Between Credit and Claims
Insurers argue that there is a powerful correlation between insurance scores and expected claims.
If such a relationship actually existed, then we would expect that an increase in delinquencies
and bankruptcies would be matched by an increase in insurance claims. In fact, the opposite has
occurred. Despite rapid increases in bankruptcies and delinquencies since 2000, auto claims
have remained stable or declined. This suggests that the correlation between insurance credit
scores and claims is not real and that insurance scores are a proxy for some other factor that is
truly related to claims.
20
Auto Insurance Report, “Residual Market Growth Continues Despite Strong Voluntary Profit,”
August 29, 2005. Note, the cited AIPSO data covers 46 states.
32
Industry Claim 2: Most Consumers Benefit
Most people benefit from insurance scoring.
Most people have good credit and can benefit from insurance scoring. It can help
consumers qualify for lower insurance rates and in some cases, even offset a less than
perfect driving record.
Most consumers pay less because of insurance scoring.
An NAII member company found that insurance scoring helps it offer lower premiums to
nearly 70 percent of its policyholders. Insurance scores enable insurers to price products
with greater accuracy, and with every customer paying according to his or her potential
for loss.
Facts:
Insurance Credit Scoring Hurts All Consumers
There are two basic public policy purposes of insurance. The first is to provide individuals,
businesses and communities with a financial security tool to avoid financial ruin in the event of a
catastrophic event, whether that event is a traffic accident, a fire or a hurricane. The is essential
financial security tool is accomplished by the spreading of risk over a large number of consumers
and business and is typically performed by insurers accepting the transfer of risk from
individuals and by spreading the individual risks through the pooling of very large numbers of
individual risks. The pool or risks is diversified over many types of perils and many geographic
locations.
The second essential purpose of insurance is to promote loss prevention. Insurance is the
fundamental tool for providing economic incentives for less risky behavior and economic
disincentives for more risky behavior. The insurance system is not just about paying claims; it is
about reducing the loss of life and property from preventable events. Historically, insurers were
at the forefront of loss prevention and loss mitigation. At one point, fire was a major cause of
loss – no more, in large part due to the actions of insurers in the 20
th
century.
Insurance credit scoring hurts all consumers by undermining the both goals of insurance. It hurts
the goal of providing an essential financial security tool by making insurance less affordable and
available to the consumers most in need of the tool. It undermines the loss prevention role of
insurance by removing the ability of insurance rating to provide economic incentives for less
risky behavior and economic disincentives for more risky behavior.
Good Credit Histories Don’t Equate to Good Credit Scores
Insurance credit scoring is inherently unfair because a good credit history does not equal a good
credit score or favorable insurance treatment. This occurs because insurance credit scores are
based not just on bankruptcies and delinquencies, but also on other factors unrelated to credit
management. For example, credit scores are often based on the type of credit (consumer finance
loans are less favorable than bank loans), the number of credit cards (there is a magic number
that is optimal, even if the consumer only uses the retail store cards once to get the first time 10%
purchase discount), length of time credit has been established (which is another way of charging
33
younger people more), length of time since last account opened (which penalizes families that
have just moved or refinanced their mortgage) and the number of inquiries (which penalizes
consumers who shop around for the best rate – behavior that should be rewarded and not
punished with higher insurance rates.) While the insurance industry offers a rationale for each of
these factors, the fact is that insurance credit scoring casts too wide a net and penalizes people
engaged in behavior we would all consider good financial management.
Over the course of the 1990’s consumer debt grew dramatically as lenders made credit more
easily available to many consumers. The number of credit card solicitations grew from 1 billion
to 5 billion annually. Lenders moved to low- or no-down payment mortgages. Although lenders
are certainly free to make business decisions about loaning money, consumers should not be
penalized with higher homeowners or auto insurance premiums because of those decisions.
To illustrate the problem, Fannie Mae recently began requiring a 10% down payment for 30 year
mortgages on manufactured homes. Previously, consumers could get a loan with no money
down. In defending the proposal, Deborah Tretler, vice president of single family homes for
Fannie Mae, stated, "We don't serve borrowers well when it is easy for a borrower to get into a
home under very flexible terms, only to have them lose their home, their credit ruined and their
homeownership dreams turned into a nightmare.”
21
It is not only lenders’ lending decisions that make insurance scoring unfair, it is also lenders’
reporting decisions to credit bureaus. In some cases, lenders report only partial information
about loans to credit bureaus. For example, some major credit card vendors do not report card
limits, to prevent competitors from learning about their customers. But by failing to report credit
limits, the insurance credit scoring models often use the current balance as the limit – with the
result that the consumer appears to be maxing out his or her credit line. Which, in turn, lowers
the insurance score.
In another example, Sallie Mae, the nation’s largest lender for student loans with millions and
millions of borrowers, has decided to report loan information to only one of the three major
credit bureaus – again, to protect its customer list. If a consumer who has a good student loan
payment history seeks auto insurance and the insurer happens to use a credit bureau that Sallie
Mae has not reported to, the consumer gets a lower score than he or she should because a lack of
information penalizes a consumer in an insurance score.
Every Consumer Organization and Most Agent Groups Want Insurance Credit Scoring
Banned
The National Association of State Farm Agents, Inc. (NASFA) hereby resolves that we
are opposed to any insurance company using credit scoring for the purpose of property
and casualty underwriting and rating. We believe credit scoring is part of a marketing
scheme designed to curtail market share, avoid rate regulation and it improperly
emphasizes credit as an underwriting characteristic without sufficient demonstration of its
21
“Mortgage regulations could stop some would-be homeowners,” by Genaro C. Armas of the
Associated Press in the September 12, 2003 issue of the Austin American-Statesman.
34
reliability for underwriting purposes. There is tremendous opportunity to mischaracterize
potential insurers and inadvertently or intentionally illegally discriminate. We further
support legislation to prohibit credit scoring for the purpose of property and casualty
underwriting and rating.
The National Association of Professional State Farm Agents and The United Farmers Agents
Association and other agents’ groups oppose insurers’ use of insurance credit scoring. Every
consumer organization opposes insurance credit scoring – Consumer Federation of American,
U.S. Public Interest Research Group, state PIRGs, Consumers Union, AARP and many more.
Consumers Union recently wrote:
22
Even though insurance companies cannot use race or ethnicity to decide who gets insurance
and how much it will cost, evidence shows that insurance scores disproportionately affect
certain minority groups and low-income consumers, which raises concern that scores can
serve as a proxy for race or ethnicity. Research shows that people in areas with a high
concentration of minorities are more likely to have lower credit scores.
The consequences are far-reaching. The economic stability of our cities and our nation
depends in part on access to fairly priced coverage. Insurance is based on the concept that
spreading the risk helps society protect itself from economic devastation and more quickly
recover from catastrophes. When insurance costs are inflated for the wrong reasons, people
are unfairly cut off from access to its protection. The whole community suffers, and those
who cannot afford insurance struggle to recover if disaster hits.
Another hurricane season is already upon us. Based on past years with similar
conditions, the National Oceanic & Atmospheric Administration estimates that two to
four hurricanes could affect the U.S. in 2006. But there's more trouble on the horizon than
just bad weather. In any state that allows insurers to use credit information to rate and
underwrite homeowners- and auto-insurance policies, consumers are already in the
middle of a storm, and most of them don't know it.
The devastation caused by Hurricanes Katrina, Rita, and Wilma shows us that people
without adequate insurance may face compounded tragedy. Since economic losses caused
by catastrophe can send a credit score plummeting, even consumers who can afford
insurance today may feel the repercussions of credit scoring in their premiums tomorrow.
Consumers Union advocates have been urging legislators and regulators in several states to
ban the practice, and we'll continue those efforts.
Polls Show the Public is Opposed to Insurance Credit Scoring
In a poll of Texas consumers conducted from April 28, 2003 through May 10, 2003, 68% voiced
the opinion that the Texas Legislature should “ban insurance companies from using a
homeowner’s credit history to decide whether it will insure a person or to adjust a premium,”
compared to 23% who voiced support.
22
Consumer Reports, August 2006, Page 61
35
Insurers Hide their Use of Insurance Credit Scoring
If insurers really believed that the public supports the use of insurance credit scoring, why don’t
we see any insurers’ ads or marketing efforts that promote their use of insurance credit scoring?
Why don’t we see any ads that even mention insurance credit scoring?
Most Consumers Don’t Get Lower Rates
Data from actual filings refute the industry claim. My analysis of actual rate filings shows that in
many cases, the so-called “discounts” consumers receive from insurance scoring are more than
offset by increases in the base rate. The fact is that, because insurance scoring does nothing to
reduce insurance claims, insurance scoring simply redistributes premiums among different
consumers. And in most cases, the number of consumers who see a premium reduction is the
same or less than the number who see a premium increase.
Industry Claim 3: Insurance Scoring is An Objective Tool
Insurance scoring provides an objective tool for decision-making.
This tool does not discriminate against any specific group of customers. It avoids
subjective value judgments because the information is based solely on credit-related
material.
It provides an objective tool for decision-making that does not discriminate against
specific groups or individuals.
Insurers are interested in having available as many tools as possible to assist them in
making a fair and objective decision about whom to insure and at what rate. The
development of an insurance score only takes into account credit-related information and
does not consider race, gender, religion, marital status and birthplace.
Insurance Scores are reliable.
The Consumer Data industry Association, formerly Association of Credit Bureaus,
reports that less than 1 percent of all credit report challenges result in a change once the
inquiry has been fully investigated. Studies have found that credit reports are more
reliable than motor vehicle records. The use of credit reports is routine throughout the
financial services industry and is widely accepted by consumers.
Insurance Scores are Not Correlated to Income
March 1999, Statement of the American Insurance Association, “On the Lack of
Correlation Between Income and Credit Score When Tested Against the Average or
Median Score”
The precise objective of the company analysis was to determine the extent to which the
credit score is correlated to income. AIA presented important, new evidence that credit
scores do unfairly discriminate against or even negatively impact lower income groups.
Indeed, research revealed that the lowest income groups have the highest average credit
score.
36
The analysis concluded that credit score is not significantly correlated with the income
for the AIA company’s policyholders.
Facts:
Selection of Factors in Insurance Scoring Models Involves Judgment and Bias
The mere fact that insurance scores are produced by a computer model does not mean insurance
scores are objective. If the factors that go into the scoring model discriminate against low
income and minority consumers, then the model itself will be biased against such consumers.
Ad discussed above, two independent studies confirm that insurance credit scoring is highly
correlated to income and race.
Insurance Scoring is Arbitrary
There are many example of illogical and arbitrary results from insurance scoring:
• Because your credit score depends on having the “right” kind of information in your
credit report, you can have a perfect credit history and still get a bad credit score.
Contrary to insurer credit scoring myths, your credit score has nothing to do with
your “financial responsibility.”
• Because your credit report can vary dramatically among the three major credit
bureaus, your credit score can vary from good to bad depending upon which bureau
provided your insurer with information.
• Because your credit score is based on many things other than how timely you pay
your loans, you score can vary dramatically depending on what time in the month
your credit report was ordered.
• Because your credit score depends on what type of credit you have, you can get a low
score even if you have a perfect payment record. If you have a credit card with a tire
company, a loan from a consumer finance company like Household or Beneficial, or
have an installment sales contract from a used car dealer, you get a lower score
regardless of whether you pay on time. But if you have a gas station credit card, you
score is higher!
37
• Because your credit score depends on the presence of loan information, you get a
lower score if you pay in cash or don’t borrow much or if you use lenders that don’t
report to credit bureaus. Many younger consumers were penalized with higher rates
due to so-called “thin” credit files because the Sallie Mae – the student loan lender to
millions – decided it would only report payment history to one of the three major
credit bureaus.
• Because your credit score depends on the ratio of your debt to your credit card limit, a
consumer who uses one credit card to maximize frequent flier miles gets a lower
score than another consumer who charges the same amount but does it on three or
four cards.
Industry Claim 5: One of Many Factors
It's just one of many factors.
Most companies that use insurance scoring treat it as just one of several factors in the
underwriting decision. Generally your insurance score alone is not likely to keep you
from getting insurance or cause you to pay more for it, although it can help you get
insurance.
Facts:
Insurance Credit Scoring Affects Your Rates – Why Else Would Insurers Use It?
This industry argument is truly a red herring. The fact that insurance scores are one of many
factors does not change the fact that a consumer’s insurance score affects his or her premium
and, typically, is the most important factor in determining that premium. If insurance credit
scoring were simply a minor factor and not likely to affect the insurer decision to offer insurance
or affect the insurer decision about the price of insurance, why would insurers fight so hard to
use it and put up with all the requirements of federal and state law regarding the use of consumer
credit reports and insurance scoring?
Industry Claim 6: Rewards Responsible Financial Behavior
Insurance scores reward responsible financial behavior, not just the length of credit
experience.
Insurance scoring is designed to examine credit management patterns and the process
used provides an objective evaluation of a consumer's credit history whether it is long or
short. When a consumer does not have enough history to generate a score, this
information often will not be considered as a positive or negative characteristic.
Fact:
38
This argument represents a reprehensible blaming-the-victim strategy by insurers. In fact, a
credit history is not a measure of financial responsibility and a good credit history does not
equate to a good credit score.
A Credit Score is Not a Measure of Financial Responsibility
• Limited Info in Credit Report
o No Utility Payment History
o No Rental Payment History
o No Savings Information
o No Insurance Purchase Information
• Credit Score Factors Unrelated to Payment History
o Type of Credit
o Length of Credit
o Inquiries
o Balance to Limits
o Thin Files
• After the Fact Rationale
Insurance Credit Scoring Penalizes Victims of Economic and Medical Catastrophes
Insurance credit scoring is inherently unfair because it penalizes consumers who are the victims
of economic or medical catastrophes, such as job loss, divorce, dread disease or terrorist attack.
For example, in the aftermath of the September 11 attack, hundreds of thousands of people
working in the travel-related industry lost their jobs. Out of this group, thousands had to increase
borrowing to offset loss of income or loss of health insurance. Many filed for bankruptcy. It is
unfair for insurance companies to further penalize these victims by raising their homeowners and
auto insurance rates.
One of the myths perpetrated by insurers to legitimize the use of insurance credit scoring to
legislators is the myth of the immoral debtor. Insurers argue that good credit scores reflect the
financial responsibility of consumers. And they ask why should financially responsible
consumers subsidize the rates of consumers who are not financially responsible? As explained
further below, this argument fails because a good credit history does not equate to a good credit
score. Stated differently, an insurance score is simply not a measure of financial responsibility.
Regarding the “immoral debtor,” data on the causes of bankruptcies reveal that the
overwhelming majority of bankruptcies result from job loss, medical problems and divorce.
Fully 87% of bankruptcies for families with children arise from these three reasons. And the
remaining 13% includes reasons such as natural disaster or crime victim.
23
In their recent book, The Two Income Trap, Elizabeth Warren and Amelia Tyagi study the
growth, composition and causes of bankruptcy. They were astonished to find that the number of
women filing for bankruptcy grew from 69,000 in 1981 to nearly 500,000 by 1999. As they
23
2001 Consumer Bankruptcy Project, cited on page 81 of The Two Income Trap, Elizabeth
Warren and Amelia Tyagi.
39
researched the causes of this phenomenon, they documented the fact that financial strain on
families – particularly families with children – resulted from dramatic increases in the cost of
housing, health care and schooling combined with deregulation of interest rates for loans and
business decisions made by lenders for easy credit. They found that married couples with
children are more than twice as likely to file for divorce than couples without children and that a
divorced woman raising a child is nearly three times more likely to file for divorce than a single
woman without a child. They concluded that “having a child is the single best predictor that a
woman will end up in financial collapse.” Their research shows that the insurer rationalization
for insurance credit scoring – “financial responsibility” – is indeed a myth refuted by the facts.
Industry Claim 7: Consumer Protections Exist
The NCOIL Law, as adopted in many states, provides necessary consumer protections.
The Fair Credit Reporting Act provides consumer protections.
Facts:
The NCOIL Model Provides Little or No Consumer Protections.
The NCOIL model law, adopted in many states, allows insurers to continue their insurance
scoring practices with few or no substantial consumer protections. I discuss this issue at length
in my testimony before the Colorado Legislature in 2004, available on the CEJ web site:
www.cej-online.org.
Insurers Seek to Avoid Telling Consumers About Insurers’ Use of Credit Scoring
Adverse Action Notices: Insurers have resisted providing adverse action notices to consumers
who suffered higher rates because of insurance credit scoring. Insurers claimed that a new
business customer – even a customer charged the highest rate because of her credit score – was
not entitled to an adverse action notice.
40
Insurers Oppose Laws That Allow Consumers to Freeze Their Credit Information Because
of Identity Theft
New York recently adopted a credit information security freeze law, described by its sponsor as
follows:
"This security freeze acts as a barricade against those who would commit fraud," Senator
Steve Saland (R-C, Poughkeepsie), co-sponsor of the legislation, said. "Identity thieves
have already preyed on thousands of New York consumers, stealing personal information
that leaves consumers severely at risk. This law enables consumers to avoid victimization
by empowering them to place security freezes on their consumer reports."
But the New York measure is the only credit freeze legislation passed in the nation this year that
does not exempt insurers. Nine other states have passed credit freeze legislation in 2006,
(Colorado, Florida, Illinois, Kentucky, Wisconsin, South Dakota, Utah, Kansas, and Vermont),
and all of them allow insurers to continue to access credit information for underwriting and other
legitimate business purposes, according to the Property Casualty Insurers Association of
America (PCI), which has asked Gov. Pataki to veto credit freeze legislation.
PCI says including insurers in the freeze provides no benefit to consumers while increasing costs
for the industry.
"While PCI supports the effort to prevent identity theft, the application of credit freeze
legislation should be tailored to address areas in which there is a prevalence of identity
theft," said Kristina Baldwin, regional manager and counsel for PCI. "The security
provisions in this legislation have no practical application or consumer benefit in the
context of insurance."
According to Baldwin, it is "highly unlikely" that illegally procured credit information
would be used to purchase insurance. She cites a Federal Trade Commission study in
January that found that 99.6 percent of identity theft complaints were related to areas
other than insurance.
"Consumers obtain little or no benefit from having a security freeze which applies to
insurers. The insurer and the consumer would experience increased burdens, costs and
inconveniences associated with this credit freeze legislation. It is important to bear in
mind that additional insurance company burdens and costs are ultimately borne by all
policyholders through higher premiums. In short, the burdens associated with applying
credit freeze provisions to insurers are not outweighed by the very limited consumer
benefits which would be achieved through applying credit freeze provisions to insurers,"
Baldwin added.
The arguments are, of course, a non-sequitor. If a consumer has been a victim of identify theft,
then an insurers’ use of that that consumer’s credit information can hard the consumer because
the credit report has been damaged. Why would a consumer want an insurer to use her credit
report when it has been damaged by identify theft? Why would an insurer want to use such a
41
report? And why would insurers oppose giving consumers a tool to protect themselves from use
of their credit information when they suspect they have been the victim of identify theft?
Insurers’ actual insurance credit scoring practices and policies are profoundly anti-consumer.
The security freeze position is the latest example of insurers placing their interests above those of
consumers.
The recent Supreme Court Decision about Adverse Actions Contradicts Congressional
Intent and Denies Consumers Essential Consumer Protections.
As with the security freeze issue, insurers have tried to keep consumers in the dark about
insurance scoring practices by denying consumers adverse action notices required under the Fair
Credit Reporting Act. Some insurers refused to provide any new business applicant with an
adverse action notice – even if the consumer suffered a high premium because of insurance
credit scoring. The recent Supreme Court decision in Safeco v Burr and GEICO v Edo did
determine that insurers did need to provide adverse action notice to new business consumers who
suffered an adverse action, but defied congressional intent and incorrectly defined what
constitutes an adverse action. Despite a clear and simple definition of insurance adverse action
endorsed by state insurance regulators and the Federal Trade Commission – a consumer suffers
an adverse action if she suffers less favorable treatment that she would have received if she had a
more favorable credit report – the Court argued that too many consumers would get adverse
action notices and endorse a standard based on a so-called “neutral” credit score. Since there is
no standard for “neutral” credit score, the Supreme Court decision allows insurers to effectively
define which consumers get adverse action notices.
- 1 -
811.776.9430
Why Low Credit Scores Predict More Auto
Liability (& UM) Claims: Two Theories
1
Patrick Butler, Ph.D.
†
April 30, 2009
Submitted as testimony to the hearing on credit score rating by the
National Association of Insurance Commissioners (NAIC)
Summary. To help inform NAIC deliberations in regard to the hearing’s third issue—
“how current economic conditions have affected policyholder premiums related to credit-
based insurance scores”—this testimony considers two explanations for the fact that auto
liability claims vary inversely with driver credit scores. Theory 1 attributes the
correlation to a direct connection between financial negligence and driving negligence,
but this testimony identifies difficulties for Theory 1 and offers an alternative
explanation. Theory 2 proposes that since people (at all income levels) with low credit
scores must economize, many do this by a reduction in car owning without a proportional
reduction in driving. Such economizing raises the average miles per car and consequently
the number of liability claims per 100 car-years. Both theories are also critiqued with
respect to explaining other predictors such as driver sex and accident record. At stake is
NAIC backing for an effective response to the conflict between mandatory insurance and
ability-to-pay. Theory 1 suggests a need for strong price regulation to cross-subsidize low
credit-score, presumably more negligent drivers while Theory 2 explains why effective
regulation of credit score rating—and of other measures of financial status such as
education and occupation levels—might be difficult and ineffective. Theory 2 instead
suggests encouragement by regulators of an informed, free market demand by
consumers—and an entrepreneurial response by insurers—for an odometer mile exposure
unit as an optional alternative to the car year exposure unit for private passenger cars.
* * * *
Mandatory liability insurance has long been demanded by the public and, despite
steadfast opposition by major insurers, has been increasingly adopted over time by state
legislatures. But concern that insurance also be affordable leads to attempts to control
some pricing variables. A recent example is legislative efforts to prohibit the use of credit
1
The analysis of this testimony was cited by the July 2007 Federal Trade Commission (FTC) “Report
to Congress on Credit-Based Automobile Insurance Scores.” Although the report presents a truncated
version of the economic logic (page 32, citing a 2006 academic paper of mine), it does not consider the
inevitability of the correlation of more claims with lower credit scores caused by the need to economize on
car insurance.
†
Insurance Project Director, National Organization for Women, 1100 H Street, NW 3
rd
Floor,
Washington, DC 20005, tel. 202.628.8669 x148, email [email protected], cell 512.695.5136.
- 2 -
811.776.9430
scores in pricing. In response, insurers commissioned a study by Miller and Smith (2003)
of a random sample comprising nearly 2.7 million car-year records from the files of
national insurers. The sample shows that the cars owned by drivers with the lowest credit
scores produced 2.5 times as many liability claims per 100 car-year exposure units as the
cars owned by the highest score drivers. But this also means that credit score pricing
charges more to those generally on tighter budgets, which contributes to pressure for
regulating prices.
To help resolve the conflict between affordability and free-market pricing, this essay
further examines why lower credit scores predict more liability claims. Two theories are
brought to bear on this question. The prevailing explanation, Theory 1, is that a lower
credit score predicts more driver negligence. The basis is that each liability claim requires
a negligent act by the insured car’s driver to cause the accident. Since the cars of low
credit score drivers produce more liability claims than other cars in their insurance class,
it is assumed that these drivers perform more negligent acts and therefore on average are
more negligent drivers. In a 2002 report on credit-score pricing to the NAIC, the
American Academy of Actuaries (AAA) likens the way credit scores work to the way
driver records work in identifying subgroups within insurance classes:
[H]istories of past accidents and violations do not cause drivers to have more
accidents. The rating practice that does exist is based on the fact that, as a group,
drivers who have been accident-prone in the past are likely to be accident-prone
in the future. [Emphasis original.]
But the AAA report is also arguing here that the cause of a correlation need not be
identified in order to gain approval for its use in pricing. Nevertheless, legislators,
insurance commissioners, and consumer advocates continue to call for an explanation for
the credit score correlation with claims.
As the first academic response to these calls, Brockett et al. (2005, 2007) provide
backing for Theory 1’s driver negligence explanation. They review studies about how the
“characteristics of individual risk taking . . . affect both financial decision making and
risky driving habits.” Brockett and Golden (2007) conclude that the research examined
by their article
suggests that the discussed individualized biological and psychobehavioral
correlates provide a connection between credit scores and automobile insurance
losses. Credit scores, like good student discounts and marital status, tap a
dimension of responsibility and stability for the individual that can permeate
multiple areas of behavior.
But this suggested connection entails unaddressed issues. One is that the studies
reviewed by Brockett and Golden rely on accident data based on the driver year, whereas
insurance claim data are based on the car-year exposure (statistical) unit and tied to the
driver-type classification of the car rather than to the driver driving at the time the car
- 3 -
811.776.9430
was involved in an accident. Moreover the review takes no notice that according to
periodic federal travel surveys (Hu and Reuscher, 2004) different categories of drivers and
cars represent a wide range in average annual miles and, furthermore, that within the
categories themselves drivers and household cars individually traveled from zero to
50,000 miles and more in the years surveyed. Differences in annual mile averages can
readily match reported ranges in liability claims per 100 car years from the lowest to
highest credit score categories. For instance, the 2.5 times difference in annual liability
claims reported by Miller and Smith can be matched by the 2.5 times difference in annual
miles from 6,000 miles to 15,000 miles. According to the 1995 travel survey, 30% of cars
were driven less than 6,000 miles and 25% of cars were driven more than 15,000 miles.
2
Characterizing those with low credit-scores as “high risk drivers” on the basis of
insurance records misleadingly implies that the high risk must be on the same statistical
per-mile basis used in engineering studies
3
rather than as possible consequences of large
annual-miles-per-car differences among categories of insured cars defined by
classification and underwriting rules.
Theory 1 also entails generally unaddressed problems. One is that drivers subject to
tighter budgets as indicated by lower credit scores should be more risk averse and should
be, therefore, if anything, less negligent. Moreover, insurers report that lower credit
scores also predict more uninsured motorist claims per 100 car years. These claims
require as a condition of payment the non-negligence of the insured car’s driver. The cars
belonging to lower credit score drivers must therefore be both more negligently and more
non-negligently involved in accidents.
As an alternative to the driver negligence explanation, Theory 2 proposes that low
credit scores predict more miles per insured car. Significantly, the uninsured motorist
claims problem for Theory 1 is actually a requirement for Theory 2: liability and
uninsured motorist claims must correlate positively. The more miles an insurance
category of cars averages, the more accident involvements and claims per 100 car years
the category must produce, which will include both more negligent (liability) claims and
also more non-negligent (uninsured motorist) claims. This means that compared to an
overall class average miles per car the sub-class of cars belonging to financially-
constrained drivers must be averaging more miles per car.
2
Because the distinction between insured and uninsured cars is not included in the federal travel
surveys, the average mileages for categories of household cars reported by the surveys do not necessarily
represent the averages for matching insurance categories, especially in places where the proportion of
uninsured cars is large.
3
For example, Williams (1999) shows how per-mile risk rates vary strongly with driver age. Age 17
drivers average about 30 state-reported accident involvements per million miles compared with adult driver
involvements of 4-5 per million miles. Drivers over age 79 average about 18 involvements per million
miles.
- 4 -
811.776.9430
The logical basis of Theory 2 is supported by several easily verified givens. First,
accidents are a cost of operating cars. Parked cars rarely cause accidents, but each
odometer mile driven entails a risk of accident and therefore must transfer a statistical but
real cost to the car’s insurer.
4
Statewide, liability claims historically vary directly with the
amount of driving as negatively affected by sharp increases in gasoline prices and
unemployment. Second, as demonstrated by consulting an agent’s manual of rates and
rules, premiums are charged not as a cost of operating cars but of owning them. As long
as classification and coverage are unaffected, adding or subtracting cars from a policy
results in a proportional change in premiums. Finally, premiums are paid in advance of
coverage and are never readjusted at the end of the policy period regardless of how many,
few, or no miles the car was actually driven.
According to Theory 2, traditional pay-per-car premiums must cause adverse
selection under certain circumstances. Per-car prices allow only one certain way to
economize on mandatory insurance: drive fewer cars more miles each. Inconvenience
keeps most drivers from doing this—until the pressure to economize is great. When
drivers remove marginal cars from insurance pools and start to share cars kept insured,
average miles-per-car rises. The result is that insurers correlate more liability claims per
100 car years with lower credit scores and raise prices accordingly (if for no other reason
than fear of being adversely selected against by a competitor that is pricing according to
the credit score indications).
Theory 2 also explains other predictors of liability claims insurers use. Just as more
liability claims correlate with lower credit scores, more claims are predicted for the cars
of residents of lower-income zip codes, more claims for the cars of drivers with lower
educational and occupational levels, more for installment plan premium payers, and more
for cars newly insured after having been uninsured for a period—the so-called no-prior-
insurance variable. Generalizing from these predictors, any marker of a need to
economize predicts more liability claims per 100 car years. (See the top set of predictors
in Table 1.)
In accord with the Theory 1 explanation that low credit scores identify negligent
drivers, Brockett and Golden (2007) cite the use of driver sex and the AAA report (2002)
cites the use of driver records for the same purpose. However, the logic of Theory 2
provides an alternative explanation for both of these traditional predictors, as shown in
Table 1. For example, men average more driving than women the same age and therefore
4
Measurement of the cents-per-mile class rates at which risk is transferred to insurers would require
large numbers of cars in each class (for future statistical stability of the class per-mile rate determined) by
risk-related categories such as car use, residence territory, and driver age. Under today’s car-year exposure
unit, the total cost of past claims for each class is divided by the insured car-years of exposure that
produced the claims to obtain the dollars-per-car-year basis of the future price. Under the odometer-mile
exposure unit the cents-per-mile cost basis of a class price would simply be the total cost of claims divided
by the total insured odometer miles of exposure that produced the class’s claims.
- 5 -
811.776.9430
are involved in more state-reported accidents annually on a per-100-licensed-drivers
basis.
When it comes to Theory 2 explaining why past accidents are predictors of more
claims per 100 car years, accidents may be realistically modeled as random sampling—
not of car year records from company files as employed in the Miller and Smith (2003)
study—but perforce of cars that are on the road. Although the low- and average-miles
cars in an insurance class are sampled by accident involvement, this sampling obviously
will be biased to those cars in the class that spend the most time on the road. This
Table 1
Two explanations for why credit scores and other predictors work
Predictor variable
(of liability claims
per 100 car years)
Correl-
ation
Theory 1
(Variable proxies for driver negligence)
Theory 2
(Variable proxies for
avg. miles per car year)
Credit score negative
“Lack of stability and impulsive behavior
affect both driving and credit history.”*
Zip code income negative
Education and
occupation levels
negative
Installment plan positive
No prior insurance positive
Variables are measures
of need to economize on
liability insurance,
which can be done
directly by giving up
some cars and driving
the insured cars
remaining more miles
each.
Driver sex – man
(Controversial for
adults. Used where
allowed, mainly for
cars accessible to
young drivers)
positive
“[T]he psychobehavioral characteristics of
risk-taking are related to impulsivity,
sensation seeking, aggression, and
sociability with men engaging in more
overall risky behavior than women.”**
At every age men
average more miles than
women, and presumably
so do the cars they drive
relative to the cars
women drive.
At-fault accident
(Use is often
disallowed for small
claims)
positive
“[D]rivers [who were] accident prone in
past are likely to be accident prone in the
future”***
Not-at-fault accident
(Controversial, but
may or may not be
used where allowed)
positive
As sub-classes,
“accident-sampled” cars
continue to average
more miles per car than
the main classes from
which they are
separated.
Car age (not
disallowed but not
used for liability
prices)
negative
Annual mile averages
decrease with car age
* Brockett and Golden, 2007
** Brockett et al., 2005
*** American Academy of Actuaries, 2002.
- 6 -
811.776.9430
sampling process raises the average annual odometer miles of the sub-classes defined by
accident involvement, as modeled by Butler and Butler (1989). Rather than identifying
accident prone drivers in the future, accident records actually define sub-classes that
average more miles per car year in the future than the cars will average in the large
matching accident-free sub-classes.
In addition to the traditional predictors cited by AAA (2002) and by Brockett and
Golden (2007) as validating Theory 1 explanations, however, are equally reliable
predictors that if used would raise difficult questions for auto insurers. A noteworthy
example is that car age is not used for liability pricing even though liability claims per
100 insured car years decrease with car age (McNamara, 1987). If this correlation were
used in pricing, liability premiums would increase for a driver who trades an older for a
newer car. But it would be difficult for Theory 1 to explain how buying a newer car
causes a driver to become more negligent. However, Theory 2 explains that since annual
mile averages decrease with car age, so must claims per 100 car years also decrease with
car age. Trading an older for a newer car does not necessarily change the number of miles
a driver drives whether many or few, but the car they drive definitely changes to a
younger car age group that averages more miles per car.
In 1994 Harrington examined the case that mandatory auto insurance is “taxing low
income households in pursuit of the public interest.” But the case presented against such
taxing is weakened by the implication that low income drivers pay the same insurance
prices as higher income drivers. More recent work by Harrington and Niehaus (1998)
confirms that the cars of lower income drivers produce more liability claims
5
and
consequently are charged higher “taxes” per car year for mandatory liability insurance.
Moreover, according to the present study’s Theory 2, Harrington’s case (1994)
misidentifies the law-abiding choice as “pay or take the bus,” i.e., pay the price of
mandatory insurance or give up driving. Instead, the law-abiding choice that pay-per-car
pricing actually offers is not giving up driving and taking the bus, but giving up cars and
driving the remaining ones more. Hence more miles per car, more claims, and higher
prices must follow in what insurers term “hard to serve markets.” Theory 1 suggests that
more driver negligence in these markets causes the higher prices. But this suggestion
means that—other than to repeal mandatory insurance as auto insurers would have it—
there is no alternative to regulating prices to maintain affordability for the presumed
negligent driver groups insurers identify.
Instead of these undesirable alternatives, however, the strong demand by the public
for enforcing mandatory auto insurance could be accompanied by a strong demand
informed by Theory 2 that automobile insurers provide the audited odometer mile
5
In the Missouri zip codes studied, liability claims per 100 car years exposure averaged 8.25 in the
lower income zip codes which is 36% more than the 6.06 claims the other zip codes averaged.
- 7 -
811.776.9430
exposure unit (Butler, 1993)—an option insurers offer to some fleet owners—as an
option for private passenger car owners. At competitive cents-per-odometer-mile class
prices this option would constitute a free-market remedy for the upward cost-price cycle
that the traditional car-year exposure unit sets off for groups of economizing drivers.
With this option drivers could car pool or take the bus to save on insurance while still
keeping their own cars legally insured and available for use.
Critical to informing a public demand for a remedy to mandated car insurance which
many cannot now afford is engagement by insurance commissioners and consumer
advocates, as well as scholars, with the explanation offered by Brockett and Golden
(2007) and the alternative explanation provided by this essay for why low credit scores
and like correlations work to predict more liability claims per 100 insured car years.
REFERENCES
American Academy of Actuaries, 2002. The Use of Credit History for Personal Lines of
Insurance: Report to the National Association of Insurance Commissioners, Nov. 15,
Washington, DC: AAA.
Brockett, Patrick L., Linda L. Golden, and Sandra H. Dunn, 2005. “Biological and
Psycho-behavioral Correlates of Risk Taking, Credit Scores, and Automobile
Insurance Losses: Toward an Explication of Why Credit Scoring Works,” World
Risk & Insurance Economics Conference, August 9, 2005, Salt Lake City, Utah.
Brockett, Patrick L., and Linda L. Golden, 2007. “Biological and Psychobehavioral
Correlates of Credit Scores and Automobile Insurance Losses: Toward an
Explication of Why Credit Scoring Works.” Journal of Risk and Insurance, 74: 23-
63
Butler, Patrick, 1993. “Operation of an Audited-Mile/Year Automobile Insurance System
Under Pennsylvania Law,” Casualty Actuarial Society Forum, Summer 1993: 307-
338, available at http://www.casact.org/pubs/forum/93sforum/93sf307.pdf
Butler, Patrick, and Twiss Butler, 1989. “Driver Record: a Political Red Herring That
Reveals the Basic Flaw in Automobile Insurance Pricing,” Journal of Insurance
Regulation, 8: 200-234.
Federal Trade Commission, 2007. “Credit-Based Insurance Scores: Impacts On
Consumers Of Automobile Insurance.” A Report To Congress by the Federal Trade
Commission, July 2007.
http://www.ftc.gov/os/2007/07/P044804FACTA_Report_Credit-
Based_Insurance_Scores.pdf
Harrington, Scott E., 1994. “Taxing Low Income Households in Pursuit of the Public
Interest: The Case of Compulsory Automobile Insurance.” In Sandra Gustavson and
Scott Harrington, eds. Insurance, Risk Management, and Public Policy. Boston,
Kluwer Academic Publishers.
Harrington, Scott, and Greg Niehaus, 1998. “Race, Redlining, and Automobile Insurance
Prices,” Journal of Business, 71: 439-69
- 8 -
811.776.9430
Hu, Pat S. And Timothy R. Reuscher, 2004. “Summary of Travel Trends, 2001 National
Household Travel Survey,” Federal Highway Administration, available at
http://nhts.ornl.gov/2001/pub/STT.pdf
McNamara, Daniel J., 1987. “Discrimination in Property-Liability Insurance Pricing,” in
Issues in Insurance,
1-67, Everett D. Randall ed., 4th Ed.
Miller, Michael J., and Richard Smith, 2003, “The Relationship Between Credit-Based
Insurance Scores to Private Passenger Automobile Insurance Loss Propensity.”
Bloomington, IL: Epic Actuaries, LLC, available at http://www.ask-
epic.com/Publications/Relationship%20of%20Credit%20Scores_062003.pdf
Williams, Allan, 1999. “Licensing Policies for Young Drivers in the United States,” in
Automobile Insurance: Road Safety, New Drivers, Risks, Insurance Fraud and
Regulation, edited by Georges Dionne and Claire Laberge-Nadeau, Kluwer
Academic Publishers, Boston: 215-220.
Karen Weldin Stewart, CIR-ML
Commissioner
Delaware Department of Insurance
841 Silver Lake Blvd., Dover, DE 19904-2465 www.delawareinsurance.gov
(302) 674-7300 Dover (302) 739-5280 fax (302) 577-5280 Wilmington
April 27, 2009
Property and Casualty (C) Committee
Market Regulation and Consumer Affairs (D) Committee
Public Hearing on Credit-Based Insurance Scores
Arlington, Virginia
Director Michael McRaith (IL)
Commissioner Kim Holland (OK).
Dear Director McRaith and Commissioner Holland:
The Delaware Insurance Commissioner's Office worked with the Delaware General Assembly to pass a
new law in 2007 that makes Delaware one of the five strictest states in the country with regards to how
credit information can be used in setting insurance rates. The new law took effect on January 1, 2008.
Many Delawareans believe credit scoring would increase the cost of insurance for those who are already
least able to afford insurance and who are already having the most difficulty paying their bills.
Further, it has a disproportionate impact on poor and minority policyholders.
As of January 1, 2008, auto and homeowners insurance companies in Delaware have been prohibited
from adjusting their current policyholders’ rates based on changes in the policyholders’ credit ratings.
This ban is the result of legislation passed in June 2007 by the General Assembly and signed into law by
the Governor.
Policyholders can also take advantage of one provision of the new law. After January 1, 2008
policyholders can ask their insurance companies to recalculate their credit ratings once a year when their
policies come up for renewal and if their credit has improved they may see a benefit in their insurance
rates. If their credit has worsened, their insurance rates will not be affected.
Commissioner Stewart said: "The law, consistent with my long held position, is that a change in a
policyholder's credit score must only help and not harm a current policyholder. ” The
Commissioner further stated: “Thanks to our new law, policyholders have a right to ask for this possible
discount without the need to read or understand any deceiving or contradictory fine print. Why?
Because there isn't any."
Karen Weldin Stewart, CIR-ML
Commissioner
Delaware Department of Insurance
841 Silver Lake Blvd., Dover, DE 19904-2465 www.delawareinsurance.gov
(302) 674-7300 Dover (302) 739-5280 fax (302) 577-5280 Wilmington
Page 2
April 27, 2009
A regulation requiring insurance companies to notify their policyholders of this new right to have their
credit recalculated at renewal took effect on April 1, 2008.
The Delaware Insurance Department provides consumers with a guide to Delaware's new law regarding
insurance and credit. The guide is available online at http://delawareinsurance.gov/credit
.
As respects to our concerns we would like the following issues addressed during the hearing:
How do insurers handle improvement (or deterioration) in scores from year-to-year?
How do insurers handle those who wish not to have their scores checked, thin-files, or no-hits?
What variety is out there for what insurers do with the scores?
Least use: not use credit at all. Next: yes/no for eligibility for insurance. Next: use for tiering
(underwriting only, not rating). Next: Use as rating factor. Most use: integrated into the rating through
a multi-point model.
How do insurers avoid double-counting so-called negative characteristics, like urban dwellers with poor
credit, or youthful drivers with poor credit, or low-valued homes whose owners have poor credit?
We appreciate the opportunity to submit our comments and concerns and look forward to the results of
the hearing.
Sincerely,
Karen Weldin Stewart, CIR-ML
Delaware Insurance Commissioner
April 28, 2009
Commissioner Kim Holland Director Michael McRaith
Oklahoma Insurance Department Illinois Division of Insurance
P.O. Box 53408 320 W. Washington Street
Oklahoma City, OK 73152-3408 Springfield, IL 62767-0001
Dear Regulators:
As leaders of the National Conference of Insurance Legislators (NCOIL), we write in advance of your
April 30 hearing on insurance credit scoring because we also have recognized the importance of this
issue. We wish to offer insight into the 2002 model law that we developed in response—specifically, into
the model’s long-standing and well-accepted consumer protections. We also note that NCOIL will soon
revise its model to more directly assist victims of today’s economic climate.
The NCOIL Model Act Regarding Use of Credit Information in Personal Insurance was and remains a
timely and effective response to consumers who feel blindsided by insurer use of their credit information.
NCOIL believes strongly that insurers should not have free reign and that we as legislators have a duty to
promote balanced public policy that safeguards our constituents from possible abuse. The NCOIL model
act—which evolved through two years of special sessions, model drafts, and many hours of debate with
all key players—does just that. It has become the standard for state insurance scoring policy and would,
among other things:
• promote so-called credit “passes” for persons impacted by extraordinary life events—
such as divorce, illness, job loss, or death of a spouse—which could encompass fallout
from the financial crisis.
• prohibit an insurer from calculating an insurance score based on income, gender, address,
zip code, ethnic group, religion, marital status, or nationality.
• discourage insurers from taking an adverse action based on “thin” or non-existent credit—
a circumstance common among seniors, young people, and low-income consumers.
• prohibit insurers from treating negatively credit inquiries that a consumer did not initiate—
as when banks mine credit reports prior to sending credit card offers.
• prohibit insurers from looking negatively upon collection accounts related to a sickness or
other medical event for which a consumer could not pay.
• provide that insurers can only consider multiple inquiries from the mortgage or auto lending
industries in any 30-day period as one credit “hit”—as these multiple inquiries indicate that a
consumer has wisely shopped around for the best deal.
• require insurers, before taking an adverse action, to use credit reports issued or insurance scores
calculated within 90 days from the time a policy was initiated or renewed.
• direct insurers to re-underwrite or re-rate if a consumer or his/her agent requests it at
annual renewal.
• require an insurer, if a consumer challenges his/her credit report and has it corrected, to
re-underwrite or re-rate based on the new information—and return any amount of overpayment.
• mandate that insurers provide key consumer notifications—including up to four credit-related
reasons for an adverse action if credit was a factor and up-front notice that credit will be
considered in underwriting and/or rating.
• direct insurers to file their insurance scoring models with state insurance departments.
• prohibit credit reporting agencies from selling insurance-related data to third-parties that
do not deserve it.
NCOIL will consider an amendment at the July 9 through 12 NCOIL Philadelphia Summer Meeting to
target consumers whose fallen credit is traceable to the financial crisis not of their making—even though
the model’s extraordinary life events language already applies.
The amendment would correlate the time a person’s credit began suffering with decline of the U.S.
economy. It would specifically acknowledge certain major credit events, such as foreclosures, that run
counter to a person’s otherwise solid credit history. The amendment will be drafted to capture only
people who are true crisis victims, rather than people who are victims of their own poor decisions.
NCOIL will have further details as the Summer Meeting approaches.
We look forward to dialoging with you regarding suitable insurance-scoring public policy, and we
appreciate your efforts during this difficult time. As legislators and regulators—and first and foremost
consumers—we agree that too many in this country are struggling to pay bills and stay in homes. We
must stand together to protect consumers from credit misuse and encourage strong insurance markets that
benefit us all. The NCOIL model law—which has been adopted in your states of Illinois and Oklahoma,
as well as in 24 others—strikes the critical balance.
Please feel free to contact the NCOIL National Office at 518-687-0178 should you have any questions.
Sincerely,
Sen. James Seward (NY) Rep. Robert Damron (KY)
NCOIL President NCOIL President-Elect
Rep. George Keiser (ND) Sen. Carroll Leavell (NM) Sen. Vi Simpson (IN)
NCOIL Vice President NCOIL Secretary NCOIL Treasurer
cc: NCOIL Legislators
NAIC Market Regulation & Consumer Affairs (D) Committee
NAIC Property-Casualty Insurance (C) Committee
Therese Vaughan
Tim Mullen/Eric Nordman
K:/NCOIL/2009 Documents/2006367.doc
