f
A Discussion on The Use of
Credit Information and Scoring
for Insurance Underwriting
Eddy Lo
March 12, 2001
CAS Ratemaking Seminar, Las Vegas
f
Topics
1.
2.
3.
4
5.
6.
7.
8.
Introduction and Objectives
Fair, Isaac and Company, Inc.
Fair Credit Reporting Act
Predictiveness
Fairness
Accuracy
Inquiries
Current Operations
© 1999
© 1999
Fair,
Fair,
Isaac
Isaac
andand
Company,
Co., Inc.Inc.
2
f
Topics
9. Statistical Correlation
10.Scoring Definitions
11. Scorecard Examples
12.Results
13.Usage of Insurance Bureau Scores
14.Summary
15.Questions & Answers
© 1999
© 1999
Fair,
Fair,
Isaac
Isaac
andand
Company,
Co., Inc.Inc.
3
f
Introduction and Objectives


Provide facts on the use of insurance bureau
scores
Answer questions on insurance bureau
scores
© 1999 Fair, Isaac and Co., Inc.
4
f
Fair, Isaac & Company, Inc.

Founded in 1956, by



Starting out


William R. Fair
Earl J. Isaac
Better credit decisions by statistics than
traditional judgmental methods
Now

Better Decisions Through Data
© 1999 Fair, Isaac and Co., Inc.
5
f
Fair, Isaac & Company, Inc.

Values





Absolute integrity
Very high standards of excellence of
product and service
Personal commitment via championing
Collegial atmosphere
Strive for constant innovation
© 1999 Fair, Isaac and Co., Inc.
6
f
Fair, Isaac & Company, Inc.


Developed unique modeling processes
based on documented models and
proprietary algorithm
National award examples


Forbes Top 200 Small Companies list,
Honor Rolls in 1999, 1998, and other
years
Future Banker “1999 Top 25 Technology
Deals” for alliance with eCredit.com, and
other years
© 1999 Fair, Isaac and Co., Inc.
7
f
Fair, Isaac & Company, Inc.

National award examples (cont’d)




Credit Risk Management Report award
for Best Scoring Model, 1998 & 1995
ABA Bank Card Distinguished Service
Award, September 1997
Financial World Magazine “One of the
100 Best Growth Companies” 1997
President Corporate Award; Society of
Insurance Research 1995
© 1999 Fair, Isaac and Co., Inc.
8
f
Fair, Isaac & Company, Inc.

Industries served


Insurance, Finance Services,
Government, Healthcare, E-Business,
Telecommunications
Global experience

Offices and representations on 6
continents
© 1999 Fair, Isaac and Co., Inc.
9
f
Fair Isaac Worldwide
TORONTO, CANADA
Fair, Isaac
ST. PAUL
MINNESOTA
OFFICES
Representation
BIRMINGHAM, UNITED KINGDOM
Antwerp,
Helsinki, Finland
Belgium
WIESBADEN,
GERMANY
TOKYO,
JAPAN
PARIS,
FRANCE
* SAN RAFAEL,
CALIFORNIA
HEADQUARTERS
Madrid,
Spain
Istanbul,
Turkey
Phoenix,
Arizona
MEXICO CITY,
MEXICO
Kuala Lumpur,
Malaysia
ATLANTA,
GEORGIA
NEW CASTLE,
DELAWARE
Santiago,
Chile
JOHANNESBURG,
SOUTH AFRICA
Melbourne,
Australia
Sydney,
Australia
© 1999 Fair, Isaac and Co., Inc.
10
f
Fair, Isaac & Company, Inc.

Participation in NAIC White Paper
 “Credit Reports and Insurance
Underwriting”
 NAIC Subgroup visited Fair, Isaac in
August 1995
 Participated in an industry working group
 Issued acceptable principles in
October 1995
 NAIC Subgroup Chairman joined Fair,
Isaac InterACT ‘96 educational
conference
© 1999 Fair, Isaac and Co., Inc.
11
f
Fair, Isaac & Company, Inc.

Tillinghast study of




Insurance Bureau Scores Vs. Loss Ratio
Relativities
Presented to NAIC Subgroup in
December 1996
Included in White Paper Appendix
White Paper adopted by NAIC in December
1996
© 1999 Fair, Isaac and Co., Inc.
12
f
Fair, Isaac & Company, Inc.

Presentations







American Agents Alliance
American Insurance Association
Alliance of American Insurers
Association of Insurance and Financial
Analysts
Casualty Actuarial Society
Chartered Property and Casualty
Underwriters
National Association of Independent
Insurers
© 1999 Fair, Isaac and Co., Inc.
13
f
Fair, Isaac & Company, Inc.

Presentations (cont’d)







National Association of Insurance
Commissioners
Neighborhood Reinvestment Corporation
Professional Insurance Agents
Independent Insurance Agent Association
Insurance departments and legislators
Reinsurance Association of America
Others
© 1999 Fair, Isaac and Co., Inc.
14
f
Fair Credit Reporting Act
(FCRA)




Original statute in 1970
Major amendments in 1996; effective
September 30, 1997
Requires “consumer reporting agencies” to
adopt procedures governing accuracy,
relevancy, access to and utilization of
“consumer reports”
Allows consumers access to their files and a
complaint procedure
© 1999 Fair, Isaac and Co., Inc.
15
f
Fair Credit Reporting Act
(cont’d)


Requires users of consumer reports to certify
the permissible purpose(s) and use only for
certified (permissible) purpose(s); and to give
FCRA adverse action notices
Imposes accuracy-oriented obligations on
furnishers of information
© 1999 Fair, Isaac and Co., Inc.
16
f
Fair Credit Reporting Act
(cont’)

Permissible purposes




Court order or written instructions of
consumer
Use in connection with a credit
transaction involving the consumer; credit
extensions/review of accounts/collections
Use for underwriting insurance
Employment
© 1999 Fair, Isaac and Co., Inc.
17
f
Fair Credit Reporting Act

Permissible purposes (cont’d)

Use by person with other legitimate
business need for information in
connection with a business transaction
initiated by the consumer, or to review an
account to determine whether the
consumer continues to meet the terms of
the account
© 1999 Fair, Isaac and Co., Inc.
18
f
Fair Credit Reporting Act

Permissible purposes (cont’d)

Prescreening: Use for “transaction not
initiated by consumer” for “firm offer of
credit or insurance”, permit conditioning
the offer on verification of information in
credit report or application to ensure that
consumer still meets the prescreen
criteria at time of acceptance; may also
condition offer on information in
application meeting pre-established
criteria, or on the furnishing of required
collateral as disclosed in the offer
© 1999 Fair, Isaac and Co., Inc.
19
Predictiveness
f
Rank Ordering
Homeowner
1.8
1.6
Loss Ratio Relativity
1.4
1.2
1.0
0.8
0.6
0.4
0.2
0.0
low
high
Score Range
© 1999 Fair, Isaac and Co., Inc.
20
Predictiveness
f
Rank Ordering
Personal Auto
1.4
Loss Ratio Relativity
1.2
1.0
0.8
0.6
0.4
0.2
0.0
low
high
Score Range
© 1999 Fair, Isaac and Co., Inc.
21
f
Predictiveness



Low scores correlate with high loss ratio
relativities
High scores correlate with low loss ratio
relativities
Validated by


Insurers
Tillinghast
© 1999 Fair, Isaac and Co., Inc.
22
f
Fairness

Data Elements Used





SEPTEMBER
3
4
5

6

10 11 12 13
17 18 19 20
24 25 26 27



Balances
Collections
Delinquencies
Inquiries
Limits
Payment Dates
Payment Due Dates
Public Records
Trade Line Open and Close Dates
Trade Line Types
© 1999 Fair, Isaac and Co., Inc.
23
f
Fairness (cont’d)

Data Elements Not Used





SEPTEMBER
3
4
5

6
10 11 12 13

Age
Disability
Gender
Health Status
Income
Location
Marital Status
17 18 19 20
24 25 26 27
© 1999 Fair, Isaac and Co., Inc.
24
f
Fairness

Data Elements Not Used (cont’d)





SEPTEMBER
3
4
5

6
10 11 12 13

Nationality
Net Worth
Occupation
Race
Religion
Sexual Orientation
Zip Codes
17 18 19 20
24 25 26 27
© 1999 Fair, Isaac and Co., Inc.
25
f
Fairness (cont’d)

Income Study by an Insurer
Flat relationship between income levels and
scores
Virginia Bureau of Insurance
 ‘Use of Credit Reports in Underwriting’, 1999
report
 To the Senate Commerce and Labor
Committee of the the General Assembly of
Virginia
 “… Nothing in this analysis leads the Bureau
to the conclusion that income or race alone
is a reliable predictor of credit scores thus
making the use of credit scoring an
ineffective tool for redlining. …”


SEPTEMBER
3
4
5
6
10 11 12 13
17 18 19 20
24 25 26 27
© 1999 Fair, Isaac and Co., Inc.
26
f
Accuracy

FCRA mandate correction process

1992 Arthur Anderson Study




Commissioned by Associated Credit
Bureaus
Based on 15,202 declines
2% dispute on declines
MVR accepted by most regulators

higher error rates
© 1999 Fair, Isaac and Co., Inc.
27
f
Inquiries
Homeowner
1.2
Loss Ratio Relativity
1.1
1
0.9
0.8
0.7
0.6
0.5
0.4
0
1
2
3
4
5+
Inquiries
© 1999 Fair, Isaac and Co., Inc.
28
f
Inquiries
Personal Auto
1.2
Loss Ratio Relativity
1.1
1
0.9
0.8
0.7
0.6
0.5
0.4
0
1
2
3
4
5+
Inquiries
© 1999 Fair, Isaac and Co., Inc.
29
f
Inquiries



Predictive of loss ratio relativities
Fair, Isaac includes consumer-initiated
inquiries
Fair, Isaac excludes inquiries for



Marketing / prescreening
Account reviews
Insurance
© 1999 Fair, Isaac and Co., Inc.
30
f
Current Operations
Insurer
Underwriting
Credit Bureau
Credit
Database
Scorecards
Subscriber Requests
Scores and Services
Royalty
Maintenance
Fair, Isaac
Scorecards
© 1999 Fair, Isaac and Co., Inc.
31
f
Statistical Correlation

Personal property
 230,000 policies with claims
 1,000,000 policies without claims
 11 archives
© 1999 Fair, Isaac and Co., Inc.
32
f
Statistical Correlation (cont’d)

Homeowner univariate analyses
 Number of adverse public records
 Months since most recent adverse public
record
 Number of trade lines 60+ days
delinquent in last 24 months
 Number of collections
 Number of trade lines opened in the last
12 months
© 1999 Fair, Isaac and Co., Inc.
33
f
Statistical Correlation
Homeowner HO - 3
Number of Adverse Public Records
1540
1600
Loss Ratio Relativity
1400
1200
1000
1000
800
600
400
200
0
zero
one or more
96%
© 1999 Fair, Isaac and Co., Inc.
34
f
Statistical Correlation
Homeowner HO - 3
Months Since Most Recent Adverse Public Record
1800
1678
1600
Loss Ratio Relativity
1400
1226
1200
1000
1000
800
600
400
200
0
no public record
less than 48
48 or more
96%
© 1999 Fair, Isaac and Co., Inc.
35
f
2000
Statistical Correlation
Homeowner HO - 3
Number of Trade Lines 60+ Days Delinquent
in Last 24 Months
1804
1800
Loss Ratio Relativity
1600
1400
1293
1200
1000
1000
800
600
400
200
0
ze ro
one
two or more
89%
© 1999 Fair, Isaac and Co., Inc.
36
f
1800
Statistical Correlation
Homeowner HO - 3
Number of Collections
1686
1600
Loss Ratio Relativity
1400
1200
1000
1000
800
600
400
200
0
zero
97%
one or more
© 1999 Fair, Isaac and Co., Inc.
37
f
1800
Statistical Correlation
Homeowner HO - 3
Number of Trade Lines Opened in the Last 12 Months
1658
1600
1503
Loss Ratio Relativity
1400
1147
1200
1220
1000
1000
800
600
400
200
0
zero
one
two
three
four or more
60%
© 1999 Fair, Isaac and Co., Inc.
38
f
Statistical Correlation (cont’d)

Personal auto
 350,000 policies with claims
 1,000,000 policies without claims
 6 archives
© 1999 Fair, Isaac and Co., Inc.
39
f
Statistical Correlation (cont’d)

Personal auto univariate analyses
 Number of adverse public records
 Months since most recent adverse public
record
 Number of trade lines 60+ days
delinquent in last 24 months
 Number of collections
 Number of trade lines opened in the last
12 months
© 1999 Fair, Isaac and Co., Inc.
40
Statistical Correlation
f
Personal Automobile
Number of Adverse Public Records
1400
1225
1200
Loss Ratio Relativity
1000
1000
800
600
400
200
0
zero
one or more
97%
© 1999 Fair, Isaac and Co., Inc.
41
f
1600
Statistical Correlation
Personal Automobile
Months Since Most Recent Adverse Public Record
1339
1400
1182
Loss Ratio Relativity
1200
1000
1000
800
600
400
200
0
no public record
97%
less than 18
© 1999 Fair, Isaac and Co., Inc.
18 or more
42
f
Statistical Correlation
Personal Automobile
Number of Trade Lines 60+ Days Delinquent
in Last 24 Months
1600
1444
Loss Ratio Relativity
1400
1238
1200
1000
1000
800
600
400
200
0
zero
one
two or more
86%
© 1999 Fair, Isaac and Co., Inc.
43
f
Statistical Correlation
Personal Automobile
Number of Collections
1600
1494
1400
Loss Ratio Relativity
1200
1000
1000
800
600
400
200
0
zero
one or more
96%
© 1999 Fair, Isaac and Co., Inc.
44
Statistical Correlation
f
Personal Automobile
Number of Trade Lines Opened in the Last 12 Months
1400
1270
1200
1083
Loss Ratio Relativity
1000
1000
800
600
400
200
0
zero or one
two or three
four or more
82%
© 1999 Fair, Isaac and Co., Inc.
45
f
Scoring Definitions

A score for an insurance risk




Is a numeric summary
Of the impact on loss ratio relativity
Based on a certain set of predictive
characteristics of the risk
A model/scorecard is an algorithm, a table, or
a piece of computer software



That will calculate a score
Based on a certain set of characteristics
Provided for a risk
© 1999 Fair, Isaac and Co., Inc.
46
f
Scoring Definitions (cont’d)

The 4 reason codes for a score are the 4
reasons that contributed most significantly,
positively or negatively, to the calculation of a
score
© 1999 Fair, Isaac and Co., Inc.
47
f
Scorecard Examples

Simple homeowner scorecard

Overlapping characteristics

weights adjusted
© 1999 Fair, Isaac and Co., Inc.
48
f
Scorecard Examples
Simple Homeowner Scorecard
Number Adverse Public Records
zero
30
one or more
0
Months Since Most Recent Adverse Public
Record
no public record
less than 48
48 or more
0
10
30
© 1999 Fair, Isaac and Co., Inc.
49
f
Scorecard Examples
Simple Homeowner Scorecard
Number of Trade Lines 60+ Days Delinquent in
Last 24 Months
zero
one
25
10
two or more
0
Number of Collections
zero
20
one or more
0
© 1999 Fair, Isaac and Co., Inc.
50
f
Scorecard Examples
Simple Homeowner Scorecard
Number of Trade Lines Opened on the Last 12
Months
zero
one
two
three
four or more
20
10
5
3
0
© 1999 Fair, Isaac and Co., Inc.
51
f
Results

Insurance Bureau Scores Vs. Loss Ratio
Relativities



Multivariate analysis - homeowner (HO1,
HO2, HO3, HO4, HO6, dwelling fire)
Multivariate analysis - personal auto (nonstandard, standard minimum limits,
standard above minimum limits, preferred
minimum limits, preferred above minimum
limits)
Risk ranking
© 1999 Fair, Isaac and Co., Inc.
52
f
Results

Insurance Bureau Score Vs. Loss Ratio
Relativities (cont’d)
 Development vs. validation datasets
 Validation (handouts)
 Tillinghast study; Conclusion, page 5;
“…The data for all companies included in this study
except Company 2 indicates at least 99%
probability that a relationship exists. The data for
Company 2 indicates a 92% probability that there
is a relationship. A layman’s interpretation of this
result could be that it is very likely there is a
correlation between Insurance Bureau Scores and
loss ratio relativities.”
© 1999 Fair, Isaac and Co., Inc.
53
f
Usage of Insurance Bureau
Scores

Facilitate



underwriting applications
underwriting investigation
tier placement
© 1999 Fair, Isaac and Co., Inc.
54
Loss Ratio Relativities
f
Illustration of
Underwriting Acceptance
1.4
0.9
0.4
650-674
640-660
referral
675-699
700-724
725-749
750-774
Score Range
© 1999 Fair, Isaac and Co., Inc.
55
Illustration of
Underwriting Tier Placement
Loss Ratio Relativities
f
Tier 1
1.8
1.6
1.4
1.2
1.0
0.8
0.6
0.4
<650
650-674
Tier 2
675-699
Tier 3
700-724 725-749
750-774
775-799
800+
Score Range
665-685
740-760
r ef er r al
r ef er r al
© 1999 Fair, Isaac and Co., Inc.
56
f
Summary




FCRA Makes Insurance Bureau Scores
Usable for Insurance Underwriting and
Marketing
Poll says Insurance Bureau Scores are
favored
Tillinghast Study confirms loss ratio
relativities and Insurance Bureau Scores
relationship
Credit reports more accurate than Motor
Vehicle Reports
 Credit report accuracy further enhanced
by corrections
© 1999 Fair, Isaac and Co., Inc.
57
f
Summary


Insurance Bureau Scores summarize credit
history succinctly and nothing else
The relationship between how people
maintain their credit and property is simply
common sense
 Good credit managers are good risk
managers
 Credit management reflected in Insurance
Bureau Scores
© 1999 Fair, Isaac and Co., Inc.
58
f
Summary

Insurance Bureau Scores deliver a fair shake
 Insurance Bureau Scores do not look at
race, creed, gender, marital status,
income, age, etc.
 Insurance Bureau Scores do not worsen
discrimination nor add to it
 Scoring remedies discrimination
 Insurance Bureau Scores can control
discrimination
 Insurance Bureau Scores do not unfairly
discriminate
© 1999 Fair, Isaac and Co., Inc.
59
f
Summary


Insurance Bureau Scores help to open up
markets
Scoring leads to precision underwriting
 Insurance Bureau Scores facilitate
consistent underwriting
 Insurance Bureau Scores don’t make
decisions, people do
 Insurance Bureau Scores provide input to
refine decisions
 Insurance Bureau Scores provide more
objectivity and accuracy
© 1999 Fair, Isaac and Co., Inc.
60
f
Summary




Insurance Bureau Scores help underwriters
focus on risks needing attention most
Insurance Bureau Scores help to reduce
premium subsidies/inequity
Insurance Bureau Scores strengthen insurer
solvency
Fair, Isaac expertise to share
© 1999 Fair, Isaac and Co., Inc.
61
f
Materials
1. Reasons and Codes
2. Answers to Your Questions about
Insurance Bureau Scores
© 1999 Fair, Isaac and Co., Inc.
62
f
Questions & Answers