Docstoc

particularly in the form of credit scoring

Document Sample
particularly in the form of credit scoring Powered By Docstoc
					Use of Credit Scoring in Auto Insurance
Pricing: Consumer Advocate Perspective
Contents


•Consumer Advocate Stance
•Violation of Risk Classification Statement of Principles
•Rationale for Limitation/Prohibition
•Solutions/Compromises
Consumer Advocate
     Stance




               Pwc
Robert James Hunter, Jr, FCAS, MAAA
Director of Insurance, Consumer Federation of America


  “The CFA has serious doubts about credit scoring as a
  reasonable way to determine insurability and price. One of our
  primary concerns is that no one can tell us why credit scoring
  works. They can only allege that there are strong statistical
  relationships between credit scoring and insurance results.”
                       2001 Winter National Meeting of the NAIC
Birny Birnbaum, FCAS, MAAA
Director, Center for Economic Justice

 “Insurers’ use of consumer credit information — particularly in
 the form of credit scoring — is inherently unfair and violates
 basic risk classification principles.”
                                       2002 Spring CAS Meeting
  Risk Classification
Statement of Principles




                    Pwc
Risk Classification Purposes


I.    Protection of the insurance program’s financial soundness
II.   Enhance fairness
III. Economic incentive
Statistical Considerations


I.    Homogeneity
II.   Credibility
III. Predictive stability
Operational Considerations

I.    Expense
      - increases overall system costs (I.e., credit reports and licensing)

II.   Constancy
III. Availability of coverage
IV. Avoidance of extreme discontinuities
      - creates excessive tiers

V.    Absence of ambiguity
VI. Manipulation
      - use of credit repair agencies or rapid re-scoring

VII. Measurability
      - credit scores varies by source/industry algorithms vary significantly
      from those used by credit bureaus
Hazard Reduction Incentives

• Risk classification systems should provide incentives for
  insureds to reduce expected losses and mitigate incurred
  losses
  - surcharges for speeding and discounts for anti-theft devices

• Desirable, but not necessary, feature
Public Acceptability Considerations


Risk classification systems:
1. should not differentiate unfairly among risks
2. should be based upon clearly relevant data
3. should respect personal privacy
4. should be structured so that the risks tend to identify naturally
   with their classification
Causality


•   Demonstrable cause and effect relationships increase public
    acceptability
•   Often statistically difficult to prove
•   Causality is not a risk classification system requirement
•   Plausible relationship between class characteristics and the
    hazard insured against
•   Risk classification characteristics should be neither obscure
    nor irrelevant
Controllability


•   Desirable quality for a risk classification system
•   Associated with undesirable qualities (i.e., manipulation,
    impracticality, irrelevance to future cost predictability)
    Rationale for
Limitation/Prohibition




                   Pwc
Unfairly Discriminatory


•   Greatest impact on poor and minority communities
    - quality of credit scores is not equally distributed across socioeconomic,
    age and racial classes

•   Similar to using age and home value as a rating factor —
    impacted poor insureds with older, less expensive dwellings
•   Sophisticated methodology for redlining
•   Redistributes premium from one group to another — zero net
    effect on total premium
Unfairly Discriminatory (continued)


•   1999 National Consumer Credit Survey by Freddie Mac:
        30% of consumers have “bad” credit records
        13% of consumers have “intermediate” credit records
        57% of consumers have “good” credit records

        36% of consumers (income <$25,000) have “bad” credit
        33% of consumers (income $25,000 to $44,999) have “bad” credit
        25% of consumers (income $45,000 to $64,999) have “bad” credit
        22% of consumers (income $65,000 to $74,999) have “bad” credit
Unfairly Discriminatory (continued)


•   1999 National Consumer Credit Survey (continued):
        48% of African-Americans have “bad” credit
        16% of African-Americans have “intermediate” credit
        36% of African-Americans have “good” credit

        34% of Hispanics have “bad” credit
        15% of Hispanics have “intermediate” credit
        51% of Hispanics have “good” credit

        27% of Whites have “bad” credit
        12% of Whites have “intermediate” credit
        61% of Whites have “good” credit
Inherently Unfair to Consumers

•   September 11th attacks
•   Causes of bankruptcies and delinquencies
    - correlation of -89.1% and -66.5% to 1987-99 C/W auto incurred LR,
    respectively

•   ID theft
•   Bank decisions to loosen or tighten credit standards
•   Overall economic conditions
•   Regional variations in credit characteristics
    - MGIC Investment Corp. mortgage study during 1989-91 recession

•   Claim financial responsibility predicts propensity for loss
Effectively Deregulates Insurance


•   No regulation of credit scoring vendors
•   Vendors refuse independent testing
•   Simple, univariate correlation is insufficient (LR only
    dependent variable)
•   Credit history has greater impact on insureds’ premiums than
    driving record, vehicle use or miles driven
•   Insurers can increase premiums by re-tiering insureds without
    rate increases
Solutions/Compromises




                  Pwc
Issues


•   Permit independent analysis of the correlation claim
•   Perform multivariate analysis utilizing exposures and claims
•   Cease applying credit tier factors to Comprehensive coverage

				
DOCUMENT INFO
Shared By:
Categories:
Tags:
Stats:
views:41
posted:11/16/2012
language:English
pages:21