A View From The Press Box

This is just another example of how truth is being pushed out in favor of emotion more and more.

Credit variables relate to insurance USAGE--i.e. how often one claims. They do not penalize for credit crises, but they do penalize for using lots of credit because that is highly correlated to insurance usage.

And compared to other data sources, credit is MORE accurate because people CARE more about it. That's part of why it works so well--the data's good.

The main thing insurers need to do is stop anyone when they begin saying "How can credit say whether or not you are going to have an accident..." and point out that it is whether or not a claim results that matters, and credit does a GREAT job figuring that out.

Intuition and uncertainty do not mix, but that's why insurers will always be under assault on this. Everyone thinks they know better when they have no clue.

Todd R. Bault, FCAS
Senior Analyst, Non-Life Insurance
Sanford C. Bernstein, LLC
1345 Avenue of the Americas
15th Floor
New York, NY 10105
Ph: 212 756-1857
Fx: 212 848-2370
todd.bault@bernstein.com

The comments herein are part of a larger body of investment analysis. For our research reports, which contain information that may be used to support investment decisions and appropriate disclosures, please see our website at http://www.bernsteinresearch.com

There is now a new spin on credit scoring--that it is a bankruptcy predictor....another scam to generate additional interest rates for banks and or additional premiums for insurance companies.

Credit scoring is a discrimminator to small-business owners, minorities and women. As a former Commissioner of Financial Institutions for Louisiana, I can attest to the fact that most small banks do not use these scores as they omit financial capacity, new worth, life insurance and the like, to name few.

Easy.

If there is a strong relationship between a credit score and relative risk/insurability, then that data should be made public. On what coverages? Physical damage, collision, BI, PD, UM, UIM, all?

People intuitively understand the relationship between a MVR and relative insurability and maybe frequency of accidents (at fault?), age, sex, location of the garage, etc. and the premiums they pay. Those are facts.

Alas, credit scores are, in far too many cases, prone to error and circumstance and thus are not linked by the average insured to insurance premiums in any meaningful way.

Prove a solid link of one's credit score to insurability/risk to the satisfaction of insureds, and resistance will evaporate.

If the industry can't sell it to a cynical public, then do some research and come up with something better than a usually misleading credit score. Maybe a driver's housekeeping in the insured vehicle glove compartment may prove to be a correlation. That should sell.

Bob Holland

It seems to me the question to be answered is, does the use of credit accurately predict losses?

If the answer is no, then why have so many insurers switched to using it? And why have their loss ratios improved?

If the answer is yes and the stats back it up, then let's quit having this debate.

The stats show there is definite link between poor credit and poor driving habits. Basically it is a life-style choice people make, and the way some people treat credit relates to the way they drive.

In some states the insurers have to rerun the credit check every 3 years, so if the person has a better score they get a better rate.

One could also make the argument that just because someone has speeding tickets does not mean they will have a accident or a claim, and basically cost the insurer nothing. So should we change the system for those also? I beleive most will say no.

In automobile insurance, lower credit-based insurance scores correlate strongly with higher claim frequencies per 100 car years for all coverages, including (significantly) Uninsured Motor Vehicle (UMV).

The prevailing explanation is that car owners with credit problems are more negligent drivers than those with better credit.

The alternative explanation I propose as more congruent with facts is that low credit signals a need to economize. To save on the fixed expense of insurance, drivers with low credit cut down on owned cars more than they cut down on the miles they drive. More annual miles per car for a low credit group means more claims per 100 car years.

More UMV claims--accompanying more liability and physical damage claims--conflicts with the greater driver negligence explanation.

This Demand Law explanation is more fully developed and applied to other claim-frequency predictors on NOW's automobile insurance website below.

Patrick Butler, Ph.D.
Insurance Project Director,
National Organization for Women, Washington, DC.
www.centspermilenow.org

David asks a valid question: "...why have so many insurers switched to using [credit scores]?" I think I know the answer.

First, because it comes from an easily available existing database and requires little, if any, personnel intervention to obtain--making the underwriters feel "warm and fuzzy" because they checked an objective source of consumer information.

Second, because it adds a million dollars a year to the bottom line of Fair Isaac and three major national repositories who sell the information, and, as stakeholders, they will promote the science like the Academy Awards (even if it may have some flaws) for the sake of profits.

As to the second part of this thesis--"Why have their loss ratios improved?"--again, a valid question.

If, in fact, there is evidence of such improvement at one carrier, there is no assurance that a different carrier would enjoy the same improvement. In my view, it may simply be the law of "unintended consequences" at play. They may have added credit scoring, but they may have also taken some other causative measure that contributed to the improvement of which we are unaware.

In sum, I think underwriting insurance and financial transactions requires human involvement. I am not comfortable allowing a machine to interpret ostensibly empirical data into any meaningful score that purports to accurately predict my future behavior. Garbage in, garbage out.

There will never be a substitute for knowing your customer, and human underwriting by experienced underwriters will always be the most reliable method for assessing risk.

Michael Gold, Director
Underwriters Bureau, Inc.
http://www.underwritersbureau.com

One thing to note is that we are not just talking about credit scores here. As I explain in this post on credit based insurance scores at http://www.edmblog.com/weblog/2005/09/data_mining_in_.html we are also talking about using credit bureau data to predict insurance risk.

This is different from using a credit score, as it uses the data collected by credit bureaus to predcit not credit risk, but claims risk.

My expert colleagues tell me that credit data only accounts for about 15 percent of risk in the models. There's lots more on this topic at http://www.insurancescore.com

In November, the voters of Oregon will decide on an initiative to ban insurance companies from using credit scoring to determine insurance rates. Oregon law already says companies cannot use credit scoring to raise rates or drop existing customers. Credit information can only be used when a consumer first applies for insurance.

The PIA Oregon/Idaho supports the continued use of credit histories and scores by insurers.

Risk classification is an elemental priority for insurance because it is rooted in a concern for fairness: one classification should not subsidize, nor be subsidized by, any other risk classification. Where the risks of loss are different, the public expects, and deserves to see, differential treatment in costs and availability of insurance.

Companies must be able to accurately evaluate and classify the risks presented. Failure to do the job right could mean the unnecessary loss of business or the loss of millions on the financial statement.

The insurance-buying public benefits from competition in underwriting just as it benefits from competition based on price and service. Those with poor credit histories and scores present significantly higher risks of loss than those with favorable credit histories and scores.

Eliminating the use of credit information would be unfair to the great majority of insurance customers who carefully manage their business and personal finances.

Responsible individuals and well-run businesses end up subsidizing the insurance costs of those that aren’t as careful. Without credit scoring, rates are higher for everyone.

Removing tools that have proven to be accurate will only result in unfair cross-subsidies between risk groups.

Sam, consider the following two papers on credit scoring.

The first--one of the first public analyses of credit data--shows pretty clearly how effective is the technique for auto and homeowners. It also shows exactly what kind of data is used, and what is not.

http://www.casact.org/pubs/forum/00wforum/00wf079.pdf

--The second, perhaps more importantly, is the result of the Texas credit scoring study that concluded, despite considerable a priori bias against, that there is no detectable unfair discrimination by race.

So it's accurate and not unfair--like I said in an earlier post--emotion versus reality.

http://www.tdi.state.tx.us/reports/pdf/credit05sup.pdf

None of this will convince those hardened against the position, but more information is better than less.

Todd R. Bault, FCAS
Sanford C. Bernstein, LLC