Credit-based scoring

Rough Notes, Aug 2002 by Hicks, Bruce

Correlation high, credibility low?

I thought I had paid close attention to the credit-based scoring issue, but now I'm convinced that I've missed some important point. The more I read about the issue, the more I read about correlation between credit scores and expected loss. The more I read about correlation, the more I expected to read about other justifications for using credit-based scoring. When I finally read from a knowledgeable insurance source that essentially "the correlation is the justification ..." well, I decided to stop reading and start writing because I'm confused.

Insurance people from the left, right and center have told anyone listening (peers, regulators, consumer advocates, legislators) that studies have shown a high correlation between credit histories and loss ratios. I understand that. However, I also understand why some people might refuse to make the leap of faith between the existence of that relationship and the use of credit scores as an insurance underwriting tool.

For one thing, it's counterintuitive. In other words, the statement doesn't "feel right." Answer the following questions as quickly as possible:

An insurance company states: "I'm sorry, we can't offer/renew your insurance because

Does It Feel Right? (Yes/No)

"You've had three speeding tickets in the last two years."

"Your MVR says that your license was revoked."

"Your application says that you are not the legal owner of the car."

"You have caused two serious accidents in the last three years."

"Your credit-based score is below our standard of xxx."

Did all of the statements "feel right" to you? If so, you should have answered each question "yes" quickly. If you answered "no," or if there was a discernible pause before answering "yes," then maybe you also feel that something's a little off. An important point is that if an insurance person can have such feelings, how do you think things are perceived by the general public as well as by the people with the charge to protect the public's interests?

The industry's continuous touting of the statistical correlation between certain elements in a person's credit history and their future profitability appears to fly in the face of credibility. It raises a number of questions.

What about cause and effect?

In my opinion, any element that is described as an underwriting tool should bear a direct relationship to determining the quality of an exposure to loss. Most folks outside the industry seem to be sold on the idea that the type of car, or the amount of driving experience or how much you drive are valid components of underwriting. How quickly a phone bill or charge card bill for Christmas presents is paid just doesn't have the same logical hook.

But maybe the hook exists. After all, underwriting criteria that are accepted today were questioned in the past. I happened to find an American Insurance Association ad that ran in the May 1980 issue of Rough Notes. It was titled "Why should two men with identical cars and similar driving records have different auto insurance rates?" It then explained how rates were affected by geography, specifically urban vs. suburban locations. The justification struck me as logical and easy to understand. I didn't get a feeling that it was "counterintuitive." One difference found in today's underwriting criteria fight over credit-based scoring is that no one is talking about causation, only correlation. Why can't we apply the same method to encourage acceptance of credit-based scoring?

How was correlation first recognized?

I've read comments from several sources mentioning that at least one authoritative study was made to prove a positive relationship between lack of claim activity and higher credit-- based scores. Well, who first came up with the idea to collect such information? Was the relationship first discovered and then applied by enough insurers, or were credit scores used first and the relationship discovered (hoped for) afterwards? Sharing more information about the events that prompted the use of credit scores in underwriting might help the idea gain acceptance from insureds and regulators.

How can companies be sure they are treating the information on the same basis?

Insurers consistently maintain that credit-based scoring improves pricing and underwriting. Is this position credible when companies do not have:

* a consistent method for integrating scores into automated underwriting programs?

* information on how their peers apply the information in underwriting decisions?

* objective information on how credit-based scores impact loss ratios?

This possible uncertainty undermines the better pricing argument. Different ways of handling credit-based information will create difficulties in making rating comparisons. Diminished uniformity also will lead to difficulty in the ability of agents, consumers and regulators to make meaningful comparisons.

What if credit-based scores involve serving self-interested parties?

Companies that supply credit-based scores claim that their method of developing the information is proprietary. Currently, at least one company not only sells the credit history to consumers, but also sells information on how to improve their credit history in order to maintain or improve their score. Selling the same core information to achieve two separate goals raises a legitimate question as to the bias and purpose of the scoring information. Creating scores that disqualify more consumers creates a new market and may influence the formulation used in the scoring.