Taking a shot at credit scoring. (Comment).

The underwriter's eternal quest for the silver bullet has always failed, and will always fail, to include the most recent silver bullet, credit scoring: Correlation without causation is statistically fool's gold. Is a credit score a causative factor of automobile accidents? ("By the Numbers," Best's Review, November 2002)

From the underwriter's viewpoint, a book of automobile business is nothing more nor less than a probability model subject to the universal laws of probability statistics: correlation, causation and access to the statistical universe.

Further, loss ratio is not an end unto itself, but a monetary reflection of the underwriter's conformity to the underlying probability statistics. The loss ratio, weak or strong, is a direct result of those statistics.

I am not privy to the statistical analysis of the purveyors of the credit-score system. But, if the purveyors of that system did not remove two groups--teen and DUI drivers--from the statistical analysis, there is, because of the claim severity of those two driver groups, a strong possibility that the conditional probability array and loss ratio are distorted.

Assuming only at-fault accidents the key to consistent underwriting profits is marginal probability of the occurrence of the auto accident (risk event), for the underwriter pays nothing if the accident does not occur.

Based upon the statistics provided by the Insurance Institute of Highway Safety, the marginal probability of an accident in the year 2000 was 3.35% or 33.5 accidents per thousand licensed drivers. However, one key statistic is missing: How many of these accidents are caused by drivers with repeat accidents, say, over a 10-year period, and how many accidents are caused by drivers who have been accident free over the preceding 10 years?

The key underwriting question: What features accurately reflect the causative differences between the high probability accident driver (3.35% of drivers) and the low probability driver (96.65% of drivers)? And, you can rest assured that a credit score is not a causative difference.

And, discovering those causative characteristics is the key to a low loss ratio and consistent underwriting profits. Can those causative differences be discovered? Yes.

And for those underwriters who say it is impossible to discover the causative difference, may I suggest that those underwriters re-read Napoleon: "Impossibility is only a word in the dictionary of fools."

James H. Garvin.

Gavin Insurance Agency

Cincinnati.