Examining 2.7 million auto records across the country, the research found consumers' credit-based insurance scores are "unquestionably correlated" to their propensity for auto insurance loss, the study authors wrote.

Their findings were unveiled at the NAIC meeting by the four major insurance trade associations: Downers Grove, Ill.-based Alliance of American Insurers; the American Insurance Association in Washington, D.C.; the National Association of Independent Insurers in Des Plaines, Ill.; and the Indianapolis, Ind.-based National Association of Mutual Insurance Companies.

"This study was just completed in the last couple of days and it's now publicly available. It looked at 2.7 million records from across all different states, which makes it the largest publicly available actuarial study that's ever been done on insurance," said David Snyder, vice president and assistant general counsel at AIA and one of the presenters of the study at NAIC's industry liaison committee this past weekend.

The study, now publicly available at www.epicactuaries.com/publications.htm, did not satisfy some consumer advocates present at the NAIC conference, and its findings will add more fuel to the ongoing debate on credit scoring.

One of the main findings from the study is that insurance scores--which incorporate consumer credit scores--are highly correlated with the loss propensity. As a general rule, as the insurance score goes up, the loss propensity comes down for auto insurance coverage, such as liability, collision and comprehensive, it said.

The study also looked at whether credit scoring measures risks that are already considered by other rating factors. And while its findings suggest that there is some overlap, insurance scores offer "still more accurate risk assessment," honing the accuracy of traditional rating variables.

And credit scores were found to be one of the top three risk factors for every line of coverage studied: bodily injury liability, property damage liability, personal injury protection, medical payments, comprehensive and collision, Mr. Snyder said.

"It was an independent, sophisticated study which answers fundamental questions involving credit scoring, which are: "Does it really measure risks, is it a proxy for anything else, and how important is it compared to other rating factors?'" he told National Underwriter.

He also emphasized that the study used a "multivariate" analysis, which involves examining all risk factors simultaneously, looking at different factors against each other, to adjust for any interrelationship between insurance scores and other factors. After adjusting for such variables, consumers with lowest insurance scores were found to have 33-percent-higher losses compared to the average, while those with highest scores had 19-percent-lower losses than the average, he said.

For the study, record samples were taken from policies in effect between July 2000 and June 2001, according to Charles Schmidt, spokesperson for the Alliance of American Insurers.

The data included information about the policy itself, and each vehicle and driver insured on the policy, as well as each claim on the coverage, while ChoicePoint Inc., Alpharetta, Ga.-based underwriting and claims data provider, offered insurance scores.

"Going into this project, previous research suggested that there is a significant correlation between insurance scores and risk of loss. However, the results were more compelling than we anticipated," said Michael Miller, one of the report's authors.

"If ever there were doubts about the importance of insurance scores in the accurate underwriting and rating of automobile insurance, those doubts were erased by the findings in this report," he said.

Mr. Snyder also argued that this study validates the "balanced" legislation and regulations effort that most states have engaged in.

"The study says unmistakably that credit scoring is one of the most predictive factors for risk. This study absolutely supports the careful, deliberative public policy approach the vast majority of states have been engaging in," Mr. Snyder said.

The study shows that credit scoring does not merely pick up something else that's already being measured, and it's not a proxy for anything else, added Mr. Snyder.

"That would be a sufficient business justification. This study establishes firmly and beyond any doubt that there is a legitimate business purpose for credit scoring, namely accuracy and predictability in pricing the risk. There is no intentional discrimination," he said.

"Even if credit scoring affects different income and racial groups differently--and I am still not taking for granted for a second that it does--it still furthers a legitimate business purpose," Mr. Snyder said, "and that is acceptable."