By explaining why there is an association between credit scores and insurance losses, insurers and actuaries can promote a better understanding of why credit scores are a useful underwriting tool, attendees of the Casualty Actuarial Society (CAS) Predictive Modeling Seminar heard.
Dr. Patrick Brockett, professor at The University of Texas at Austin, said one of the most important developments in the past two decades in personal lines insurance is the use of credit scoring as a classification and rating variable.
“This is maybe the most fundamental innovation in the classification area that has come up,” he said.
Yet despite the success of credit scoring as an underwriting tool and statistical evidence of the correlation between credit scores and insured losses, Dr. Brockett observed that the use of credit scoring remains under attack because it is not clear why there is an association between scores and losses.
The outcome of this debate is important for actuaries, he noted. This is because the acceptability of the Actuarial Standard of Practice on Risk Classification (ASOP No. 12) is being called into question by the new underwriting variable “credit score.”
“Actuarial Standard No. 12, which actuaries use to examine whether certain classification variables are acceptable or not acceptable, says that it is not necessary to prove causality. If you can prove there is a strong statistical relationship, that should be sufficient, provided it doesn’t violate certain protected relationships,” he explained.
But with the introduction of credit scoring as a variable, certain states for the first time have said this was not a reasonable criterion.
“The outcome of this debate ultimately could have a slippery slope argument for the use of gender. Or, why do married people have fewer accidents than single people? What is the causal relationship? This is an important issue that transcends credit-scoring,” he said.
Dr. Brockett went on to explain that like the insured loss, credit scoring is also a function of a set of variables.
“Insured loss is a function of auto-specific and driver-specific characteristics including a person’s personality and biological characteristics. Likewise, the credit score is a function of certain credit-specific attributes and the person’s personality, psychology, and biology.
“The correlation between the insured losses and credit scores is high and positive because they both draw on a similar vector of personality and psychological traits,” he explained.
Dr. Brockett noted that a review of the literature makes clear that there are intrinsic underlying individual biological and psychological characteristics of risk-taking in both financial behavior and driving. “The connector between insurance losses and credit scores is the psychological dimension.”
Basic chemical and psycho-behavioral characteristics, for example, a sensation-seeking personality type, are common to individuals exhibiting both higher insured auto losses and poorer credit scores.
This connection can be used to better understand why credit scoring works. “There is a commonality that leads us to a possible explanation of why credit scoring works,” he said.
Dr. Linda Golden, professor at The University of Texas at Austin, added: “What we’re talking about with risk-taking is a personality trait. Biochemistry influences personality.
“Our biochemistry may be the determinant of our personality, which then may have a strong influence on risk-taking impacting our credit scores, helping to explain in the bigger picture why credit scores predict,” she said.
Dr. Jesse Leary, Assistant Director for Consumer Protection, Bureau of Economics, U.S. Federal Trade Commission, gave an overview of the recently released FTC study on credit-based insurance scores and auto insurance.
Dr. Leary explained that one of the key findings of the study was that credit-based insurance scores are effective predictors of risk. For example, the study found that those in the lowest decile of credit scores were 1.7 to over 2 times riskier than those in the highest.
The study also found that scores are distributed differently across racial and ethnic groups. As a group, African-Americans and Hispanics tend to have lower scores than non-Hispanic whites and Asians. Therefore, the use of scores likely has an effect on the average insurance premiums that these groups pay.
“Using scores raises the average predicted risk of African Americans by 10 percent and Hispanics by 4.2 percent,” he said.
Dr. Leary noted that little of the relationship between credit scores and claims comes from the relationship between scores and race/ethnicity.
The FTC could not develop an alternative scoring model that would continue to predict risk effectively, yet decrease the differences in scores among racial and ethnic groups. “The results of these efforts indicate there is no readily available alternative scoring model that would achieve those results,” he said.
The session was moderated by Richard Smith, consultant, Towers Perrin.
Source: Casualty Actuarial Society
Was this article valuable?
Here are more articles you may enjoy.