NAII Says Alaska Study on Use of Credit Scoring is Flawed

A study commissioned by the Alaska Insurance Division on the use of credit-based insurance scoring did not find a connection between
insurance scores and income or ethnicity. Moreover, the report is riddled with shortcomings and inaccuracies, according to the National Association of Independent Insurers (NAII).

“The Alaska study provided no conclusive evidence that insurance scores are anything less than objective, accurate underwriting and rating devices benefiting most policyholders with lower premiums,” said Diana Lee, senior research consultant for the NAII.

The 41-page study was conducted at the request of lawmakers and the former insurance director to review how the use of credit-based insurance scores impacts auto and homeowner consumers in Alaska. A broad, on-going review of the insurance scoring practice, including market conduct examinations, is expected by the insurance division.

“While the report is obviously flawed, it is important for public policymakers to note that the study itself cautions that its data is ‘not adequate’ and its findings may be ‘premature,'” Lee said. “Through its own admission, the study concludes that the methodology used was insufficient to show a correlation between insurance scores and income and ethnicity.”

“NAII is concerned that legislators and regulators in Alaska and other
states will rely on the inconclusive results in this report to draw
incorrect and exaggerated conclusions and form public policy
recommendations,” Lee added. “We implore public policymakers to keep in mind that the Alaska report- like so many others before it- found no basis for any conclusion that credit history is skewed toward ethnic minorities or low-income individuals. Our analysis of the survey methodology and results reveals no credible evidence that the use of insurance scoring violates any Alaska law. Conversely, insurers have demonstrated that the use of credit history allows them to reward customers less likely to incur losses with lower premiums.”

According to NAII analysis, the division’s methodology for conducting the study was seriously flawed. Census data by zip code was used instead of individual policyholders’ race and income. Even more disturbing, the study did not consider any information on credit history or insurance scoring. Instead, the categories of “preferred” (consumers that present the least risk to an insurer), “standard” (average risk level) and “nonstandard” automobile insurance risks (highest risk level) were used as proxies for credit history. Moreover, it was assumed that all preferred policyholders had good credit histories, all standard policyholders had average credit histories and all nonstandard policyholders had poor credit histories.

“The study made these assumptions without considering the influence of other factors that may have explained why policyholders were preferred, standard or nonstandard, such as accidents, age and traffic violations,” Lee said. “The study observes that changes in the number of policyholders in these markets may be caused by the insurers’ use of credit information, but then admits that ‘the data collected in the survey is not adequate to clearly determine the extent to which these changes are the result of the use of credit history.'”

In the report, the division itself acknowledges that further study is
needed, little basis exists to support claims such as “unequal effects
exist on consumers with varying income and ethnic characteristics,” and findings may be “premature to determine whether the policyholder
distribution between preferred, standard and nonstandard markets is due primarily to credit history or to other underwriting and rating factors.”

Other inaccuracies and misleading statements in the Alaska report include:
· The report claims that insurers were unable or unwilling to provide
adequate justification to support the use of credit history so that
insurance scoring models could be kept confidential as proprietary trade secrets. Insurers have expressed a willingness to allow insurance regulators access to their models, provided the regulators do not disclose the models to the public. Current Alaska law does not provide insurance companies with the protection that the details of their credit scoring models will not be shared with competitors.
· Rather than evaluating “individual” policyholder demographics such as age, race and income, the insurance division looked at various groupings within zip codes. These conclusions cannot be validated since broad generalities are made on a zip-code basis. To do this type of study correctly, individual information must be evaluated to determine whether a disparate impact on individual policyholders truly exists. The insurance division should have conducted a more thorough examination, using a multi-variate analysis to determine how much weight credit actually does have versus “other factors.”
· To determine a distribution for a particular zip code, the insurance division combined the data of “all” participating companies that write in that zip code – regardless of whether the company uses credit or not – to determine the percentage of policyholders designated as preferred, standard and non-standard class. The auto insurer respondents to the survey include 20 companies that write non-standard business, 24 companies that write standard business and 27 companies that write preferred business. The insurance division did not describe which companies are the preferred, standard and non-standard writers.

The report does show that, based on a comparison of the “preferred” market percentages found in the tables in the report – such as before and after credit – the industry can conclude that before credit history was used, there were fewer policyholders in the “preferred” category.

Topics Carriers Alaska