On August 4, 2020, the Big Data (EX) Working Group (the “Big Data WG”) of the National Association of Insurance Commissioners (“NAIC”) met by conference call as part of the NAIC Summer 2020 National Meeting. The conference call focused on updates from the Casualty Actuarial and Statistical (C) Task Force (“CASTF”) and the Accelerated Underwriting Working Group on the progress each has made on their respective workstreams.
As we previously reported, CASTF is working on a white paper on best practices for regulatory review of predictive models and analytics by insurers. The development of the white paper is driven by a need to promote comprehensive coordinated review of predictive models across the states. During the Big Data WG conference call, a presentation was given by the Chief Actuary of the Louisiana Insurance Department regarding the draft white paper which addresses (1) proposed changes to Product Filing Review Handbook (the “Handbook”) to include best practices for review of predictive models and analytics filed by insurers to justify rates, and (2) proposed state guidance for rate filings based on complex predictive models.
The white paper is intended to encapsulate long standing review practices that many regulators have been engaging in for years or even decades and is not intended to impose any new requirements. The Big Data WG had previously identified certain principles that CASTF would adhere to in drafting the white paper – specifically that there be no change to regulatory authority and autonomy of the states, that states share information and expertise and discuss technical issues, and that the states maintain confidentiality of information in accordance with state laws.
In drafting the white paper and with these principles in mind, CASTF identified four best practices for regulatory review of predictive models. First, regulators should ensure compliance with state rating laws – i.e., rates shall not be excessive, inadequate, or unfairly discriminatory. Second, regulators should review all aspects of a predictive model including the underlying data, assumptions, adjustments, variables, input and resulting output. Third, regulators should evaluate how the model interacts with and improves the rating plan (e.g., obtain a clear understanding of how model output interacts with non-modeled characteristics/variables used to calculate a risk's premium). Fourth, predictive models should enable competition and innovation.
The draft white paper includes (i) proposed changes to the Handbook to include the above-mentioned best practices, (ii) proposed state guidance, including 79 information elements that a regulator should know to meet the objectives of the best practices when reviewing a generalized linear model (GLM), (iii) a glossary of terms, and (iv) a sample rate-disruption template. Proposed changes to the Handbook include, for example, addressing certain common drafting issues such as the use of the phrase “rational explanation” and other correlation terminology used in rate model filings for which the paper now requests regulators provide additional explanation. The draft white paper also emphasizes that information provided by regulators about rating models should be kept confidential in accordance with applicable state laws, and it also addresses the applicability of the best practices to other lines of business and non-GLM based models.
The Big Data WG is now considering comments on the draft white paper which have been submitted by interested parties, and will seek to issue a final draft of the white paper by the end of September 2020.
The Big Data WG also heard an update from the Accelerated Underwriting Working Group regarding the progress it has made with regard to its charge to consider the use of external data and data analytics in accelerated life underwriting, which is discussed here.
Originally published by Clyde & Co, August 2020
The content of this article is intended to provide a general guide to the subject matter. Specialist advice should be sought about your specific circumstances.