“Insurance pricing is a high-wire act,” CAS says. Actuaries have to quantify and differentiate among a massive variety of risk variables while avoiding unfair discrimination. “As regulation and society’s understanding of discrimination evolve, however, it is necessary for us to keep abreast of changes in the manner in which discrimination is defined and adjudicated.”
The CAS research has generated four papers – two published this week, two more to be published on March 31 – that define, quantify, and propose methods for addressing unfair discrimination where it is found to exist.
Confusion around insurance rating is understandable, given the complex predictive models being used today, which can lead to inappropriate comparisons and inaccurate conclusions. Algorithms and machine learning hold great promise for helping to ensure equitable pricing. However, research has shown these tools also can amplify biases that manage to creep into their programming.
Recent Colorado legislation requires insurers to show that their use of external data and complex algorithms don’t discriminate against protected classes, as well as other state and federal efforts to address perceived bias in pricing.
The actuarial discipline and the insurance industry are well positioned to continue helping policymakers and corporate decisionmakers understand and address these inequities.
The CAS papers published this week are: