A multivariate energy-based fairness adjuster for premiums

Fairness in insurance has received increasing attention, particularly in light of regulations calling for non-discriminatory premium estimation, such as the EU Gender Directive (2012). This research explores group fairness mechanisms designed to mitigate group-level disparities. We propose a method based on the energy distance, a multivariate metric that enables fairness adjustments across multiple sensitive attributes simultaneously, even where they are non-binary. To maintain overall prediction balance, our approach integrates an autocalibration step to correct for biases in the total predicted number of claims. Moreover, our method supports fairness adjustments for new policyholders without requiring retraining of the fairness model. We evaluate the proposed methodology in the context of car insurance pricing, where demographic factors such as age and gender are commonly used for risk assessment and premium determination. Results show that our method effectively reduces group-level disparities while preserving the predictive accuracy.