Model selection with Gini indices under auto-calibration

Abstract

In general, the Gini index does not give a consistent scoring rule. Therefore, maximizing the Gini index may lead to a wrong decision. The main issue is that the Gini index is a rank-based score that is not calibration-sensitive. We show that the Gini index allows for consistent scoring if we restrict it to the class of auto-calibrated regression models.

A Discussion of Discrimination and Fairness in Insurance Pricing

“…• Adverse selection and unwanted economic consequences of non-discriminatory pricing should be explored. Non-discriminatory prices typically fail to fulfill the auto-calibration property which is crucial for having homogeneous risk classes, see Wüthrich [19].…”