Graphical tools clarify insurance model accuracy and fairness
A Practical Guide on Graphical Model Validation
Machine Learning
Summary
Insurance companies use models to estimate how much people should pay for coverage, but it's important to check if these models are fair and accurate. The authors explain popular ways to check this using pictures and statistics that compare what the model predicts with what actually happens. They focus on making sure these checks are done in the right way, depending on how insurance policies are counted. This helps companies set fair prices and keep their models reliable.
What this means in practice
- •For actuarial modelers: Validate insurance pricing models with clear visual and statistical tools tailored for policy or exposure weighting.
- •For financial risk teams: Assess calibration and discrimination of risk models in insurance portfolios to improve premium accuracy and risk selection.
A survey. It maps existing work.
Authors
Mario V. Wüthrich
Abstract
This manuscript formalizes the most popular model validation tools used in general insurance actuarial modeling. These include graphical tools like calibration plots, actual-vs-expected plots, lift charts, Murphy diagrams, as well as classical statistical tools such as Bregman losses, deviance losses, elementary losses, Murphy's decomposition and Gini scores. Particular emphasis is placed on whether calibration and discrimination are studied under a policy-weighted or an exposure-weighted population measure. This distinction is crucial in ensuring that premium schemes are calibrated on the correct scale.