CS-18 - How to Properly Vet Natural Catastrophe Models

Catastrophe models are widely used in the insurance industry. Initially, they were primarily used on the portfolio level but increasingly they are being used for pricing individual risks. The kinds of standard validation that are suitable for aggregate level analysis are not sufficient when using them for this purpose. A key question for both pricing actuaries and regulators is how to vet such emerging models and what are the best practices on how to use them.

In this session we will look at practical techniques to evaluate model output for reasonability with an emphasis on location level output. Techniques will be described both for instances when the user has access to a single model and when the user has access to multiple models. The approaches described will lean heavily on the use of geographic information systems and will describe how building models based on catastrophe model output can give us greater insight into how the models work. The approaches that will be described are capable of being implemented by an actuary at a company or department of insurance who does not have the resources to validate each step of the modeling process.