Imelda Powers, Global Chief Catastrophe Modeler
Since commercial catastrophe (CAT) models were first introduced in the 1980s, they have evolved as new scientific discoveries and claims insights emerged. Despite the sophisticated nature of each new generation of CAT models, occasionally a model misses a significant loss driver for a particular peril. This occurs when a previously hidden attribute reveals itself through unprecedented intensity. Lessons from such surprises stimulate model improvements as our understanding of the physics of the peril and its damage potential, increase. Through this process, models mature over time.
Some examples of well-known “failures” include the following:
Building a Resilient Internal CAT Model
A company’s internal model often comprises catastrophe loss estimates from a mix of standard formulas, claims experience and commercial models. Most insurers rely on one core commercial model because of cost and resource considerations, but may obtain other modeled estimates to provide second or third opinions. Beyond a model’s scientific merits, there are several preliminary considerations when choosing which core model to use:
(1) Setup cost – IT infrastructure and expertise to support the application.
(2) Security/data protection – whether to host the applications in-house or with third party providers.
(3) Ease of use – intuitive interface, availability of support and proper documentation.
(4) Resilience of vendor – whether the vendor is capable of continuing service in crisis situations.
These considerations are important and can help a company decide between two reasonably rated models.