Trevor Maynard addresses the importance of a new open framework for modelling catastrophe risks
Catastrophe models have strengthened the insurance industry by stretching our imagination. These models came into growing use after Hurricane Andrew in 1992 led to several insurance company insolvencies. Now we realise that the largest hurricanes could cause losses well in excess of 100 billion dollars, and those offering catastrophe insurance and reinsurance must hold capital that matches the level of risk. We are better off with these models than without them, but while they have grown in sophistication, insurers have increasingly wanted more transparency.
Interface with science
Catastrophe models build up their loss estimates by modelling the hazards explicitly. They then combine this with engineering knowledge of how buildings will behave as the earth shakes or the winds blow. Good data is essential, as usual, as the property's age, construction and use all combine to materially affect the level of damage. These models provide a perfect interface between science and business.
Companies must own their own assumptions of such parameters when they underwrite catastrophe risks. Arguably this was always the case, but Solvency II has formalised it. Solvency II requires model users to validate why the model they have chosen is appropriate for their company. It is no longer acceptable to pass the responsibility for parameter choices and model methods to a third party. This has led to some challenges because traditionally modelling firms held some of the details of their models close to their chests. This was quite understandable, as the detail was their intellectual property, after all.
The modelling companies responded to Solvency II well by publishing the impact of their key assumptions, but insurers are increasingly making their own adjustments to models and may start to need even more detail. Lloyd's, as one of 22 partners, has been funding a project with the working title of Oasis to create an open source catastrophe modelling framework. Following two years of development, the system is now available to members for testing to explore functionality and request further development. Eventually the Oasis framework will be fully open and available to all. Membership is open to any company for a small membership fee.
The framework will allow model developers anywhere to create a model in a format ready for insurers to use. This is analogous to the proliferation of apps that Apple stimulated by providing a language to the developer community on the iPhone.
Academics studying natural hazards will be able to convert their insights into model updates whose impact can be tested as soon as they are ready. As a result, many alternative views of risk can be available to the insurance industry and will help to communicate the levelof model uncertainty in the results.
We hope the framework will stimulate a large catastrophe modelling community including both users and model developers. We could see an explosion of models for emerging markets, which will be timely. Just over a year ago, Lloyd's published research which argues that there are many regions in the world where the current penetration of insurance falls too far short of the economic impact of disasters. Models and a good knowledge of the risks in the region are a necessary first step to underwriters considering whether to provide significant protection against these exposures.
The actuarial profession may wish to get more actively involved going forward. Once Oasis is fully open it would enable a GIRO working party to explore these models and perhaps enhance them. The next few years will be very exciting in the cat modelling space.
Trevor Maynard is head of exposure management and reinsurance at Lloyd's with responsibility for catastrophe aggregation monitoring and the natural hazards element