By Ian Scott, senior risk modeller at Capsicum Re
This was the topic of discussion at a panel I attended back in March. Our moderator was Hjörtur Thráinsson, a well-respected modelling expert at Munich Re, and he proposed that there is no such thing as mature CAT models due to models being changed regularly, wide ranges of different loss predictions between supposedly similar models and their relative short time within the (re)insurance market place but I disagree.
CAT models are essential tools within our industry, enabling us to intelligently select and analyse risk, pricing, capital adequacy and profitability. The amount of research and science behind these models is vast but therein lies a key differentiator between mature and immature models.
There will always be unknowns and a lot of assumptions have to be made; CAT models don’t accurately predict disasters occurring. We only have to go back to 2011 with the Tōhoku earthquake in Japan, which measured nine on the Richter scale, and was followed by a devastating tsunami for an example of a failed catastrophe model. This was the fourth largest earthquake on record (since 1900) yet this combination of events was deemed highly improbable by scientists. Similarly the Northridge earthquake in California in 1994 occurred on a fault line that was previously unknown, demonstrating that we can never guarantee that an event won’t happen or know exactly when and where an event will occur. This is why it is so important to constantly re-evaluate assumptions made so that we can develop these models.
So in light of this, can there ever be a mature catastrophe model?
The answer’s yes. A mature CAT model is one that has been peer reviewed throughout the industry, that has been subject to scrutiny and sensitive analysis and one where every part of the catastrophe model has been investigated. Based on this definition, the two most mature models are US Atlantic Hurricane and US Earthquake and because they are the most frequently used models, they are under constant analysis and review.
However we can be and should be doing more to encourage the sophistication and evolution of these catastrophe models and this is where we can learn from the scientific community.
Whenever a research paper is published it is peer reviewed and this facilitates a culture of openness. The implementation of Solvency II has encouraged the insurance industry to follow suit as it requires everyone involved in the CAT modelling process to have a clear understanding of the assumptions used to build the CAT models and their inherent uncertainties. There is also an initiative, supported by Lloyd’s, called Oasis LMF, whose mission is to create a community that will increase the supply of catastrophe modelling information; encourage openness and transparency in model building and create an open community which will stimulate innovation within the insurance industry.
It might feel counter intuitive to share claims experience and data as it’s our proprietary knowledge but it is necessary if we are to create more robust, open CAT models and it serves to benefit our community and clients in the long term.
CAT models are vital in our line of business and it is in our best interests to ensure the predictions produced by the models are as accurate as they can be. We need to constantly assess, challenge and update the models as it is through this collaboration, that we will improve validation techniques and the accuracy of these models.