Understanding the uncertainties is key to the effective use of catastrophe modelling


The widespread response upon hearing of the upcoming changes to North America earthquake modelling is “please, not another disruptive model change”.
How have we found ourselves in a place where a significant advancement in our knowledge is perceived so negatively? It stems, I believe, from a misunderstanding of how modelling works, specifically how it treats uncertainty – and a failing of the exposure management community to effectively communicate how those uncertainties impact their work.

Our historical record and knowledge of all elements that go into making up a financial loss from a catastrophe is limited. Modelling purely attempts to represent what we know, and quantify what we don’t and the result is a glimpse into the realm of possibility. Updates merely improve and refine the view, but they rarely change the landscape.

One of the responses to hearing of the change was, “the nature of earthquakes hasn’t changed, so why does the modelling need to?”. As a discipline, exposure management needs to get much better at clearly communicating what modelling is trying to do, and how and where the models need to be combined with other risk management techniques to deliver a complete picture.

The catastrophe model validation process has already established the need for syndicates to form their own view of risk behind each insured peril, and at Asta we support our syndicates through this, both with guidance and a robust governance framework. This process should be used seriously by carriers to assist them in forming a clearer opinion on the nature of the perils they insure and the role that models play in their assessment.

High frequency events, with strong, detailed claims histories are the place to start when building an internal view. In this space, models are the facilitators of a rapid and robust technical rating process; which given the significant amount of existing data already in daily use is unlikely to deliver a risk rating that is much altered by an upgrade to the model.

It is for rare, extreme events, or those with limited historical record, where models act as sophisticated crystal balls. Here the input of externally sourced opinion, via a commercially available model can be essential.

However, relying solely on external opinion is not enough as risks of this nature often drive our biggest underwriting decisions, the level of capital we need to hold and the risk of ruin we face. So while we can and should look to the models for help and guidance it remains our responsibility to appropriately use that information to inform the specifics of our own business.

It is beholden on us to specifically understand the material uncertainties in our assessments, however precise the model results may be. At best we may then be able to manage them – ensuring we are aware of the entire risk landscape and not just the view in our lenses – perhaps even mitigate them; but at the very least we need to communicate them, in the right way, at the right time.

Only once we’ve established how to handle the conflict between the model’s answer and the reality of our own portfolio, will scientific advancements, and any relevant changes to the models, be seen in a more positive, progressive light.