Any Way The Wind Blows – science really matters….
Just over ten years ago, at the end of 2005, the insurance industry was facing one of its biggest challenges in modern history. That year had seen the most active Atlantic hurricane season in recorded history, with 28 named storms having formed in the Atlantic basin – an astounding eight more than the next most active season in the record. The impact of the season was nothing short of devastating, with almost 4,000 deaths and record damage of about $159 billion. Even more concerning was that these events had immediately followed a highly active 2004 season that, with total losses of over $57billion, was the costliest season ever on record at that point. Both the 2004 and 2005 seasons had seen a record-equalling six U.S. hurricane landfalls each.
If deriving a simplistic view of return periods of U.S. hurricane landfalls from the long-term Hurricane Dataset (HURDAT), at the end of 2003 some insurers and reinsurers would have had the chances of six U.S. hurricane landfalls in a single season pegged at approximately a 1 in 137 year event. If assuming activity between years was independent – an assumption that, at least on the surface, doesn’t seem unreasonable given the historical statistics – the probability of seeing the next two seasons having six landfalling storms back to back would have a probability pushing 1 in 20,000. Thus, what followed the events of 2004 and 2005 was a great deal of soul searching by modelling agencies, underwriters and ratings agencies.
Ultimately, there are serious discussions to be had about the robustness of science underlying any view of North Atlantic Hurricane Risk.
Many in the insurance industry, along with climate scientists, looked to causal mechanisms that could explain how such a situation could have occurred, with theories citing everything from a “perfect storm” of climate conditions through the phase alignment of multiple internal atmospheric oscillations, to forcing from the impacts of climate change and warming oceans. Fearing that the activity was unprecedented, and that we could be entering a period of time in which history was not a good predictor of future risk, some within the insurance industry began to look to alternative methods to provide a more nuanced view of Atlantic hurricane risk, with an aim to create more reliable predictions of near-term activity.
Tools vs Truths
At XL Catlin, however, we have been questioning the robustness of the science behind this reaction.
Unusually, XL Catlin has its own in-house science team; we, together with the business, independently look at not only the significance of vendor model changes, but all models and predictions that come out every season.
We do this because, as a reinsurer well-known for underwriting North Atlantic catastrophe risk, it is important to us that the latest thinking on this peril is embedded within our internal view of risk.
Like all major insurance and reinsurance firms, we subscribe to catastrophe models provided by commercial vendors to provide underwriters with up-to-date data on natural catastrophe risk.
We have invested time and money in building a science team that can critique current methods used by mainstream catastrophe modelling vendors, and have cultivated our ability to utilise catastrophe models as far as it is appropriate – to use them as tools rather than truths – to derive our own view of the risk. This has led to us taking a unique approach to how we model and ultimately underwrite this class of catastrophe risk.
Precision vs Accuracy
Ultimately, there are serious discussions to be had about the robustness of science underlying any view of North Atlantic Hurricane Risk. Unfortunately, given the paucity and uncertainty of historical data, coupled with a lack of understanding of multi-annual effects on the climate-hurricane system, many views of the future risk are impossible to invalidate. However, a crucial job for the science team here is in communicating to our underwriting teams exactly what level of science we can trust to be robust, and what conversely is an interesting but highly uncertain scientific question. These ideas help us to define how much science should be used in decision-making, whether the right questions are being asked, and whether we have an appropriate balance between precision and accuracy in our view of risk.
The net effect of our decision to overlay our own independent view of risk on top of the latest modelling data is to give our underwriters a superior analysis of the predicted risk of hurricanes in the North Atlantic. Our underwriters are able to take this sophisticated information to make sensible underwriting decisions that make sense for XL Catlin and for our clients.
Our scientific philosophy is nicely summarised by the mathematician John W Tukey – “Far better an approximate answer to the right question, which is often vague, than an exact answer to the wrong question, which can always be made precise.”