The 2017 Atlantic hurricane season, which caused more than US$300 billion in damages, will go down as the most expensive ever, and the most active since 2010. Insurers’ results, which show catastrophe losses in the billions, may shock outsiders. However, for those on the inside, losses were high but not unexpected. Insurers take on risk every day. Never paying out for a loss means insurance isn’t serving its purpose. What’s important is our ability to assess risk through data analysis, modelling, actuarial and underwriting expertise, so that when claims of this magnitude do come through, we are prepared. For the most part this season, insurers were prepared: exposure management did its job.
Nonetheless, there are still lessons to be learned from H.I.M. so that we can continue to grow and improve.
Improper use of catastrophe models.
Risk modelling has advanced to the point that models exist for most material perils in an insurer’s book, but like the perils that they represent, they each has their own traits that must be understood. Some models, such as the new marine model developed by Lloyd’s, maintain a strong link to the ‘real’ loss scenarios on which they are based. Following a major event this allows direct questions to be asked and actionable conclusions drawn.
More typical probabilistic catastrophe models are not designed this way. The starting events on which analysis is based are themselves generated, using Monte Carlo simulation across an array of connected potential distributions. The goal is to produce enough “points” on a curve that the essence of the hazard is adequately represented, but the generated scenarios themselves are just a means to an end, and contain limited value in their own right. When a real-world event occurs, heavy caution is required in using these models in a way that they were not designed for. While they may still have a role to play, it is still largely up to the underwriter to call upon their expertise, experience and knowledge to pick the loss estimate. When Hurricane Katrina hit, many made huge errors by placing too much emphasis on one or two loss scenarios coming directly from a model. This season it’s been heartening to see that lessons have largely been learned, and there has been greater understanding in the market that catastrophe models should be used as a guide rather than a crystal ball; but it’s still something to be careful of.
Don’t lose sight of smaller losses.
While most of the big property losses were within modelled ranges, albeit at the high end, some insurers were caught off guard by large accumulated losses from smaller risks. As is usually the case, the greatest attention is placed on the highest value areas, which can lead to some complacency for smaller product lines. In the case of H.I.M., the Lloyd’s yacht market, which was already in a fragile state due to years of rate reductions, was hit with outsized losses. Hurricane Irma proved especially devastating, particularly as the destruction of commercial yachts resulted in a double hull and business interruption claim. One of the challenges of exposure management is looking at a big book of business and trying to figure out how all the classes interact and where exposure accumulations lie. While certain lines may be farther down the priority list, complacency can quickly come back to haunt you.
Improved communication between insurers and regulators.
Insurers and regulators always find themselves in the same tug of war. Regulators demand to know immediately after, if not before a hurricane hits, how much loss will be experienced. Insurers are reluctant to provide precise figures with so much uncertainty present. Even with sophisticated models, estimating losses to actual events takes months, not hours; and with often complex positions on contract wordings and binary trigger reinsurance, small movements can result in large movements either way for insurers.
Following H.I.M. both sides should consider how they can work together more cohesively. This could be for example, an agreed approach to providing numbers early on, but with greater understanding of the potential for uncertainty and a guarantee that firms won’t be penalised for getting initial estimates wrong. Or it could be working with regulators to increase their understanding and willingness to wait longer to ask in depth questions. Both sides have valid concerns. It’s a question of finding a solution that brings the two together.
The biggest lesson for exposure management professionals from H.I.M is that we can never become complacent. Extreme natural catastrophes are very rare, typically following many hundred-year cycles. 2017 gave us a challenge that we stood up to, but that doesn’t mean that there aren’t big surprises waiting just around the corner. Exposure management, and the discipline of catastrophe risk modelling, have come a long way since I entered the insurance industry in the early 2000s, but there is still an even further way to go.
Alan Godfrey started his career at Amlin in 2004 after studying mathematics at the University of Cambridge. In 2006 he set up and led the company’s catastrophe modelling team, which by the time he left had grown to 40 full-time employees covering Reinsurance, Property and Marine classes, based across multiple international locations.
Through this role Alan gained extensive knowledge of the uses, strengths and weaknesses of the main catastrophe models, as well as the developing best practice in Exposure Management. With particular focus on the operational efficiency and effective use of capital, he provided support to Amlin in achieving one of the first Solvency II approved Internal Models. Alan joined Asta in 2015 as Head of Exposure Management.