Please note: After several successful years, the Open Minds blog will be closing. For further details, please visit our FAQ

Currently showing: Climate/natural disasters > Disaster risk

07 Sep 16 08:15

If you're an insurance company risk manager, it's important to know how hard you can be hit by a natural catastrophe like a European winter storm. But it's equally or even more important to understand how often you can be hit. It will determine how much capital you need to hold or, alternatively, how much and which kind of reinsurance you decide to buy. While companies strive to do this out of self-interest for a long time, the Solvency II requirements have added an increased regulatory interest in the topic in recent times.

Swiss Re remains the only reinsurer that understands its cat risk assessment inside-out, as all our models are built in-house by our Cat Perils team. This requires my colleagues and me to stay on top of scientific research, and occasionally to investigate beyond, in order to get results with a high relevance in respect of the problems of the insurance industry.

Last year, we released a publication entitled "Winterstorms in Europe: Messages from forgotten catastrophes." It combined two key elements: Firstly, we investigated generally known, but (previously) poorly understood major winterstorms of the late 19th century and established a best guess of the corresponding wind intensities across Europe. Secondly, we made use of Swiss Re's capability of running such catastrophic events through its own model for an estimate of insurance losses, if these events were to happen today. One of the three events investigated in detail would be a true test to the UK insurance industry. Insured UK property market losses of around 10bn GBP would considerably impact insurer's balance sheets. Furthermore, handling an estimated 3-4 million claims would severely strain company resources from an administrative point of view.

We are just about to release a follow-up publication looking specifically at the topic of seasonal clustering of winter storm events. How probable are years of 2, 5 or even 10 events in a row? And given one strong event, is there an increased likelihood of more severe events to follow? Without spilling all the beans prior to the release of this publication here some key facts and findings: The new Swiss Re study used publicly available climate model results (CMIP5 initiative) from eight different science institutions around the world. These correspond to 7000 years of simulated European winter storm activity. The results underpin and further quantify earlier studies: winter storm events do come in clusters, and the UK lies at a latitude where clustering is particularly evident. Maybe even more importantly, the results show increased clustering for stronger events. Or put in other words: if you're hit hard once, there's an increased chance to be hit hard once more!

The findings are obvious: the temporal aspect of European winter storms must not be neglected as it aggravates the risk of catastrophic loss years. Risk managers clearly need to define their risk tolerance for the annual loss burden in light of possible event clusters. Together with my team I'm looking forward to discuss more details with our clients and help them shape their specific view of risk related to European winter storm clustering.

Category: Climate/natural disasters: Disaster risk, Floods/storms

Location: Europe

1 Comment

Urs Leimbacher - 17 Oct 2016, 1:03 p.m.

Thanks for sharing these insights, Peter. Frankly, a bit of a frightener with the winter approaching - and mindful of the loss scenarios from European winter storms that Swiss Re has published in the 2015 "lessons from the past" publication you mention.

On the positive side, I'm always amazed about the ever more granular analysis that our models enable. Kudos to the colleagues who work on our models. Their findings help our clients to better assess their own portfolio's risk exposure and to act accordingly - being smarter together!

If you would like to leave a comment, please, log in.