Catastrophe models are good for the insurance business
John Dizard analyses the art of pricing risk in the wake of Hurricane Harvey and Irma
Hurricanes Harvey and Irma are now being described as a moment of vindication for climate change models. Well, up to a point. While insurance people believe that climate change is a real phenomenon, and at least partly caused by human action, they do not use climate models as their principal method of pricing storm risk.
Rather, for the most widely insured and then re-insured “perils”, they use catastrophe models, which incorporate modest but steady increases in assumed losses due to climate change. Climate models do not generate results that can be used directly to price risks, because they have problems modelling extreme weather events and are too expensive to run on the required scales of time and space. The most commonly accepted climate models divide the world into cells that vary from about a kilometre to hundreds of kilometres. Such constructs are not currently able to model the observed number of hurricanes, nor their strength or where they make landfall.
Policies that cover windstorms are generally priced annually, while climate change happens over much longer time periods. Catastrophe models are mostly built on the observed weather record of recent decades. Those models also incorporate long-term climate trends, economic growth and coastal construction, windstorm hydrodynamics and the effects of storms’ passage over land. They have a spatial resolution of 10 meters to 10km, and model year-to-year loss probabilities, or even losses just after storms take place.
Catastrophe modellers are compulsively careful with their projections, and are careful to note ranges of error and uncertainty. However, there can be certainty that models are good for the insurance business.
As one Bermuda catastrophe reinsurance executive says, “Assume this is a $100bn loss year for the industry. The forward-looking industry models that take extreme events into account say that will occur every six years. The history of hurricanes for the past 100 years says this happens every 20 years. Since 1900 there have been only 33 hurricanes that have cost more than $10bn in damage.”
Other insurers might use different timescales or historical loss estimates. Perhaps the most challenging problem for insurers is to keep rates up during long periods of low storm or earthquake losses, despite competitive pressures, so as to have the company and industry resources to deal with the occasional mega-disaster.
In the 12 years since Hurricane Katrina, the industry has built substantial reserves and capital. According to a London reinsurer, “The consensus was that the industry was overcapitalised by $50bn-$60bn before Harvey’s landfall. If there really is an insured loss of $30bn for Harvey and $20bn-$25bn for Irma, and if we assume on top of that the insurance-linked securities business reloads [capital] with a few billion, that excess capital is gone.”
You could also consider those tens of billions as a safety margin. Robert Muir-Wood, the chief research officer of RMS, a prominent catastrophe-modelling consultancy, asserts that “before the models became available after Hurricane Andrew [1992], insurance companies would go bankrupt. Actually, insurers have more or less stopped going out of business after catastrophes.”
The increased reliance on commonly accepted catastrophe models made it possible to tap outside capital, including insurance-linked securities. You could call that a benefit of greater accuracy or the cause of chronic underpricing.
The cat models (not climate models, remember) can be useful just after a catastrophe to estimate ultimate losses before the next pricing season. That is how, for example, that the size of justifiable claims made on Harvey and Irma losses can be more accurately estimated before many reinsurance policies are renewed in January. Also, the US hurricane season is not over until October, so depleted reserves and capital can be drained by more storms, never mind earthquakes.
The pleasurable frisson of being a reinsurer comes from a season that is just loss-y enough to increase rates dramatically without the actual pain of your own company losing capital. Sort of like going to a theme park ride rather than a war.
At the moment the reinsurers believe they are close to getting that outcome. Many were at the annual “Rendezvous de Monte Carlo” insurance convention that took place in Monaco over the Hurricane Irma weekend. “On Friday,” one of them recounts, “we expected a $150bn event. By Wednesday it seemed more like $24bn.”
That would be a thrilling but not terrifying rollercoaster.
By the end of November the reinsurers should have an idea how much they can increase rates for next year. They are pencilling in increases of up to 30 per cent for reinsurance on Florida wind and storm surge policies, though not much, if any, rises for lines of business unaffected by the storms.
The surprise could come in the scale of auto insurance losses in Houston. As one underwriter points out, “Unlike Florida before Irma, nobody left Houston before Harvey. There could be auto insurance companies going out of business thanks to Harvey.”
Whatever the covered losses turn out to be, the world can expect significantly higher premiums, and more attention to catastrophe modelling, as well as climate change headlines.