Existing risk models are not accurately capturing the growing risk from severe convective storms in the U.S. — including tornado, hail and thunderstorm — causing insurers and reinsurer to be blindsided by increasing losses, according to a study released Wednesday by Swiss Re.
“There are many models to asses the risks from “main” perils such as tropical cyclones and earthquakes, but there are few that model severe convective storms,” said the recent Sigma study released by Swiss Re. “The few severe convective storm risk assessment models that do exist have many shortcomings.”
According to the study global insured losses from natural catastrophes and man-made disasters were $35 billion in 2014, down from $44 billion in 2013. That is significantly below the $64 billion-average 10 year average.
However, Swiss Re points out that global insured losses from severe convective storms rose by an average annual rate of 9% in the period 1990 and 2014. The U.S has been especially prone to the storms, with insured losses from severe convective storms averaged $8 billion annually between 1990 and 2014. The past six years have been more prone to the peril, with U.S. convective storm losses exceeding $10 billion per year.
Rather than more storms hitting the U.S. Swiss Re cites increasing exposures tied to economic development, population expansion and urbanization as the reason for the loss development that existing catastrophe models fail to capture. The study added unless models develop, losses from “severe weather events may well continue on an upward path.”
Admitting that modelling severe thunderstorms is “inherently difficult” since the historical event archives are reliable or non-existent, modeling firms need to move ahead with research and testing.
“Today, computing capability has reached a level where generating these millions of years of events possible,” the study says. “But this is a recent development, which explains why risk assessment modeling for severe convective storms is still a new art”.
“Models are only good as the data and assumptions that go into them,” the report states. “This is why more attention needs to be devoted to building a longer and higher quality historical archive of events and losses, and also to developing probabilistic models for a broad ranges of perils.”