A panel of academics at ILS Bermuda’s Convergence 2020 conference has slammed the re/insurance industry’s catastrophe prediction models as not fit for purpose.
Cat models that use historical inputs are based on “short and incomplete” data that would be misleading, even if the data were comprehensive, because of the impact of climate change, said Professor Kerry Emanuel, professor of atmospheric science at the Massachusetts Institute of Technology.
Speaking on a panel titled The Effects of Climate Change on Wind, Flood & the Earthquake Zombie Hypothesis, that was chaired by Samantha Medlock, a senior counsel who sits on the Climate Crisis Select Committee for the US House of Representatives, Emanuel argued that climate data was only reliable going back as far as the 1970s.
Recorded data before then is so inaccurate that it is of little use to actuaries, he said. Even if models had a long and accurate data set to draw on, climate change means historical data is a poor indicator of present risk, he added. He called on re/insurers to turn to physical models that calculate risk without using historical data.
Emanuel noted that three separate teams of researchers, working independently of each other, had calculated that Harris County in Texas, which had been struck by Hurricane Harvey, was three times more likely to flood now than it had been in the 1980s.
The general mispricing of risk is having profound social and political consequences. Failure to incorporate more accurate models is putting people’s lives at risk by encouraging them to live in areas that are susceptible to natural disasters, said Emanuel.
“The people killed by Hurricane Katrina arguably died because risk was underestimated,” he said.
However, incorporating more accurate models could devastate some house prices by increasing the cost of home ownership in areas that are identified as being at greater risk of flooding than previously acknowledged. A 1 percent increase in the cost of home ownership could lead to a 20 percent fall in house prices, noted Matthew Eby, founder and executive director at First Street Foundation.
Elby said his model, which is not based on historical data, showed there are many more US homes at risk of flooding than traditional, historical models imply. The model indicates that 14.6 million properties are in a flood risk area, compared to the 8.7 million properties suggested by historical models. Within 30 years another 2 million properties will be at risk of flooding, he added.
Elby hit out at “quasi monopolistic” modelling companies and said the industry was in need of disruption from startups with a new approach. Existing modelling firms “have no incentive to change the way they estimate risk,” he added.
Models become more inaccurate when looking at long tail risks, or low frequency, high severity, events, Emanuel added.
Professor Ross Stein, chief executive officer and co-founder at Temblor.com and adjunct professor of geophysics at Stanford University, noted that although earthquakes are different from floods and hurricanes, in terms of not being a climate-related phenomenon. However, they suffer from similar shortcomings in the predictive models used by actuaries, he said.
Different countries use very different models, creating a lack of consistency in approach and making it impossible to use historical data for these rare occurrences across geographic regions.
Stein noted that airlines had excellent safety records because airplanes have black boxes allowing for detailed analysis every time there is a problem. “Only 1 percent of houses have seismometers,” he noted. “We should have a black box in every building.”
Convergence 2020, ILS Bermuda, Kerry Emanuel, Samantha Medlock, Matthew Eby, Ross Stein, Massachusetts Institute of Technology, First Street Foundation
To read more articles like this one, visit Bermuda:Re+ILS.