Complexity of emerging risks outstripping cat models’ capacity to forecast losses

Traditional modelling techniques will not work for emerging risks, Oxford University researcher Peter Taylor told delegates at a seminar in London, England, entitled “Catastrophe Modelling from a European Perspective.”
The Reinsurance Association of America and the International Underwriting Association (IUA) of London jointly organized the seminar.
Taylor’s presentation asked whether traditional cat modelling is up to the task of helping to underwrite risks associated with very complicated emerging risks.
In particular, he noted 16 potentially massive social upheavals arising from, among other things, famine, pestilence, war, drought, demographic shifts, machine intelligence, human-constructed pathogens, nuclear fallout, severe weather and a loss of biodiversity.
The challenge in using current cat models to forecast these events is that underwriters have little or no data to go by in predicting the social consequences of these risks.
In some instances risks related to machine intelligence, for example there are few or no precedents to provide a quantitative basis on which to underwrite the risk.
Some of the above risks are based on complex systems changing faster than human institutions.
“When things go bad, they can go very bad in all sorts of ways,” Taylor observed in the notes for his presentation. “We don’t have much knowledge of the knock-on effects that could affect insurance” in the areas of contingent bodily injury claims or a financial system failure, to cite two examples.
The inability of traditional models to forecast hurricane seasons accurately is a case in point, Taylor said.
He presented one chart in which three significant forecasters predicted no more than 15 named storms in 2005, when the actual number of named storms came just short of 30.
And in 2006, forecasters called for between 13 and 17 named storms, when in actual fact there were nine. “The track record isn’t good, folks,” Taylor noted of the models’ ability to predict severe weather.
In the end, Taylor said models would have to become much more sophisticated, drawing on knowledge learned from “complex science.” Complex science has been used to predict extreme correlations and fluctuating phenomena related to the stock market, for example.