Climate forecasting 'absolutely useless'
Ron Dembo has worked on risk-based algorithms for more the past three decades. He set up three companies after academic research at Yale University and MIT, and his latest venture, riskthinking.AI, models climate change scenarios. He says he takes issue with how the finance industry looks at environmental, social and governance (ESG) data.
Q: What’s wrong with ESG data?
A: There are so many holes. Just to give you an example, rating agencies, indices providers — you name it — their data lacks consistency. It’s not completely science-based and there’s a lack of multi-factor analysis. [Mr Dembo’s company ultimately competes with those providers.]
When you’re looking at a climate event, this is usually the result of several factors coming together. A fire event is a combination of droughts, heat and high winds. The question for the future is: when will we see a particular combination of a drought, heat and high winds that will cause another fire in Australia or in California? Most of what I see out there, if not all, is backward-looking and based on past data.
Q: No more forecasting then?
A: We need to get away from forecasting. Forecasting will be an absolutely useless task going forward. I’ve just written a book about it, Risk Thinking; it deals with what others have tackled — Mervin King and John Kay in Radical Uncertainty, and there are a few other books. In a radically uncertain world, probabilities are meaningless; correlations are meaningless. You need to look at scenarios.
Q: Care to offer a practical solution?
A: In the finance world you have a standard called GICS [global industry classification standard, created by indices provider MSCI and rating agency Standard Poor’s]. GICS allows any investment manager to look at equities in the same way, or at exchange-traded funds in the same way. That’s one of the things missing in climate change analysis. We developed a standard for environmental factors much like GICS, and we’re going to make it open-source for not-for-profits.
