“Forward-looking data” isn’t all mumbo-jumbo

We’re all players in the prediction game, whether we like it or not.  This is particularly true of financial risk managers, where the game is played for multi-billion dollar stakes.  It’s often an unrewarding one.  Risk managers have perhaps the least fêted roles on Wall Street.  If their predictions hold and their risk management strategies perform, they’re rarely celebrated as winners (beyond their paychecks!) On the flipside, if they miss something, like the dreaded “black swan event”, they’re often made the scapegoats – even when ‘winning’ is all but impossible. 

Their jobs have been made even harder now because of the theorized threat posed by “green swans” – climate-driven events that have unpredictable, super-disruptive economic and financial effects. 

Forgetting the economy, just predicting how the physical climate will behave as greenhouse gases build up in the atmosphere continues to be a challenge.  Take the ongoing debate over whether climate models accurately forecasted last year’s temperature extremes.  One paper says 2023’s scorching September was an “extremely rare event in the latest generation of climate models.”  Meanwhile, a CarbonBrief article says the apparent acceleration in warming is “broadly in line” with model projections. 

Add in how the physical climate could impact local, national, and international economies over time and the forecasting job gets even tougher.

What’s so difficult is that everything is in flux, everywhere, all at once.  The economy is changing along with the climate, but not at the same time or same rate.  Guessing the potential financial impacts of climate change on, say, a factory near the banks of the Danube River is therefore like trying to bullseye a moving target from the front seat of a rollercoaster – one that springs loop-de-loops on you when you least expect. 

Of course, this shouldn’t – and doesn’t – stop us from trying.  The question is whether we’re going about climate-related financial risk prediction the right way.  This was at the heart of Tony’s post on Sunday, which was in turn a response to Dr Ron Dembo’s Forbes post arguing that seemingly unpredictable climate risks may be anticipated using the stochastic method.

It seems to me the two are talking past each other.  Dembo argues that the stochastic approach produces true “forward-looking data” that is uncoupled from past observations.  Tony believes the “forward-looking data” label is a misnomer, as no matter your modeling approach the outputs are related to the inputs, which consist of historical data.  To my eyes, both leave out important context.