Maybe We've Been Wrong About Climate Change

August 2, 2021

A key element of managing investment risk involves a concept called “tail risk.” It is used to measure the probability of an extreme scenario that would cause returns to be well outside of their normal range. The global pandemic that caused a rapid 30% drop in stock prices is an example of a tail event that occurred just last year. The near meltdown of the entire U.S. financial system in 2008 is another.

In my experience, typical industry risk models do a solid job of accounting for day-to-day risks but a fairly poor job of accounting for tail risk. This is likely because the models are designed by humans, and humans are subject to bias. We tend to anchor by assuming that the future will look something like the past. We also tend to overestimate our ability to identify the factors that will drive future outcomes. The hubris embedded in this mindset tends to undervalue tail events, which usually have little historical precedent and by definition have a very low probability of occurring.

Tail events can be either positive or negative. Because no one is going to lose sleep over a surprise 25% gain, negative tail events understandably garner more focus. If appropriate guardrails are not in place, a negative tail event can result in a complete wipeout (e.g., Long Term Capital Management in 1998).

Recently, we have experienced several weather scenarios that not long ago would have qualified as tail events. However, the frequency with which phrases like “extreme” and “record-breaking” are being used implies that such events are now more likely the normal course. In other words, the tragic recent events in Germany probably cannot be chalked up as the proverbial “100-year flood”.

All of this gets back to the models. The models have trained us to think of climate risk in terms of how many degrees of warming we can tolerate over the rest of the 21st century. As a result, climate pledges tend to coalesce around goals that stretch out to 2040 or 2050. This is fine in a baseline scenario, but probably nowhere close to good enough in a tail scenario.

Competent risk managers don’t take the output produced by a model and assume it represents the full range of possibilities. Instead, they assume something significantly worse could happen and assess whether that outcome could be absorbed without devastating consequences. If the answer is “no”, they take decisive and immediate action to reduce risk. If a tail event never happens, so be it. But eventually, it almost always does.