More extremes are coming… everywhere

If you have followed the stream of climate change science and have some notions of statistics, there is a chance that you noticed something astonishing in the distribution of rising global temperatures. Extreme temperatures (especially the hotter ones) are becoming more and more common. Patterns are breaking down. One of the most famous and striking visuals of this is probably Robert Rohde’s acclaimed Youtube video, from which you can extract the two following screenshots of the average global temperature distribution in 1875 and 2010:

Global average temperature distribution in 1875 vs. 2010 where I have outlined the so called ‘fat tail’ of extreme high temperatures in 2010.

A recent LinkedIn post from energy guru and Carbon4 CEO, Jean-Marc Jancovici put the nail in the coffin:

LinkedIn post from Carbon4 CEO J-M Jancovici. Translation: ‘Jan 2020 beat all record in Europe, says Corpernicus, the European Observation service. Non only the climate is becoming warmer, but its variability, and hence extreme temperatures, change even faster…’

Jean-Marc basically sums up this Copernicus article.

From geopolitics to finance, to national politics in the West (see Barclays’ 2016 “The Politics Of Rage”) the frequency of extremes does not seem to affect only our climate. One of the main objections to this theory is that financial markets’ average volatility has declined continuously for many years (which is true), an impression which is reinforced by 12 consecutive crisis-less years and the longest economic rebound in American history. The explanation to this calm seems rather boring. Constant central banks’ intervention — everyone has in mind Mario Draghi’s 2012 legendary “whatever it takes” — and multiple fiscal stimulus programs of various governments seem to have conspired to “solve the economy” by reducing to virtually to zero any possibility of a 1929 and 2008-type disaster. Edward Chancellor, recently quoted on Tweeter by John Hussman notes that this is actually nothing new. About hundred years ago i.e. on the eve of the Great Depression, it was so consensual among economists to consider that “the Federal Reserve, with its ability to control interest rates and conduct ‘open market operations’” was in the “1920’s the’ remedy do the whole problem of booms, slumps and panics’” that “Barron’s weekly envisaged a ‘new era without depressions’.”

The task of controlling an inherently unpredictable processes is not only an illusion, but a dangerous one. I recently posted a Twitter thread on this topic which serves as the basis for the present article. In a nutshell, since central banks are limited by a set of factors (human and monetary, especially their currency), their actions are necessarily confined to a limited zone of market events and events in general. You would not expect a central bank to find a COVID-19 vaccine or take charge of the precarious US/Russia geopolitical relationships. Hence when, for instance, a central bank decides to tackle elevated market volatility, this can only amount to “compressing” an area around the average to reduce its variance as such:

The effect of trying to control equity market returns reduces their variance locally around the average: they turn the red curve into the blue one.

Since it cannot take care of all the events everywhere in the world at all time, its action will necessarily forsakes some of them, and for productivity reasons, the events which fall out of its scope are the rarest, i.e. in the tails of the distribution.

If we assume that the variance of the events of the universe oscillates around some sort of natural average, the total variance of all events is constant over time. This is arguably a contentious point which is discussed further below but if we admit it for now, as the area under the distribution is always 1, reducing the variance of its centre means increasing the area under its tails.

Hence the resulting action of a central bank on the whole distribution would turn the red distribution in the chart below into the blue one:

Provided that variance is mean-reverting reducing the “natural” variance around the average increases the distribution tails (chart source: https://www.riskprep.com/all-tutorials/36-exam-22/145-understanding-kurtosis). Note: the two distributions have the same variance.

The thickening of the tails is measured mathematically by the increased fourth moment of the distribution which is called “kurtosis” (the kurtosis risk is what Nassim Taleb calls the fat tail risk).

Let’s wrap it up. Contrary to general belief, as the local nature of government and central banks’ actions imposes that they are limited to a neighbourhood around average events, it is obvious that those actions cannot lead to a reduction of the probability of extreme crises, but only to their increase. Since central banks can only reduce volatility locally, they mechanically increase the probability of black swans.

Does it make sense to assume that the variance is mean-reverting? There are pretty strong arguments supporting this thesis and the article quoted in Jancovici’s LinkedIn post is just one of them.

Nature cites the thickening of the global temperature distribution tails as one of the direct results of human activity on nature, an activity which precisely aims at locally reducing the hazards that humans face in their daily life and gain better general comfort.

Nature explains that human’s action on its environment increases temperatures but also the tails of global temperatures’ distributions (source: https://www.nature.com/articles/srep05884)

It has also been one of the IPCC reports’ conclusions for many years that climate change will increase the extremeness of weather events (although not their probability if we are to believe current models).

At a recent conference at Télécom Paris, Jean-Marc Jancovici said in essence that only one thing is certain with climate change, it is that it will increase uncertainty. Precisely, an upsurge in black swans does reduce our ability to predict our future. Since 2003, and even more so since 2008, “3-sigma” events (in normal distribution) on the financial markets are more and more common, as shown by this Morgan Stanley chart:

Although financial markets’ volatility has decreased over time since 2008, extreme events are on the rise

Like in the cartoon illustrating this piece, this all seems to be true in a more general fashion.

The limited nature of our actions, be it because of human, financial, conceptual or energy resources, guarantees that any attempt to control a random process will likely end up in tears.

What is the next such “unintended consequence moment” on horizon? If you are reading this, you probably remember the 2008 Great Recession. What you might not know, as it was far less advertised, was that global conventional oil production peaked around 2008 (acknowledged by the IEA in their WEO 2009). If you don’t think this is just a coincidence — and I will one day show why it’s probably not, in the meantime you can read this excellent primer on the subject –, bad news : peak shale oil peak is approaching and will probably happen before 2030. If you read Jancovici’s piece on energy, it’s unlikely that you will doubt that its consequences will be comparably devastating to those of the Great Recession of 2008.

Trader on emerging FX (Asia) markets in London for 15+ years. Centres of interest: markets, macro-economics, physics, probabilities, philosophy.