In our 2010 book “Alchemists of Loss” Kevin Dowd and I examined the record of Wall Street risk management leading up to the 2007-08 financial crisis and concluded that it was seriously lacking. The universal assumption of “Gaussianity” caused participants to lose sight of the possibilities of “fat tails” where the probability of bad outcomes is seriously understated and “long tails” where potential outcomes are far worse than the model shows. It is by no means clear that Wall Street has cleaned up its act (certainly, it has not employed us to help it do so)! It is, however, increasingly clear that the mis-assessment of risk spreads far beyond Wall Street.

The classic example is the origin of World War I. Nobody in 1914 wanted a European war. Certainly, nobody in 1914 wanted the European war that actually occurred: the most destructive in world history, both in the short-term to its participants and in the long-term through loosing the scourges of Communism and its later counterpart Nazism upon the world. Only Woodrow Wilson can be said to have benefited from it, by having his country participate in a relatively short and successful war and then being empowered to impose his authoritarian leftist fantasies on the entire world through the Congress of Versailles. However, Wilson was notably not involved in 1914’s decision making.

In terms of the risk distribution facing the leaders of Europe after the assassination of Archduke Franz Ferdinand at Sarajevo on June 28, 1914, the “tails” were both fatter and longer than they calculated. They were “fatter” because the endless succession of medium sized crises in Europe since 1905, from the Algeciras dispute onwards, had made all the major participants more belligerent and resentful towards each other than in a truly peaceful environment, so the probability of a war was far greater than policymakers assumed. They were “longer” in that the insane alliance system built up by the Franco-British “Entente Cordiale” of 1904 and its “Triple Entente” successor with Russia in 1907 had made Germany and Austria-Hungary correctly paranoid about their encirclement by the threatening coalition of Russia, France and Britain. Hence, any war that took place, being existential for the Central Powers, was likely to be much longer, more embittered and more destructive than imagined at the time. What was needed, probably from Britain as the least irretrievably compromised Great Power, was a Castlereagh, tirelessly travelling across Europe to establish good relations with the opposing powers. The idle and intellectually underpowered Sir Edward Grey was utterly inadequate to this task.

Nevertheless, even had today’s computerized risk modelling been available in 1914 (by some wonderful steampunk development of Charles Babbage’s “Analytical Engine” perhaps) and been applied to international relations, it would have given policymakers the wrong answer. The Gaussian distribution, as used by all current computerized modelling systems, whether in the markets or in climate change modelling, has risk “tails” that are both thin and short. It greatly underestimates the probability of something going badly wrong – it fails to account for the hidden correlations between different risks — and it greatly underestimates how bad things could get – anything more than 3 standard deviations from the mean is wrongly shown to be vanishingly unlikely.

The new science of non-parametric statistics is showing signs of correcting the follies of Gaussian models, at least for market applications where vast accurate price histories are available. In this approach, the modeler does not make any assumptions about the distribution that a collection of data follows but allows that distribution to fall out from the data itself. Thus, if you have a collection of stock or bond market data that includes the financial crises of 2008, 1987 or 1929, it will under this approach tell the modeler that the distribution is not Gaussian with an unexpectedly high standard deviation, but something much more prone to fat or long tails. It will suggest a fuzzy logic-based distribution if the data shows fat tails to be likely (for example, in securitized home mortgages), or a Cauchy distribution if the data points to a prevalence of long tails (for example in stock prices over a “crash”). If the non-parametric model is sufficiently sophisticated, it may point to yet another different type of distribution, previously unconsidered by the modeler.

Another current example where decision-maker opinion is warped by Gaussian assumptions is the Russia-Ukraine War. Now that President Trump has been elected, there are still two abnormal “tails” that make the situation more dangerous. The “fat tail” which now seems of high probability is that at Trump’s instigation, the United States more or less drops out, at which point the European countries step up their backing for Ukraine; however that backing is unlikely to be sufficient to bring victory for Ukraine, so a prolonged “World War I” situation occurs, where Ukraine loses in the end, but only after several years of attrition, immensely costly in human and material terms. The “long tail” is that either Trump loses control in the United States or he is persuaded to step up sharply military assistance to the beleaguered Volodymyr Zelenskyy, to give Ukraine a decisive “win”. In that case, the probability of all-out nuclear war with Russia is not zero; the distribution’s tail is “long.” Those tails must be managed, so peace is essential; if that requires replacing the Marxist clown dictator Zelenskyy with a mainstream Ukrainian politician such as Yulia Tymoshenko, the price is infinitely worth paying.

Trump’s tariff initiative is a good one, but as with all new major policy changes, it has tail risks. The fat tail, more likely than perhaps might appear, is that it fails, for example through Congress lumbering into action to block his initiative, demonstrating yet again its complete inability to achieve anything useful. The U.S. would then relapse into the globalist “woke” mess in which it has been wallowing for the last two decades, losing competitiveness through “climate change” nonsense and giving China an unopposed geostrategic victory. The “long” tail, very unlikely but not of zero probability, is that high U.S. tariffs combine with climate change rubbish in other countries to spiral the global economy into a permanent 1930s, in which living standards inexorably decline and everything stops working, as it did this week in Spain.

The solution to avoid both tails is for Trump to settle on a program of universal low tariffs, at the 10% to 20% level and to do so quickly enough and coherently enough that any Congressional rebellion is quelled, so they can get back to their proper job of cutting spending and passing a budget. In that case, both the globalist decline and the protectionist crash would be avoided.

The existence of tail risks is one very important reason why any kind of world government should be avoided at all costs. As the World Economic Forum has shown, any such government would pursue an appalling policy of fashionable socialist nostrums – the WEF motto of “you will own nothing and you will be happy” would be its central goal, reducing ordinary people to serfdom. With a monopoly government, there would be no way to assess or avoid tail risks, and no way of providing a policy counterexample such as President Trump’s United States that would show how appallingly wrong was the fashionable conventional wisdom. Far from avoiding tail risks, such a government would embrace them, leaving no way out beyond universal violent – and justified – revolution.

It may be decades before decent non-parametric mathematical models appear that can assess “fat” and “long” tail risks outside the financial markets where quantitative data is plentiful. Certainly, conventional wisdom will suppress them. Just as Wall Street wants the freedom to trade excessively without worrying about existential risks (for which experience has shown they will be bailed out at the expense of the rest of us), so policymakers will fiercely resist the introduction of models that show the true cost of Edward Grey’s idleness or Boris Johnson’s Ukraine fecklessness, killing a peace deal in March 2022. Yet mathematicians should build such models. For the rest of us, enslaved by fashionable leftist folly among decision-makers, mathematics in the form of such models has the potential to set us free.

-0-

(The Bear’s Lair is a weekly column that is intended to appear each Monday, an appropriately gloomy day of the week. Its rationale is that the proportion of “sell” recommendations put out by Wall Street houses remains far below that of “buy” recommendations. Accordingly, investors have an excess of positive information and very little negative information. The column thus takes the ursine view of life and the market, in the hope that it may be usefully different from what investors see elsewhere.)