Fighting (and Modelling) the Last War

“History must be lived forwards, but can only be understood backwards”

In military parlance, the concept of ‘fighting the last war’ is understood as leadership applying the tactics and lessons from the last major conflict to the current situation. The costs can be catastrophic. German cavalry in 1914 appeared on the battlefield in the spiked Pickelhaube helmet, spears, and beautiful mounts. While they may have looked dashing, their ineffectiveness by 1915 led to the Cavalry Corps effectively being redesignated to infantry units given the emergence of machine gun and trench warfare.

German Cavalry, 1914

German Cavalry, 1914

Otto von Bismarck in the Pickelhaube, circa 1880

Otto von Bismarck in the Pickelhaube, circa 1880

France’s famous Maginot Line was reminiscent of fortified trench systems from World War I, even as the Germans had developed a new method of combat, “Blitzkrieg”, that effectively rendered the Line nearly useless. Korea, Vietnam, and more recent conflicts have shown that this phenomenon is not something that modernity has inherently ‘figured out’, instead for thousands of years, kings, generals and officers have lost battles, wars, nations and empires by fighting the ‘last war’.

 Any given Sunday, you will probably see some NFL coach fighting the ‘last war’ – an outmoded and predictable set of plays that will likely lead to a few wins and his job. In the trenches of WWI, the consequences were catastrophic. Millions of lives were thrown into the meat grinders of the trenches due to inept leaders not adapting to the new rules of battle.

 Central banks are the modern ‘omniscient’ force fighting the last war. By 2008, three letter acronyms (TLAs) unrecognizable to ‘even’ an equity salesman on Wall Street just a year or two prior were commonly overheard in gyms, taxis and coffee shops. Since then we’ve had TARP, multiple rounds of QE, operation twist, ZIRP, and if you’re ‘lucky’ enough to live in Europe, NIRP. The beginning of the reversal of these policies had the bond and equity markets screaming bloody murder by December 2018, enough so that the Fed is about to reverse course. Now the market waits with bated breath for a new round of Fed rate cuts, cessation of QT, and the contemplation of other ‘tools’ (cowbell) to push asset prices higher. Why do they think they can get away with this? Why do they think this will ‘solve’ for any problem? Specifically, because the specter of inflation hasn’t been spotted for decades. But more generally, because markets have been formed into political utilities. I highly recommend Epsilon Theory’s work on this topic.

But why do they think it will work? Central bankers only know how to fight the last battle, while pontificating not only that newly invented gadgets will work, but we know better than you, so back off. The GFC led to all sorts of models to explain, monitor and prevent new crises from emerging. And these are typically calibrated to solve for the types of specific problems that came about in the GFC that 11 years ago lead to massive asset price deflation. It’s amusing (but mostly terrifying) that the Fed thinks that right now their return to loose monetary policy is being proactive to the extent it will protect against a new crisis as opposed to being reactive, reactive in the sense the Germans ditched the horses and spiky helmets for trenches and machine guns after being mowed down. But the Central Bank plan is sadly misguided. How do a bunch of super Team Elite braniacs with massive resources and nearly unlimited power make such modelling errors?

In his excellent book “The End of Theory” Richard Bookstaber identifies 4 phenomena that lead to potentially catastrophic modelling errors:

 1) Emergent phenomena

o   Any number of independent actions cause catastrophic, unforseeable and novel events. Unintended consequences also create previously unseen crises. Think of people who act in a personally rationally manner in their frame of reference, but the sum of all the group actions results in chaos (people rushing to leave a burning theater).


2) Non-ergodicity

o   Probabilities and distributions of certain events are non-static over time. This is fundamental of both physics and picking up a gun and playing Russian roulette. Casinos would have a hard time doing business not only if the odds only changed every game, but if no one knew what they would be in advance (including the house). But much of our natural and social world is non-ergodic, yet we apply ergodic models to them.


3) Radical Uncertainty

o   This is what Rumsfeld famously characterized as “unknown unknowns”. Effectively, you have no way of knowing that you don’t know about the existence of certain phenomena. Newton didn’t know he didn’t know about the ‘existence’ of particle physics.


4) Computational Irreducibility

o   Economists (among others) assume that all phenomenon can be reduced to formulas and models that will predict future outcomes. We know this to be false across many natural and unnatural phenomenon, for instance the Three Body Problem.


Even simple systems involving a small number of interactions (human or otherwise) are typically irreducible. Consider the calculative permutations among hundreds of nations, thousands of companies or billions of people in a non-reducible subset!

We do understand that the above modelling errors ARE going to happen. We cannot erase non ergodicity in non-ergodic things for instance. We CAN acknowledge the nature of the model being applied versus the data set.

Decentralized Command

 In his legendary book “About Face” U.S. Army Col. David Hackworth at length describes the mechanisms of compounding errors due to centralized command during his time in the US Army in Vietnam.

“As a rule, the judgment of those on the ground was respected less and less in the post Korea years. Centralization of command, which had characterized the trench war of  ’52-’53, was alive and growing stronger every day. And it wasn’t just know-nothing commanders making the decisions (which would have been bad enough); now it was a combination of computers and peacetime procedures that ignored the human variables: initiative, potential and personal growth of an individual soldier” (Hackworth, 323)


The concept of decentralized command was his prescription (I am simplifying) for many of the policy and execution errors he saw the US military (and D.C.) making in Vietnam. The centralization of capital allocation through central bank policy is NOT the answer to winning the ‘war’ against deflation and anemic growth that so terrify central banks. The financialization of markets and zombification of corporations sounds a lot like the things identified by Hack (ignorance of initiative, potential and personal growth) within the U.S. military. Model errors will be present as long as we have data and models. Correct application of models, skepticism of ‘Big Data Solving’ and agent based models is the first step to avoiding these mistakes. What’s harder is we can only hope to attenuate the consequences of these errors by limiting the power and scope of those misusing models to form policy. But for now that (Prussian) horse has left the stable.

Of course all this makes the assumption that the Central Bankers are acting as independent and in good faith. A topic for later…