Storm warnings

With economists taking a reductionist approach to economics, another financial crisis is inevitable. They must learn from the mistakes of weather forecasters and changing their predictive models

David Orrell
Author: David Orrell
October 23, 2017

Weather forecasters have one advantage over economic forecasters: even if they struggle to predict storms, no one asks them to prevent them.
In his short, bracing and very readable book Can We Avoid Another Financial Crisis?, the economist Steve Keen answers his own question with a resounding ‘no’.

In fact, according to Keen, most developed countries already find themselves in one of two classes: ‘debt zombies’, or ‘debt-zombies-to-be’. The latter includes the US, UK and so on, who are already among the walking dead of debt. The former includes countries such as Australia and Canada, which survived the financial crisis relatively unscathed, but only by embracing excessive levels of private debt.

Keen does have some suggestions on how to get out of this mess, but admits that most of these measures – such as using a version of basic income to help pay off private debt – will never be accepted by politicians, who are controlled by brain-eating neoclassical economists (Keen doesn’t actually call them that, but I’m reading between the lines).

According to Keen: “Mainstream economists are the real culprits in the crisis and its aftermath, since they advise governments that credit is in fact benign, that rising private debt is no cause for alarm, that a bigger and politically more dominant finance sector is in fact good for the economy, and that the government should avoid running deficits.”

Economic reductionism
Keen explains that the basic problem is that mainstream economists have followed a particular modelling agenda, which attempts to reduce everything to microfoundations. This is consistent with the traditional reductionist approach in science, which treats the world as a mechanical system. However it breaks down for complex systems, which feature emergent properties that, by definition, cannot be predicted by breaking them down into parts.

My favourite example is clouds, which are formed when water vapour accumulates on small particles in the atmosphere. We know a lot about particles and water vapour, but we still can’t produce clouds on a computer, which is one reason weather forecasting is so difficult.

Keen suggests that we should go back to simple models. He explains the causes of the crisis, and the coming zombie invasion, based on a set of equations featuring only three variables and nine parameters. This model is far simpler than the elaborate models favoured by macroeconomists, but its behaviour is more complex because the equations are nonlinear, have non-equilibrium solutions and include the effects of the financial sector, such as creating money in the first place.

In The Money Formula, a book jointly written by myself, quantitative finance expert Paul Wilmott relates a similar story: he came across a book about the Treasury’s model of the UK economy in the library of Imperial College, London, which contained 770 equations. As Wilmott noted: “What you want to do is throw away all but the half dozen most important equations and then accept the inevitable, that the results won’t be perfect.” Sometimes less is more.

Toy model
Keen’s book persuasively shows how excessive debt creates a trap from which economies find it hard to escape – and it rightly blames the problem on economic theories. If I have a quibble with the book, though, it is that it misdiagnoses the reason why these theories evolved as they did: it’s not just about the desire for microfoundations, but also the need to tell a particular story.

To illustrate the usefulness of small models, Keen describes the Lorenz system – a toy model based on the equations for atmospheric flow – which was one of the first models to illustrate chaos. In principle – as its inventor, the meteorologist Ed Lorenz, suggested in a 1972 talk – the flap of a butterfly’s wings in Brazil could set off a tornado in Texas. “To apply Lorenz’s insight,” writes Keen, “meteorologists had to abandon their linear equilibrium models.”

There seems to be a trend amongst economists to compare their field with meteorology. After the Bank of England predicted a Brexit disaster that didn’t occur, their Chief Economist Andy Haldane described it as a “Michael Fish moment”, referring to the BBC weather forecaster’s 1987 prediction that no hurricane would hit England. According to Haldane, the forecasting error kickstarted a technological revolution at the UK Met Office which economists can copy.

However, there hasn’t really been a revolution in weather forecasting, and the equations haven’t changed that much. The big improvements have been in global data collection systems and computational power. These have little to do with Lorenz’s model and more to do with global technological developments.

Weather forecasting remains married to a reductionist approach and has done little to adopt new ideas from areas such as machine learning. Rather than copying it, economists should learn from its mistakes.

In fact, the main reason Lorenz’s model became popular in the 1990s was not because weather models were highly chaotic – they weren’t – but because it was a perfect excuse for forecast errors. It was therefore the meteorological equivalent of the efficient market hypothesis, which is popular largely because it explains why economists can never predict the economy.

This points to the real reason why the models used by economists are so backwards: the aim of models is not to predict the economy. It is to tell a particular story.

The equilibrium models of mainstream economics are based on rationality, stability and efficiency, not because these are realistic assumptions, but because they are perfect for the financial sector’s PR departments, which fund or otherwise cosset mainstream economics departments.

As Keen rightly points out, the most basic flaws of these models is that they omit the financial sector. But that is no accident. The biggest role of the financial sector is right there in the models – its own omission. Trying to predict economic storms without including money is like trying to predict the effect of a storm without including water – which is why Keen’s storm warnings are so important.