Since mid-February the All-Share index has fallen more than one-third and recovered more than 20 per cent. One way to understand such huge price swings is with a new approach to economics, known as complexity economics.
To see how this works, think about something else we’ve seen in recent weeks – the riots that sometimes accompanied the protests over the killing of George Floyd.
Why do riots sometimes happen, but often not? One answer was provided by Stanford University’s Mark Granovetter back in 1978. Imagine three types of people on the streets (In reality, there’s a spectrum: I’m simplifying somewhat). One is a trouble maker. Another is a peace maker who can restrain one trouble maker, but not more. And the third is an imitator who will copy the person closest to him. Now, imagine the trouble maker tries to throw a brick. If a peace maker is close enough to him, he’ll stop him and there’ll be no riot. But if imitators are closest, they’ll follow the trouble maker and throw more bricks. The peace maker will be overwhelmed. There’ll be a riot.
After the riot, moralising politicians will blame the criminality of the mob. But in fact, exactly the same characters are involved in the riot as in the peaceful evening.
What’s this got to do with stock market crashes?
Plenty. Let’s rename our characters. The peace maker is now a seller. The imitator is a momentum trader. And the peace maker is a value investor. Just as we get a peaceful evening when the trouble. maker interacts with the peace maker, so we get market stability when the seller interacts with the value investor. And just as we get a riot if the imitators follow the trouble maker, so we get a crash if momentum traders follow the seller. By the same logic, we can see sharp rises in prices when momentum traders follow the value investor’s buying.
This simple story doesn’t just explain the price moves we’ve seen this year. It also explains why stock markets can be stable for so long but suddenly see bursts of high volatility. And it explains why crashes are unpredictable. They can be triggered by tiny, imperceptible changes in the relationships between people.
But there’s something else. You might think the imitator or momentum trader is an irrational sheep. He’s not. For a long time before this year there was a negative correlation between market volatility (as measured by the Vix index) and subsequent returns: high volatility led, if anything, to lower returns. This meant you could have made good risk-adjusted returns by selling when volatility rose and prices fell. And many people did just this: they are called risk parity traders.
Those “imitators” who sold as prices fell and volatility rise were therefore behaving rationally. They were acting upon evidence. Their selling, however, exacerbated the fall in prices and caused shares to fall to levels which were too low – in retrospect. We can say now that the market over-reacted in March. But this was fuelled by individually reasonable behaviour. Rational people can sometimes give us irrational markets. As Richard Bookstaber wrote in his book The End of Theory “the sum of individual rational actions can be the genesis of a crisis.”
All this highlights some essential features of complexity economics. The first is that relationships, networks and interactions are crucial. Just as we get riots when trouble makers interact with imitators, but peace when peace makers do so, so we get stock market sell offs when risk/parity traders interact with sellers, stable markets when value investors interact with sellers and rising markets when risk-parity traders meet value buyers.
It’s not just trading behaviour in which interactions matter. MIT’s Daron Acemoglu has shown that recessions arise from network effects. The financial crisis caused a recession because banks were important hubs in networks; they were key suppliers of credit, so their failure dragged down other firms. More dispersed networks, in which banks were not so crucial (say because they could be substituted with peer-to-peer lenders or credit unions) would not have produced so deep a recession.
The 2008 financial crisis taught us something else – that you can’t measure systemic risk merely by looking at individual banks alone. Imagine each individual bank was following a different strategy. If one failed, others could then profit by buying its assets cheaply in a fire sale. We might see lots of individual bank failures in this world, but no great systemic risk. But this was not the world we lived in in 2007. Banks were doing similar things, holding mortgage derivatives. Which meant that when one failed, many failed – not least because they had nobody to whom to sell distressed assets. As Andy Haldane and the late Robert May concluded in a classic paper: “Excessive homogeneity within a financial system – all the banks doing the same thing – can minimise risk for each individual bank, but maximize the probability of the entire system collapsing.”
Financial interlinkages are not the only ways in which networks matter. In his recent book Narrative Economics, the Nobel laureate Robert Shiller shows how economic ideas can spread just like viruses, as they are transmitted from person to person. And David Hirshleifer at the University of California at Irvine says that in the process bad ideas can spread faster than good. The golf club bore is quicker to tell you about his successful shares than his many bad ones. Those who believe them therefore get an exaggerated sense of the returns to stock picking and the profits on speculative shares. This, he shows, can lead to price bubbles.
Andrei Shleifer and Brock Mendel, two Harvard economists, have also shown how this can happen. They show that mortgage derivatives became over-priced in the mid 2000s because mostly rational traders were “chasing noise”; they assumed that prices were reasonable because they assumed that others knew what they were doing. But they didn’t. The blind were leading the blind.
Complexity economics, therefore, tells us important things that conventional economics does not. Economic outcomes are not merely individual behaviour or character writ large. Networks, relationships and interactions are crucial.
Which leads us to a second feature of complexity economics. This is that whereas orthodox economics imagines that economies tend to move towards stability and equilibrium, complexity economics does not. Instead, it believes that positive feedback loops can amplify small disturbances. Brian Arthur at the Santa Fe Institute in New Mexico, one of the founders of complexity economics, says these “are very much a defining property of complex systems.”
A big fact, described by Harvard University’s Xavier Gabaix, corroborates this theory. He shows that stock market returns are not distributed as a bell curve. Instead, big moves are more common that such a curve predicts. This is hard to reconcile with equilibrium theory, but it fits with the prediction of complexity theory, that bursts of instability do happen sometimes.
A third feature of complexity theory is that social outcomes – be they riots or stock market crashes – not a simple product of individuals’ intentions. “Behaviour in the aggregate is more than the simple summation of individual behaviours” says Leigh Tesfatsion at Ohio State University.
Nobody tries to be poorer, but recessions nevertheless happen. This trivial fact alone tells us that aggregate economic outcomes are not simply related to individuals’ intentions. In our story of this year’s market fall everybody was behaving rationally, but the result was a market that became too cheap (we know in retrospect).
In this respect, the stock market is like The Beatles. The band made some of the most innovative pop music in history. But when it broke up, its members went onto plagiarise the Chiffons, record the Frog Chorus and give us the mindless dirge that is Imagine. This shows us that the whole can be more than the sum of its parts. Or, of course less: the England football team achieved little in the early 2000s despite containing great individual players such as Paul Scholes, Steven Gerrard and Ashley Cole. This tells us that the quality of an aggregate depends not only upon individuals alone, but upon interactions and relationships between them. That’s complexity.
We have other examples of this. I’ve shown that rational traders can produce irrational markets. But as the great physicist Niels Bohr famously said, the opposite of a great truth is another great truth. In a set of experiments in laboratory markets, Yale University’s Shyam Sunder and colleagues gave computerised traders deliberately irrational algorithms and got them to trade artificial stocks. They found that prices quickly approached those predicted by efficient market theory. This suggests that sometimes irrational people don’t necessarily produce irrational markets.
Alan Kirman, an economist at Aix-Marseilles University, has given another example. He studied his local fish market and found that for most individual traders demand for fish was independent of their prices. That most basic idea in economics – the downward sloping demand curve – did not exist for individuals. But he showed, it did exist for the aggregate market.
And in Haldane and May’s analysis, the financial system as a whole was risky and fragile in 2007 even though each individual bank, judged in isolation, did not seem so.
So, banks, fish markets and stock markets are like the Beatles. Their aggregate properties can’t be read simply from looking at individuals.
If this sounds like a fancy new idea, it shouldn’t. Adam Smith had the same idea back in 1776 when he wrote that “It is not from the benevolence of the butcher, the brewer, or the baker that we expect our dinner, but from their regard to their own interest.” Selfish people, he thought, can deliver benign results. In this sense, says Doyne Farmer at the University of Oxford, Adam Smith was one of the first theorists of complexity.
There’s another aspect of complexity, which follows from the facts that economies don’t always move towards equilibrium and that aggregate outcomes are not individual behaviour writ large. This is that outcomes are unpredictable.
One big fact, reported by the IMF’s Prakash Loungani, tells us this. It is that economists have systematically failed to predict recessions – not just the ones in 2008 but pretty much all recessions around the world since at least 1989.
How can this be? A clue lies in a game (well, more of an algorithm) devised by the late mathematician John Conway back in 1970, called the Game of Life. He got a sheet of graph paper and coloured in a few squares: he called these squares “populated.”. He then devised some simple rules. If a populated square has only one or no neighbours, it dies. If it has four or more neighbours it also dies. If it has two or three neighbours it survives. And if a space is empty but has three neighbours it becomes populated: the space gives birth.
These very simple rules generate different patterns depending on the initial pattern of squares: you can see these by playing the game online, which I strongly recommend. Sometimes, the pattern never changes – if you start with a square of four squares, for example. That’s like an economy staying in equilibrium. From other positions you can generate oscillating patterns, analogous to cycles of booms and busts. Yet others produce patterns that change but then die completely. And others produce patterns that change endlessly, in the way that Professor Arthur sees the economy – “not as a system in equilibrium but as one in motion, perpetually constructing itself anew.”
Mere squares following three simple rules can therefore produce hugely variable patterns. Isn’t it likely therefore that human beings with more complicated behaviour will also generate a greater variety of patterns than the equilibrium or mere cyclical behaviour beloved of mainstream economics?
There’s another feature of the Game of Life. Often, you cannot predict what patterns will emerge given the rules and the starting position. The only way to find out is to actually play the game.
Which opens up a new method of economic research. It’s called agent-based modelling. Just as the Game of Life discovers what happens when cells are given an algorithm to solve, so agent-based modelling treats individuals as solvers of specific algorithms and plays out what happens when those algorithms are solved.
Todd Feldman at San Francisco University gives us an example of this. He asked: what are the most expensive mistakes that investors make? It’s difficult to answer this by looking at real traders because humans make a mixture of right and wrong decisions. So instead he programmed artificial traders with specific biased decision rules and got them to trade, seeing which lost the most money. He found that the costliest error is the belief that recent trading conditions – be they bull or bear markets – will continue. This, he found is three times as expensive as our tendency to hold onto losing shares too much.
Complexity economics, then, comprises four big insights: that networks and interactions are crucial; that aggregate outcomes cannot be read off from individual character or intentions; that economies don’t always move towards stability; and that outcomes aren’t predictable.
You might think all this is obvious. Maybe. But it is a massive contrast to mainstream economics, some of which you might have imbibed even if you haven’t had the misfortune of formally studying it.
In its basic form this analyses the behaviour of “representative agents” – a single consumer and firm – who maximise their profits and utility in the face of external shocks which drive them temporarily away from equilibrium. Of course, this basic approach has been refined in many ways. But the Nobel laureate Joe Stiglitz has likened these adjustments to Ptolemaic epicycles – attempts to fix a faulty theory with increasingly awkward patches.
Such an approach, though, denies the essence of complexity economics. In thinking of 'representative agents' it removes any consideration of network effects and invites us to image the economy as being akin to individual behaviour. And the assumption of equilibrium removes the instability and unpredictability emphasised by complexity economics.
Mainstream theory has, however, led to absurd claims. In 2003 Robert Lucas – who won a Nobel prize for his work on it – claimed that macroeconomics’ “central problem of depression prevention has been solved, for all practical purposes.” Five years later, the world economy suffered the worst recession for over 70 years.
That episode alone, allied to the greater ability of complexity economics to explain key facts such as the unpredictability of recessions and oscillations in stock market volatility, surely tells us to take complexity economics seriously.
What should investors do about this?
Mainly, we should recognise the ubiquity of risk. The key message of complexity economics is that apparently stable markets can suddenly collapse or fall into long-term decline. But risk isn’t always on the downside. The same network effects that can cause price falls to feed on themselves can also cause rises to feed on themselves too. Instability works in both directions.
Luckily, we know how to mitigate the dangers of momentum-driven bear markets whilst also riding bubbles. We can use the 10-month average rule proposed by Meb Faber at Cambria Investment Management. This says we should buy when prices are above their 10-month average and sell when they are below it.
This rule does not, however, protect us from sudden crashes. For this, we need assets other than equities such as bonds, foreign currency and cash. Simple portfolios of these have been surprisingly resilient during this year's stock market sell-off. Just because the world is complex it does not follow that our portfolios should be. Sometimes, the answer to complexity lies in simplicity.
In fact, the main message of complexity economics is that we must pay more heed to risk management, because the world is more fragile, uncertain and unpredictable than we think. The notion that greater knowledge will give us great ability to predict events might well be just a fantasy.