It looks as if the global economy is recovering. If so, it would very probably be good for equities and bad for governments bonds. But it doesn't follow that investors should readjust their portfolios. This is because we cannot be confident that the recovery will continue, simply because economic forecasting is impossible.
The Queen famously asked of the 2008 crisis: why did nobody see it coming? An honest answer would have been: because economists almost always fail to foresee recessions. Back in 2000 Prakash Loungani, an economist at the IMF, studied the record of consensus economic forecasts. He identified 60 recessions between 1989 and 1999 - defined as calendar years in which GDP fell in any particular country - and found that the consensus forecast the previous April had predicted just two of them. "The record of failure to predict recessions is virtually unblemished," he concluded.
The failure to foresee the last recession was not, therefore, an isolated one. It was part of the pattern. This tells us that something is wrong with mainstream economic forecasting. But what? A big part of the answer lies in the economics of complexity - the fact that economic outcomes are the result not only of individuals' decisions, but of interactions between individuals.
To see this, let's start by thinking about how economic forecasting can be considered possible in the first place. From one perspective, it's odd that this should be so. After all, I haven't decided yet how much I'll spend next year: will I change my car? Will I buy an expensive guitar? A lot depends upon how I'll feel, and if I can't predict this, how can anyone possibly second-guess me?
The answer lies in the law of large numbers. Of course, our individual spending decisions have a large random element. But across millions of people, these elements cancel out; if I go on a spree, others might not, or might decide to save instead. With the whims cancelling out, we can think of aggregate spending as arising as if it were the response of a single consumer - a "representative agent" in the jargon - to observable forces such as wealth and interest rates. (These responses might or might not be rational - that's another story.) Equally, spending by companies can be considered as the action of a "representative company". This implies that corporate and consumer spending should be a stable function of a few economic variables and hence forecastable, but for 'shocks'.
All this, though, assumes that the law of large numbers holds. But this requires two conditions to hold, both of which are questionable.
One is that no individual's spending is big enough on its own to affect the aggregate. This is reasonable for individuals. But it's not for companies. Xavier Gabaix at New York University has estimated that variations in the productivity of the US's largest one hundred companies have accounted for more than one-third of the fluctuations in US GDP. This is consistent with a fact about the early 90s recession in the UK - that it hit hard only a fraction of companies, while almost half did well. Economists who think that macroeconomic variations arise from macroeconomic causes are therefore looking in the wrong place.
A second condition is that individual companies or consumers' spending must be independent of each other. Only if this is the case can we assume that random fluctuations in individuals' spending will cancel out across all companies or consumers. Increasingly, though, economists are questioning this assumption. "If agents are sufficiently dependent on their neighbours, aggregation doesn't remove uncertainty," says Alan Kirman of the University of Aix-Marseilles. "Interaction and interdependence are the central motor of the economy." Conventional economic forecasts, which are based on single 'representative agents' - 'the consumer', 'the corporate sector' - thus ignore a major cause of booms and slumps.
Economics, complains Thomas Lux of Kiel University, "has been blind to the role of interactions and connections between actors."
But what form do these connections take?
Most obviously, big banks are dependent upon each other, and companies and households dependent upon banks for credit. This implies that when one bank - Lehmans - failed in 2008, other suffered losses which in turn limited their willingness to lend to other banks, which in turn cut their lending to the real economy. "Relatively small initial liquidity shocks have the potential to make strong contributions to systemic risk," says the Bank of England's Andy Haldane. "Liquidity hoarding can cascade through a banking network, with severe consequences." The key word here is 'network'. What matters isn't simply individual banks' behaviour, but the linkages between them. If these are big and tight, trouble in one bank can bring others' down, but if they are small and loose they won't.
It's not just banks where interlinkages matter. In a recent paper Daron Acemoglu at the Massachusetts Institute of Technology shows that if companies are tightly connected - with one's output being a big input for others - then difficulties in one company or industry will depress output more than if there are weaker input-output links. For example, if oil companies find it harder to extract oil, prices rise to the detriment of the wider economy, whereas if it becomes harder to make, say, biscuits, there will be no such wider adverse effect.
It's not just material links between companies that matter. So too might sentimental ones. Economists have recently found proper scientific evidence to corroborate Maynard Keynes' claim that companies' capital spending decisions depend upon "animal spirits". And such spirits might be contagious - literally. Chris Carroll of Johns Hopkins University has shown that economic expectations can spread in exactly the same way that diseases do; if one company becomes gloomy, its neighbours, customers and suppliers might also, simply because company bosses talk to each other. In this way, it is possible we can talk ourselves poorer, or richer. Again, what matters isn't a 'representative' company, but the subtle and perhaps unobservable ways in which companies and their owners and managers interact.
There are also linkages between consumers. Viral marketing works precisely because some customers learn about new products from others. Your mother is more likely to get a Kindle or iPad if you already have one and so can show her how it works.
Another link is simply that many of us want to emulate others' spending - a desire which generates what Cornell University's Robert Frank calls "expenditure cascades". Big spending by some people, he says, alters others' ideas of what constitutes an acceptable lifestyle, causing them to spend more. If your friends spend £1,000 on their child's birthday party, you'll feel mean if you spend £100, so you'll become more lavish. And if your neighbours drive Audis, you'll feel a prat in a battered Mini Metro.
Of course, the idea of keeping up with the Joneses is an old one. But until recently, it's been hard to find proper evidence for it: if we see a group of neighbours spending more, how can we tell whether some are copying others or simply that the neighbourhood generally has enjoyed some good fortune? Perhaps the neatest evidence here comes from a study of the effects of the Dutch postcode lottery. Every week, this selects a postcode at random and gives a BMW to everyone in it who bought a ticket. Researchers have found that the neighbours of winners who didn't win themselves are significantly more likely to buy a new car. This is a clean sign of a network effect between consumers.
The message here is simple. Booms and slumps are unpredictable because they arise from apparently small events that can be magnified by links between different companies or consumers. If one's pessimism or difficulties lead to others' becoming pessimistic or troubled, we'll get a recession. And if one's higher spending is emulated by others, we'll get a boom. It's the connections that matter.
It's tempting here to say that economists have failed to predict downturns because they've had a mistaken conception of the economy, thinking of it as being driven by the choices of a representative agent subject to shocks rather than as the result of complex emergent processes in which interlinkages between people are key determinants.
However, I suspect it might be that even if we do give due weight to interactions and network effects, recessions would still be unpredictable simply because we cannot say in advance what surprises individuals will get, or whether the links between them will be sufficient to destabilise the economy or not.
There's a parallel here with riots. As Stanford University's Mark Granovetter showed, these happen when a sufficiently large number of people imitate the disorderly behaviour of a few, and their imitation in turn causes others to riot. In such circumstances, bad behaviour cascades through society. However, it is fiendishly hard to say in advance whether there'll be enough such imitation to tip low-level trouble into mass rioting; it might depend on subtle cues and a particular arrangement of individuals which just can't be spotted beforehand.
There's also a parallel here with stock market sell-offs. If one person's selling is imitated by another - because of herding or information cascades - then asset prices can fall further than the fundamentals would justify. Such events are impossible to foresee.
All this matters for investors. It means we shouldn't base our investment decisions upon a view of the future, because the future is unknowable. Sure, we can say that we're more likely to get moderate growth in the next few months than recession. But this is not because we know the future, but because we know the past. Moderate growth is more common than recessions. Predicting it is therefore simply playing the probabilities. There's nothing wrong with this, but it requires no great insight.
The question we should ask is not 'what's going to happen?' We cannot know. Instead it's: 'is my portfolio sufficiently resilient to the large number of possible shocks we could get?' This demands not futurology but some introspection - to know what risks matter to you individually - and a vague knowledge (which is all that's possible) of those risks, the correlations between them, and how to diversify against them. In 2008, banks spectacularly failed to solve this problem. But there's no reason why we ordinary investors should fail to do so.