To see this, think about the market's opinion about the Federal Reserve's decision on whether or not to reduce the pace of quantitative easing. This raises two problems.
One is that the Fed's decision is potentially ambiguous for shares. For example, a decision to continue QE might, in theory, be bad for the market; if the Fed believes the economy is so weak that it needs big policy support, equity investors should worry. Or it might be good: printing money supports share prices.
Judging from the market's immediate reaction to the Fed's recent surprise decision to continue QE, the bullish reaction to QE was the dominant one. But this leads to our second problem. If a reduction in QE would be a bad thing, when should we worry about it happening? It's obvious that QE cannot continue forever and so must be withdrawn sometime. But it doesn't follow at all that it's profitable to worry now. The investor who feared tapering at the start of this year would have missed out on a near 1,000-point rise in the FTSE 100 between January and May. And the investor who worried in late June would also have missed out on a rally. The time to worry was in late May and early June - not much before, and not much after.
What we have here, then, are two co-ordination games. Each individual trader is trying to answer two questions: what will others think about the effects of monetary policy? And when will they think it? The successful trader is the one who anticipates which co-ordination equilibrium others will alight upon.
In this sense, Maynard Keynes was right. Investment is like one of those now-defunct newspaper competitions in which entrants are asked to guess which woman other readers will think most beautiful: "We devote our intelligences to anticipating what average opinion expects the average opinion to be."
What matters, therefore, isn't necessarily reality - although the truth can be a co-ordination equilibrium - but others' opinions. Maybe Fed tapering would be bad for the economy. But if other investors aren't going to believe it is, there's little point in us selling just yet.
This means that investors face uncertainty not just about the macroeconomy and policy, but about other traders, too. As Stanford University's Mordecai Kurz has stressed, uncertainty is endogenous; it isn't simply something imposed upon the market by events, but rather is an inherent feature of the trading process itself.
Such uncertainty means that quite rational investors can generate price volatility that would look irrational, if it were the conscious product of an individual. For example, if I worry that you'll worry about the Fed tapering, I'll sell. And if you worry that I'll worry, you'll sell. Prices could therefore fall a long way, without much change in the objective facts; this is just what happened in late May.
This volatility will be magnified if my selling causes you to sell - if you think my selling reflects some superior knowledge. Brock Mendel and Andrei Shleifer, two Harvard economists, have shown how volatility and mispricing can happen if rational investors "chase noise". If they wrongly interpret a price move as a sign that others know something they don't, they'll wrongly jump on to a bandwagon. This, they believe, helps explain why markets seemed to believe credit risk was low in 2007. It wasn't so much that "Mr Market" was relaxed, but rather that individual traders each thought that the low price of risk meant that someone else knew something they didn't, and so didn't bet against those prices. They therefore arrived at a co-ordination equilibrium that was, in retrospect, wrong.
However, bubbles and high volatility don't happen very often. More likely, my selling meets with your buying and so prices are quite stable. Instead, volatility comes in bursts; it was low in the mid-1990s and mid-2000s, for example, but high in 2000-01 and 2008-09. Such a pattern is hard to explain if you think of the market as being like a person. But it arises naturally if you think instead of prices being the result of traders trying - not always successfully - to anticipate others' behaviour.
But here's the problem. There's a massive difference between explanation and prediction. Although it's possible to explain volatility and mispricings after the event, it's not necessarily possible to predict them in advance - not least because other investors are trying to predict our predictions, and so on.
And herein lies a reason why the "Mr Market" metaphor is dangerous. It invites us to believe that price moves are predictable, because most individuals' behaviour is reasonably predictable most of the time. If, however, price moves are examples of complex emergent behaviour and are due to investors trying to anticipate others, then predictability disappears, and with it the market’s resemblance to any individual.