It is now 31 months since the UK voted to leave the EU and we are no nearer to knowing what sort of Brexit we’ll get or even whether it will happen on schedule. The millions of words spent on predicting the outcome have therefore been noise not signal.
In fact, investors face this problem all the time. We are bombarded with news and opinion much of which is mere noise – in that it provides no good reason for us to act or change our beliefs.
Often, though, investors do in fact trade on such noise, as Brad De Long at the University of California at Berkeley and colleagues pointed out in a classic paper. This doesn’t always lose money. Sometimes there’ll be a bigger fool who’ll buy an overpriced share from us at an even higher price. But sometimes we’ll be that bigger fool ourselves. Trading on noise can cause investors to buy at the top of bubbles, as Harvard economists Brock Mendel and Andrei Shleifer have shown. And in the long-run, noise trading will incur higher dealing costs so we lose money slowly even if we don’t do so quickly.
Which poses the question: how can we distinguish between noise and signal and so avoid being a noise trader?
One way is to have prior beliefs – opinions which provide an anchor for our behaviour and so stop it from being moved by every fleeting observation.
The most extreme such prior is the efficient markets hypothesis. This says that all news and opinion which is more than a few seconds old is already discounted by shares and so is in the price. If this is the case, everything except the most immediate breaking news is noise and not signal and so we should not trade on it.
This, though, isn’t wholly true. There are a few pieces of data that do have predictive power - which are signals rather than noise. We know the direction of the aggregate market can be predicted by: the dividend yield, foreign buying of US equities; the time of year; and the ratio of global share prices to the global money stock. And there are also a few lead indicators of economic growth such as whether the US yield curve is inverted or not and the rate of growth of the narrow money stock in China and the eurozone.
Even these relationships, however, contain some noise. Correlations between them and subsequent outcomes are less than perfect. There’s uncertainty about the precise time lag between predictor and outcome. And of course we cannot be sure that past relationships will continue to hold.
For individual shares we have countless possible signals: economists claim to have unearthed over a hundred “anomalies” – systematic predictors of returns. We should, though be wary of these. Duke University’s Campbell Harvey has warned that most of these findings are “likely false.” And those that are true might not remain so. John Cotter and Niall McGeever at University College Dublin have shown that several ways of picking stocks that worked in the 90s and early 00s have since ceased to do so. This is what we should expect. If investors learn of a signal that a share is under-priced they will act on it and bid the share’s price up to a level from which it is no longer under-priced. The fact that momentum stocks did badly last years suggests that even the most successful signal-based strategy might have stopped working. Even if investors do sometimes leave money on the table, they don’t do so for long.
When faced with any item of news or opinion, we must ask: so what? Does it have any ability to predict the future? The efficient market hypothesis is wrong to claim the answer is always no. But it very often is. In failing to appreciate this, we risk trading on noise. One way to guard against this is simply to ask: if this is a signal for me to buy, why is somebody else going to sell to me? ("Because he's the noise trader" isn't good enough.)
All this said, noise trading isn’t a wholly bad thing. As the late Fischer Black pointed out, it increases market liquidity: if only informed people traded shares, very few would be traded. This, he said, gives investors the chance – if only a slim one – of beating the market. Our problem is, though, that it is fiendishly difficult to exploit noise traders without becoming one ourselves.