Has stock market predictability disappeared? I ask because of the performance of some once-useful leading indicators of returns.
A few years ago, a handful of them did a near-perfect job of calling the market, predicting decent rises in 2017 and 2019 and a fall in 2018. But they failed in 2020, forecasting a small rise when in fact the index fell over 12 per cent. In itself, that’s understandable: nothing and nobody foresaw the pandemic. What’s more significant is that these lead indicators failed this year as well, predicting a fall in the market whereas the All-Share index is 12 per cent up as I write.
In part, there’s a reason for this. One once-good lead indicator has stopped working. For many years, big foreign net buying of US equities led to stock markets falling while net selling predicted rises: for example, the market falls of 2001-02 and 2008-09 both followed record buying. This was because foreign buying was a sign of high confidence and hence that equities were over-priced.
This relationship, however, has broken down recently. Non-Americans were huge buyers of US stocks in 2020 and yet markets have since risen. This is because such buying is no longer a sign of confidence but perhaps its opposite: in times of uncertainty people flock into familiar assets – and these include Apple (US:AAPL), Amazon (US:AMZN) and Microsoft (US:MSFT). Such uncertainty and nervousness, however, can lead to prices rising in subsequent months as sentiment recovers, which means the link between foreign buying of US stocks and subsequent returns has flipped over. One once-good lead indicator is no longer so reliable.
But this is not the whole story. If we recalibrate our lead indicators to exclude foreign buying of US equities, 2021 still saw a failure. They predicted only a tiny rise in the All-Share, not the decent one we’ve actually enjoyed.
One culprit here is the dividend yield. Last December it was close to its long-term average, which predicted only moderate returns.
A bigger culprit, though, is the ratio of the money stock in developed economies to global share prices. A year ago this was below its long-term trend, which suggested that shares were over-owned and too expensive. But they have since become only more expensive.
The thinking behind these lead indicators was straightforward: sometimes, markets become too cheap or too dear and ratios of share prices to dividends or the money stock have in the past told us when this is the case. Such thinking, however, has let us down this year.
This could be just one of those things. No set of lead indicators is perfect, and there’s always some gap on average between even good forecasts and the outcome. The standard error of our lead indicators is such that you’d expect an error of the size we’ve seen this year roughly one year every six.
But that’s just one hypothesis. Another possibility is that the market is predictable, but not at horizons as short as 12 months. Another is that there’s only predictability when the market is egregiously mispriced as it was in 2000 or 2009, and last year was not one of these times.
Yet another theory, though, is that predictability has disappeared for good. Which is plausible. Investors should not leave money on the table: if they discover a way of predicting returns they should act on it and in doing so they eliminate such predictability. John Cotter and Niall McGeever at University College Dublin have shown that this often happens.
There’s only one way to test these competing theories. And that’s to use these lead indicators to forecast what’ll happen next year. They are now predicting a significant fall in the All-Share index, of around 16 per cent, with a less than 10 per cent chance of the index ending 2022 higher than it is now. This is because the slightly bullish signal of prices being above their 10-month high is offset by the fact that both the dividend yield and money-price ratio are below their average.
I stress that this is a test of a hypothesis rather than an actionable forecast – although it does pose the useful question of what bullish signals we have to offset these bearish ones. Yes, I personally hold more cash than most investors, but this is because I’m more risk-averse than most rather than because I have a strongly bearish view.
And this is the purpose of forecasts. If they are to be useful it is as tests of a theory.
Which is the issue here. It does not matter in itself whether I’m right or wrong (though I have been for the last two years in this respect): one man’s judgment is uninteresting. What matters is the question: are moves in the market predictable on the basis of currently-available information? Important as it is for investors, this question has much wider and deeper significance. The question of how much predictability there is in complex systems goes to the heart of how we should think about politics, society and the economy.
