Join our community of smart investors
OPINION

Moderate gains

Moderate gains
January 19, 2017
Moderate gains

Some simple statistical tests tell us that several indicators have in the past predicted subsequent annual changes in the All-Share index. These include:

■ The dividend yield.

■ The yield curve. When long-dated gilt yields are well above short-dated ones, share prices are more likely to rise. This is probably because the yield curve can predict economic activity: an inverted curve predicts a recession and hence falling prices.

■ The ratio of retail sales to the All-Share index. High retail sales predict good returns, perhaps because consumer spending is partly forward-looking and so high spending presages better economic times.

■ Foreign buying of US equities. High buying is a sign that sentiment is high, which is often a sign that shares are overpriced.

■ The ratio of the global money stock to global share prices. When this is high, shares subsequently rise as investors reweight portfolios from cash to equities.

All these indicators have a statistically significant connection with subsequent annual changes in the All-Share index since January 1998, implying that returns are partly predictable.

But only partly. Even the strongest of these indicators (the money-price ratio) explains less than 30 per cent of the variation in annual equity returns.

>High retail sales predict good returns, perhaps because consumer spending is partly forward-looking and so high spending presages better economic times

Worse still, each of these indicators has sometimes given us a bum steer. Most predicted the market would fall in 1999, but it rose. Most failed to foresee the market's fall in late 2002. The dividend yield largely missed the slump of 2008. Foreign buying of US equities did a good job of warning us of 2008's fall and the euro crisis of 2011-12, but failed to foresee 2015's fall. And while the money-price ratio correctly warned us of 2015's fall, it didn't predict 2016's bounceback.

The fact that some indicators work sometimes poses the question: can we combine them to give better predictability?

Yes. My chart shows how such a combination has predicted returns since 1998. This combination drops the ratio of retail sales to the All-Share index: it contains no information that isn't in the other indicators. And it adds another indicator: the ratio of the All-Share index to its 10-month average. This measure captures what measures of value miss - that sometimes there is momentum in the market so that a cheap market can get cheaper and a dear one dearer.

 

 

This combined indicator has in the past explained over half the subsequent variation in annual changes in the All-Share index. Even so, it has sometimes failed. It didn't warn us that prices would fall in 2002; failed to foresee the rise of 2005-06; and didn't tell us prices would fall in the winter of 2015-16 (it also under-predicted the slump in prices in 2008, but I don't consider this a major failure: the indicator definitely warned us of a big fall and so told us to be out of the market in good time, thanks largely to the message sent by foreign buying of US shares).

This indicator points to the market rising by around 9 per cent in 2017 - although we've already had 2.6 percentage points of this. But the historic error margins imply that there's around a one-in-four chance of the index falling. This is because the slight sell signal sent by the money-price ratio is offset by the slight bullish message of the dividend yield being above average and the index being above its 10-month average.

Such an outlook seems moderate. But this shouldn't be surprising. The most common thing to happen is that equities rise slightly over a 12-month period, so we'd expect any worthwhile predictor to often predict this.

Of course, this rests upon a big assumption - that past statistical relationships will continue to hold, which of course they might not. I'd regard this forecast less as a guide to action and more as a way of testing a hypothesis, that there's some predictability in returns.