We all know that the dividend yield on the FTSE All-Share index can predict subsequent returns, especially over longer periods. Valuations matter, that’s common sense. There is, though, another measure that has in the past predicted returns just as well.
My chart shows how this alternative measure has predicted annual changes in the All-Share index since 1997. It told us that the market was overvalued in the late 1990s, in 2007 and in early 2015 and on all three occasions shares subsequently fell. It also told us shares were cheap in 2003, 2009 and 2012 and the market subsequently rose.
Since 1997 the correlation between this measure and subsequent annual changes in the All-Share index has been 0.46 – the same as the correlation between the dividend yield and subsequent changes. For three-yearly changes, this measure has been a slightly better predictor than the dividend yield. And in fact it works as well in predicting annual returns as two other of my favourite lead indicators: foreign buying of US equities and the ratio of retail sales to the All-Share index.
So what is this measure? It’s simply time. My measure is the deviation of the index from a time trend of an increase of 3.5 per cent a year. If the index rises more than this (cumulatively) it becomes ‘overvalued’ and subsequently falls, and if it rises less it becomes cheap and so will rise eventually.
Time alone, therefore, has predicted equity returns.
This isn’t as absurd as it seems. It’s just another way of saying that markets sometimes overreact. Sometimes they rise too much and become expensive and sometimes they fall too much and become cheap – which is contrary to what the rational markets hypothesis predicts. The only question is how we measure what is ‘too much’. Common sense says we can do by comparing prices to dividends. But we can equally well do so by comparing them to a time trend.
In fact if we try to combine the dividend and this time trend, we find that the yield is statistically insignificant. This implies that all the work in the dividend yield is being done by prices rather than dividends. Dividends themselves tell us nothing about future price moves that a time trend doesn’t.
The time trend is, however, significant when we combine it with foreign buying of US equities and the ratio of retail sales to the All-Share index.
This is consistent with there being three types of lead indicator of returns. One asks: is the market over or underpriced? The time trend and dividend yield tell us this.
A second measures investor sentiment: this is what foreign buying of US equities does. Other indicators, such as discounts on investment trusts or the level of Alternative Investment Markets (Aim) share prices, do a similar job.
And a third indicator aggregates the dispersed and fragmentary hunches about the future which millions of people have. This is job done by the ratio of retail sales to the All-Share index and by the US and UK yield curves: if we take them all together, both consumers and bond investors sometimes know something about the future.
In theory, however, our time trend need not always be a good lead indicator. If the long-term growth rate of dividends were to rise, shares could rise by more than 3.5 per cent without becoming overvalued. Instead, high prices would be vindicated by future growth. Conversely, if investors were to correctly anticipate lower dividend growth they’d sell equities and our measure would make them seem cheap. But this would not mean they would rise. Low prices would instead be justified by subsequent low growth.
Oddly, though, this hasn’t happened much in the past. Whenever investors have expected high growth – as they did in the late 1990s – they have been disappointed and prices have subsequently fallen. And when they’ve feared long depressions, as in 2008-2009, they’ve been too pessimistic and so prices have risen.
Just because something has not happened before, however, does not mean it cannot ever happen. A change in trend growth would cause the time trend to fail as a predictor.
Right now, this measure is sending no signal. This is partly by the measure’s very construction: I chose a3.5 per cent trend because it’s close to the average rise in prices since the mid-1990s. But I think it’s plausible that the market’s rise so far this year has merely corrected last year’s underpricing.
Many other lead indicators, though, are still bullish: such as the dividend yield, foreign selling of US equities and the retail sales/All-Share index ratio. What we have to fear (other than short-term dips) is either that the US’s inverted yield curve proves correct again or that the old adage ‘sell in May’ proves to be.
