If you are a long-term investor investing for retirement or to leave money to your children, there’s one thing you must know: what will be the long-term returns on equities? Which raises a problem: we cannot know this. There are at least four different approaches to the problem, which yield different answers.
One approach is to simply assume that future returns will resemble past ones. Since 1900 UK equities have delivered a total return after inflation of 5.5 per cent a year. Why not assume the future will see something similar?
There’s a good reason not to. For much of the 20th century shares were cheap because investors feared, at various times, war, socialism, high inflation or deep recessions. As these fears receded in the 1980s and 1990s equities staged a huge relief rally. But with fears of such catastrophes now much diminished valuations on equities are higher than they were for most of the 20th century and so future returns should be lower. Indeed, they have been just this since that relief rally; in the last 25 years real returns have been under 4 per cent a year.
Which brings us to a second way of projecting future returns – to look at the dividend yield. It has been a fantastic predictor of longer-term returns. My chart shows the point. It plots the dividend yield on particular dates against real annualised returns on the All-Share index since those dates. So, for example, in December 1999 the dividend yield was 2.1 per cent and real total returns since then have been an annualised 2.3 per cent, while in March 2009 the yield was 5.1 per cent and real returns since then have been 6.8 per cent a year. Variations in the dividend yield alone explain two-thirds of the variation in annualised long-term returns since 1988.
With the dividend yield now 3.3 per cent, this relationship predicts a total real return on equities of 3.9 per cent a year from now on – with most of this coming from dividends rather than capital appreciation.
By happy coincidence a third approach to estimating future returns gives us the same answer. This is to assume that the market is now fairly valued so the dividend yield won’t change from its current 3.3 per cent. In such an event share prices will rise in line with dividends. If dividends grow at the same rate they have in the past 30 years – which is 0.6 per cent a year after inflation – this gives us a real total return of 3.9 per cent a year: a 3.3 per cent yield and 0.6 per cent annual capital gain.
Sadly, however, both these methods have a problem. if investors were to perceive equities as offering lower growth and/or higher risk and if these perceptions were to be proven correct then future returns will be low despite the current message of the dividend yield. To put this another way, past relationships can break down. The pattern in my chart exists only because the dividend yield has been stationary: high yields have led to yields falling and the market doing well, and low yields have led to rising yields and prices falling. But this need not happen in future; the market could suffer a permanent de-rating.
A fourth approach to projecting future returns is more theoretical. It asks: what risk premium on equities over bonds does economic theory predict? One answer, provided by Rajnish Mehra and Edward Prescott in 1985 is: a low one. This says that the risk premium should be simply the product of four things: equity volatility; the correlation between equity returns and our background risks (such as the danger of losing our job or business); the amount of such background risk we face; and our risk aversion. Reasonable estimates of these give us a risk premium of not much more than 2 per cent. Which means that with bonds yielding less than minus 2 per cent in real terms total equity returns could be around zero.
This approach is controversial, however. The risk premium might reasonably be higher than this because investors need compensation for taking on the small risk of a cataclysmic disaster such as financial crisis, prolonged depression or political unrest. If such catastrophes don’t materialise, equities should do OK.
The point here is simply that we don’t know what future long-term returns on equities will be.
It’s tempting to believe that while the short-term is unpredictable equities will behave normally over the long run and so we can look through near-term uncertainty and volatility. But this is not true. We don’t know what future “normality” will be. Instead we face what Richard Bookstaber, the author of The End of Theory, calls radical uncertainty. The future, he says, “is unknown in a deep metaphysical sense.”
Which means that accurate, optimising asset allocation is impossible. If we put our different estimations of long-term returns into a simple asset allocation model – such as that proposed by Robert Merton back in 1969 – it tells us that we should have somewhere between nothing and 80 per cent of our financial wealth in equities. Which is no help.
We cannot, therefore, have any accuracy or confidence whatsoever in building portfolios if we are a long-term investor. We just don’t have enough information about the future to make even rough estimates. The nasty fact about long-term asset allocation is that we do not really know what we are doing – and nor can we.
One solution to this is to not be a long-term investor but instead to make shorter-term tactical asset allocation decisions based upon predictors such as the dividend yield, the 10-month rule or time of year. Even these methods, however, are imperfect: they rely upon statistical relationships which hold only as averages, and which might not even do that, as they could break down in future.
Which leaves us with one other solution, suggested by the late Nobel laureate Herbert Simon. We should, he said, abandon any pretence at optimising and instead aim simply at “satisficing” – simply having portfolios we feel comfortable with, while knowing they might well be second, third (or worse) best. Which is what most of us in fact do.
