Join our community of smart investors

An uncertain retirement

We can't know what medium-term returns will be, so we can't know what is a safe rate at which to run down our wealth in retirement
August 2, 2017

How much can we safely spend in our retirement? The answer is more uncertain than you might think, because we can’t be at all sure what medium-term returns on our wealth we be, and so we cannot know for sure what a safe drawdown rate will be.

There are (at least) three different ways of thinking about medium-term equity returns, all of which yield different conclusions.

One starts from the assumption that the market is now fairly valued. If this is the case, then the dividend yield on the All-Share index should not move systematically from its current 3.6 per cent. If we then assume that dividends rise in line with GDP – that is, that there will be no systematic change in the share of profits in GDP or payout ratio – it follows that share prices should rise in line with GDP. We might expect this to growth at around 1.5 per cent per year on average in real terms. All this implies a total real average annual return of just over 5 per cent: 1.5 per cent(ish) capital appreciation and 3.6 percentage points of yield.

A second method is more pessimistic. It adds an equity risk premium to real bond yields. Figures from Credit Suisse show that equities have outperformed bonds by an average of 3.7 percentage points a year since 1900. Adding this premium to current real bond yields of minus 1.6 per cent gives us expected real returns of 2.1 per cent per year.

A third approach would use a lower equity premium. Back in 1985, Rajnish Mehra and Edward Prescott pointed out that conventional economic theory predicted a premium of less than a percentage point per year – and actual returns since then haven’t been far from this. This would point to negative real returns on equities on average.

You might think we can test these theories against the historical evidence. We can’t.

Since 1900, real total returns on UK equities have averaged 5.7 per cent per year. This seems to support our first theory. But it doesn’t.

It could be that past returns overstate true returns simply because of sampling error. Maybe we’ve had an unusual number of good years since 1900, in the same way that even over a number of draws, some lottery balls are more likely than others to come out.

Even over 116 years, luck doesn’t even out. This is because equity volatility is so large. Since 1900 the standard deviation of real returns has been 21 percentage points. This implies that there’s 1.9 percentage points of standard error around that 5.7 per cent average return. Which implies that there’s around a one-in-six chance that actual average returns are less than 3.8 per cent per year, and about a 3 per cent chance that they are under 2 per cent.

Worse still, there’s very good reason to suspect that past returns were due in part to good luck that won’t recur.

For much of the 20th century, we faced the chance of disaster: military defeat, nuclear annihilation or revolution. In many stock markets, these chances actually materialised: Will Goetzmann and Philippe Jorion say that of 24 national stock markets that existed in 1931, 10 suffered a long-term closure. Because such calamities did not strike the UK, shares staged some big relief rallies. What’s more, the 1945-73 period saw fantastic economic growth that doesn’t look like returning.

On both counts, future equity returns might well be lower than in the past.

Which brings us to the problem. Uncertainty about average returns interacts with ordinary volatility to leave us in the dark about the future.

Let’s take the best-case scenario, that history repeats itself so that average real returns are 5.7 per cent a year. This implies that there’s only around a one-in-seven chance of us losing on shares in real terms over 10 years, and only around a 5 per cent chance of a loss of 30 per cent or more.

If, however, we take the pessimistic view that real returns will be slightly negative on average then there’s a more than one-in-three chance of losing 30 per cent or more over the next 10 years.

In fact, these calculations underestimate uncertainty. I’ve assumed that returns are normally distributed (which is reasonable for the moderate losses I’m considering) and that they are serially independent so that losses one year don’t affect the chances of losses the next. Uncertainty around these assumptions adds to uncertainty around medium-term returns.

What can we do about such uncertainties?

For those of us still working, one answer is to postpone retirement and/or save more. Those in retirement might try to live modestly, or abandon hope of leaving a big bequest: in effect, you can transfer uncertainty about returns onto your children.

For me, this is also a case for holding cash despite what is for now at least a negative real return. Our downside on cash is limited to the real interest rate; the downside to equities is not so bounded. Of course, a big cash holding means you will run down your wealth in retirement. But you’ll do so slowly, and at a more predictable rate than you would if you had an equity-heavy portfolio.

There is, though, another implication here. This shows that rational retirement planning is pretty much impossible even for those people who have the discipline to save a lot and invest wisely. For me, this is a strong argument for a high state pension.