When market timing matters

My chart shows why. It shows the annualised total return you’d have had on the All-Share index (after inflation) if you had bought in each month between December 1985 and December 2014.

Imagine you’d bought in September 1987. A month later you’d have lost more than 25 per cent in the October crash. You’d have felt awful. Viewed from 30 years on, however, buying then doesn’t seem so terrible. You’d have made a real return of 4.5 per cent per year since then. Of course, you’d have done better if you’d waited. But 4.5 per cent is a very reasonable return – not least because it compounds very nicely over 30 years. And it’s better than you’d have got if you had bought at any time between 1995 and 2002.

What’s more it would by now have made very little difference whatever time you’d have bought between 1988 and 1992 because annualised returns since any time then have been much the same. All the dips and bounces in the market in those four years are pretty trivial in hindsight.

In the very long run, then, most dips and rallies don’t matter very much. So timing the market isn’t very important.

With two big exceptions.

One is that although most month-to-month fluctuations don’t matter, serious bear markets certainly do. If you’d bought at the peak of the tech bubble in 2000 you’d have made only 2 per cent per year since then. If you’d waited until 2003, you’d have made over 6 per cent. That’s a big difference, especially when compounded.

Secondly, over shorter periods, timing obviously does matter. Look at how the line gets more volatile as the time horizon shortens.

Some simple maths tells us this. Let’s assume real returns average 5 per cent per year with a standard deviation of 15.2 per cent – which is the volatility we’ve seen since 1985. If we then assume that returns are serially independent the chances of them being 3 per cent annualised or less over five years are high, at over one-in-three. But the chances of returns being that low over 20 years are much lower – under 11 per cent. In the long run, the trend dominates the noise. But in the short run – which can last many years – it’s the noise than dominates.

(In fact, over only a few years the noise is so great that we cannot even say what the trend really is, but that’s another story.)

For many of us, then, timing does matter. Which poses the question: can we do anything about it?

There’s a big danger here. If you wait for a correction, you might well miss out on returns as prices rise even higher; irrational markets can always become even more irrational. Worse still, you might throw in the towel and buy just before the correction does occur.

There are two things we can do to help protect us here.

One is to follow the 10-month average rule. If you had sold when prices fell below their 10-month moving average you’d have gotten out of equities in November 2000 and in November 2007. You would therefore have avoided the worst of the two bear markets.

Secondly, watch the dividend yield. There’s a huge correlation between the line on my chart and the dividend yield. For the whole period, it is 0.68. For longer periods – if we consider only returns between 1985 and 2010 – the correlation is a whopping 0.78.

There are two reasons for this.

One is that the dividend yield has been a great predictor of medium-term price changes: a high yield predicts rising prices and a low yield falling prices, especially over two or three years.

The second is that in the long term, most returns come from dividends rather than from price appreciation. Since December 1985 real total returns have been 6.2 per cent per year. Of this, 2.4 percentage points have come from price rises and 3.8 percentage points from dividends.

Herein lies some comfort. Right now, the dividend yield on the All-Share index is slightly above its post-1985 average, and the index is also above its 10-month average. Both tell us to be in the market now. This, of course, will change some time.