Should we abandon the concept of volatility? That’s the suggestion of Eric Lonergan at M&G. Volatility, he says, is a “fad” which is “a recipe for pseudo-science and over-confidence”.
Before seeing why he might be right, let’s remind ourselves why volatility might be a useful idea. It’s because it’s a simple if rough guide to our chance of losing money.
For example, since 1870 annual real total returns on UK equities have averaged 6.1 per cent with a volatility (or standard deviation) of 19 percentage points. These two facts allow us to estimate the chance of losing money. This would happen if annual returns fall 6.1 percentage points below average. 6.1 is 0.32 of our standard deviation. If we then assume that returns are normally distributed then statistical tables tell us this is a 37.4 per cent chance. Similarly a loss of 10 per cent or more is a 16.1 percentage point shortfall from average, which is a 0.85 standard deviation event, or a 19.8 per cent chance.
We can apply this reasoning to any likely loss, or any probability, on any asset. With another assumption and some trivial maths, we can also apply it to any time period. Volatility, then, is a simple number which can tell us a lot.
Of course, the assumption that returns are normally distributed doesn’t hold in the real world. We know that extreme losses are more common than a normal distribution predicts. But this doesn’t discredit the concept of volatility: we can apply standard deviations to many other probability distributions including the power laws that describe extreme losses.
Nevertheless, we still have the question: why use volatility? Why not simply count the number of losses and use this as a guide to risk? For example, since 1870 there have been 46 years in which shares lost money in real terms. That suggests a 31 per cent chance, not the 37.4 per cent chance predicted by volatility plus the assumption of a normal distribution.
One reason not to do this, of course, is that the distribution of past returns might not be a good guide to future ones. I suspect equity investors had a lot of luck in the 20th century that might not persist this century.
But, of course, the same applies to volatility: past volatility needn’t be a guide to future volatility.
This is especially true if we measure volatility over short periods, simply because it varies so much.
Perhaps the most egregious example of how this can mislead us came in 2007 when David Viniar, then chief financial officer at Goldman Sachs, said of one of his hedge funds “we were seeing things that were 25-standard deviation moves, several days in a row”. Even allowing for extreme returns to be more likely than a normal distribution implies, however, a 25 standard deviation move should be the sort thing we see only around one day every 350 years. If you’re seeing such an event regularly, you’ve horribly mismeasured volatility. In particular, you’ve extrapolated from a period of unusually low volatility.
It’s in this sense that Mr Lonergan is right. Volatility can indeed cause overconfidence if we fail to appreciate that past volatility isn’t a guide to the future.
There’s worse. As Mr Lonergan says, “volatility is not risk”. Let’s take just four examples of this.
- Bonds. The risk that matters with these is that of default – of the borrower not repaying interest or principal. The bond’s volatility does not tell us how great this danger is.
- Liquidity risk. A big reason for the 2008 banking crisis was that mortgage derivatives suddenly became impossible to sell. Their apparently low price volatility in previous months gave no clue to this catastrophe. Retail investors have a similar problem. In bad times, property becomes less liquid – a danger not captured by measures of its volatility.
- Cyclical risk. Housebuilding stocks are not usually especially volatile. But they run the danger of doing horribly in a recession: Taylor Wimpey, for example, lost over 95 per cent during the 2007-08 crisis. Many investors might want to avoid this danger because recessions are times when they need financial wealth to hold up in order to protect them from the risk of losing their job or business. Volatility, however, doesn’t tell us how great this danger is (except to the extent that share price volatility does tend to rise during recessions – although by then it might well be too late).
- Political risk. Utilities have traditionally been low-volatility shares, But this doesn’t mean they are safe. A future government might regulate them more harshly or even nationalise them on unfavourable terms. Whatever the chance of this happening, volatility does not tell us what it is.
So, maybe volatility does have some use as a rough and ready estimate of the chances of particular price moves. It is, however, of only limited value as a measure of risk. Just because volatility is a precise number does not mean it is a good guide to that risk. Precision and truth can be two different things.
