Quanting It Up With Gold

More specifially, with gold price changes in a given timeframe. Chris Dillow has compared the frequency of gold price drops since 1979, on a percentage basis, with two tables: one representing the normal distribution and the other, a cubic power-law distribution. He finds that outlier drops are far more frequent than the normal distrbution predicts, but are roughly in line with a cubic power law.

The message is simple. A cubic power law seems to apply to all major assets, including gold, which means that big losses are more likely than bell curve thinking predicts.

But so what? Does it follow that you should hold less gold?

It does, if you believed returns were normally distributed, and if you are loss averse - that is, more fearful than the average investor of big losses.

For others, though, this fact might might be so worrying, for two reasons.

First, it's not just big losses that are more common than a normal distribution predicts. So too are big gains. Indeed, these have actually been even more common than a cubic power law predicts.

Second, probabilities must equal one. A higher chance of a big loss must therefore mean a lower chance of something else. That something is a small loss. A normal distribution says we should have had 1,254 daily losses of one standard deviation or more (1.23 per cent in sterling terms) since January 1979. In fact, we've had just 720.

So, although gold offers a big chance of a big loss, it also offers a bigger chance of a big gain, and a smaller chance of a small loss than a bell curve predicts. Whether this increases or decreases the allure of the metal is a matter of taste.