It's widely thought that most investors should hold some gold because its low correlations with equities and bonds mean it is a nice way of spreading risk. Such a view, however, runs into a problem. History shows that uncorrelated assets can sometimes suddenly become highly correlated and so once-safe portfolios can quickly become dangerous; this disaster hit both Long-Term Capital Management in 1998 and US banks' mortgage-backed securities in 2008. This poses the question: could a similar thing happen to gold?
Recent experience suggests not. Looking at the last five years of monthly returns, the correlation between the All-Share index and gold (in sterling) has been minus 0.15; that between the All-Share and gilts has been minus 0.24; that between gold and gilts has been plus 0.29.
These numbers are what you'd expect if markets are driven by simple 'risk on' or 'risk off' trades. When risk is on, investors pile into shares and out of gold and gilts. And when risk is off, they do the opposite. Shares then move in the opposite direction to gold and gilts, while gold and gilts move together – to some extent – because both are in their different ways safe-haven assets.
In such markets, gold is a useful portfolio diversifier.
However, asset prices haven't always been driven merely by 'risk on' and 'risk off' trades. For much of the 1990s, correlations between all three were positive, albeit often only slightly so. Combined with its high intrinsic volatility, this meant that gold was a less useful diversifier.
That experience warns us that low or negative correlations need not be permanent. There are – at least – two ways in which all three might again become positively correlated and, worse still, all fall together.
One would be if – or when – global monetary policy tightens. This would see bonds sell off as the Federal Reserve stops its 'operation twist' and possibly if the Bank of England reverses its quantitative easing and sells gilts. Bond prices would also fall (yields rise) as investors anticipate higher short-term interest rates, and because the same better economic conditions that prompt monetary tightening also cause investors to dump safe assets. Such tightening would also hurt gold, to the extent that its rise has been driven by cheap and plentiful money which would end. Whether it's bad for shares is, however, more doubtful; tighter money is bad, but the brighter economic outlook that causes it is good.
But how likely is such a scenario? One thing that speaks against it is that quantitative easing has not caused positive correlations between gold, equities and gilts. This, though, might be a happy accident. All three have risen since 2009, but just not in the same months, so monthly correlations have been low. It's quite possible that slight differences in timing might generate positive correlations, with all three doing badly at the same time.
A second way in which all three might fall together would be if China's export-led growth were to significantly slow down. This would hurt gold to the extent that its price depends upon Asian demand. It would hurt bonds because the counterpart of China's large trade surplus has been hefty buying of western bonds, so a lower trade surplus might mean less buying of bonds. And it could also hurt equities, to the extent that global growth slows as China does, or to the extent that a change in our paradigm for that growth adds to economic uncertainty.
Now, it's unclear how likely or immediate either of these scenarios is. But in one sense, this is not the point. The point is that correlations are not fixed data but are instead the products of varying macroeconomic conditions. It is therefore dangerous to assume that past, benign correlations will continue. A good justification for holding cash, despite its poor returns, is that it protects us from the risk that correlations will turn nasty.