Take, for example, a portfolio comprising 50 per cent in the All-Share, 40 per cent in gilts and 10 per cent in gold, rebalanced every month; this is an idealised example, which I use merely to highlight the benefits of diversification. (In ignoring corporate bonds, I'm not losing anything, as these are, in effect, a mix of gilts and equities with low-grade bonds being more equity-like and higher grade ones more gilt-like.) Since January 1986, this has given an annual return of 9.5 per cent with a standard deviation of 8.9 percentage points. The All-Share has given us 10 per cent returns with a 15.9 percentage point standard deviation in this time. So our balanced portfolio has given us similar returns to shares but with much less risk.
This has been achieved in two ways. First, investors who diversified out of equities did not lose much. Since 1986, gilts have returned only 1.2 percentage points a year less than shares and gold only five percentage points less; including the late 80s and 90s doesn't flatter the metal.
Secondly, correlations between assets have been low. Since 1986, the correlation between monthly returns on equities and gilts has been zero, and that between equities and gold has been 0.2. These imply that, quite often, losses on shares have been offset by gains on other assets.
However, both of these forces favouring the diversified investor might weaken in coming years.
We can be pretty sure that gilt yields won't fall by eight percentage points in the next few years, as they have since 1986 - and, indeed, there are good reasons to expect them to rise. And while gold is largely unpredictable in the short term, Hotelling's rule tells us that the least bad assumption to make is that its price will rise in line with current interest rates - which implies very low returns; Harvard University's Robert Barro has found that this rule has predicted gold's long-run returns quite well.
If we assume that the equity premium over gilts will be the same in the future as it has been since 1900 - 3.7 percentage points, according to Credit Suisse - then we have a doubly worrying implication. First, returns on a balanced portfolio will be lower in future than they have been since the mid-80s. And, secondly, because equities will outperform gilts and gold by more than they have recently, the opportunity cost of diversifying out of equities will be higher; we'll have to sacrifice more returns in order to hold asserts that diversify equity risk.
This sounds bad enough. But there's worse. Correlations between assets might be higher in future than they have been recently.
It would be easy to tell scary stories about inflation or debt crises which would hurt both gilts and equities. But there are less dramatic ways in which the correlations could be higher.
Most obviously, an end to quantitative easing (QE) - or a reversal of it - could be bad for all three main assets. Insofar as cheap and easy money has inflated all three, a reversal of QE would hurt them all. Luckily, central banks will only withdraw QE when economies are stronger - which should mean that share prices will be higher. But it's possible that uncertainty about whether economies can cope without monetary support would hurt equities while the lack of that support will hurt gilts and gold.
A second source of higher correlations would be a return of duration risk. Equities and long-dated bonds have one thing in common; both offer cash flows in the distant future. This means that if investors were to reduce demand for long-duration assets, both would fall together. The fact that we haven't seen this recently doesn't mean it can't happen. Before the mid-90s, it was common for equities and gilts to rise and fall together as duration risk aversion fell and rose. If or when interest rates rise - thus reducing the net present value of future cash flows - we might see a return of varying duration risk.
A further danger comes from China. It's widely agreed that, over time, China's economy will slow down and be driven more by household spending and less by exports. This could be bad for bonds, to the extent that lower Chinese savings would reduce the global savings glut that has forced bond prices up. But it could also worry equities partly as slower Chinese growth reduces the growth of demand for commodities and so hurts mining stocks, but also because a new economic paradigm would create uncertainty.
Now, I'm not saying all of this will happen for sure. I'm just saying that it is possible. Correlations and the gains from diversification are not fixed parameters. They are products of economic and market conditions and so can change. Such changes would be a nasty shock to investors who have been used to bonds and equities both offering good returns or at least being negatively correlated.
So, what could we do about this? If the 1970s is any guide - the last time bonds and equities seriously fell together - we will see an increased demand for physical assets such as housing, commodities or collectibles. These, however, carry three dangers. Some (most obviously housing) are expensive now; we might not get the inflation that benefited them in the 1970s; and they carry liquidity risk - the danger that we'll not be able to sell them quickly if we need to.
Instead, there's a duller and less lucrative possibility - to hold onto cash even though it offers nugatory real returns. It's trivially true that cash protects us from the risk of asset prices falling. But it also protects us from correlation risk - the danger that previously uncorrelated assets will become correlated. This risk has been forgotten in recent years. But it shouldn't be.
