Erik Eyster at the LSE and Georg Weizsacker at Humboldt University in Berlin ran a series of experiments in which they offered subjects a choice between two portfolios: one that had a high weighting in a risky asset and a low weighting in a safer asset which partly hedged that risk; and another with a more even weighting between the two which took better advantage of the hedge. They found that almost all subjects preferred the high weighting in the risky asset.
Defensives beat the market
In itself, this might show just a preference for risk. However, they then offered their subjects a choice of two portfolios, one of whose payoffs replicated the high-weighting in the risky asset, the other of which replicated the more balanced portfolio. Given this choice, as many subjects chose the former as the latter. Many people's choices were therefore obviously inconsistent.
This tells us that people don't appreciate the value of low or negative correlations; this caused them to miss the chance to diversify in the first choice. "People tend to ignore correlations," conclude Messrs Eyster and Weizsacker.
What's more, when people are prompted to think about correlations, they do so the wrong way. Michael Ungehauer and Martin Weber at the University of Mannheim show that people estimate correlations simply by looking at the frequency of co-movements, which causes them to underrate the importance of assets that move in opposite directions when markets fall a lot.
All this is no mere quirk of a laboratory experiment. In another new paper, economists at AQR Capital Management show that in stock markets around the world you can earn high risk-adjusted returns by buying stocks that have low correlations with the market: betting against correlation pays off. This suggests that investors underrate the virtue of spreading equity risk which leads low-correlation stocks, or defensives, to be underpriced.
Correlation neglect causes people to make two errors.
One is to fail to diversify sufficiently among assets with low correlations. For example, holders of risky assets such as emerging markets equities often pass up the chance to mitigate that risk by holding bonds.
The other is to over-diversify when assets are highly correlated. For example, investors hold expensive actively managed funds alongside investment trusts that hold similar assets and so are highly correlated with those funds. This leads people to incur unnecessary charges.
We should, therefore, pay more attention to correlations. But how?
You don't need to do fancy maths or go to the bother of getting data on past correlations. In fact, doing so might actually be misleading because correlations can change.
Instead, you can do so informally. Think of the future as being a set of possible states of the world: economic growth or recession; high or low inflation; investors being scared or risk-tolerant; and so on. Then ask: in which of these states would my portfolio do badly? And: are there any assets that might do well in such states?
So, for example, if you hold lots of emerging markets shares you would suffer in a world in which investors become reluctant to take risk. But in such a world, government bonds might well do well. Or if you have lots of UK property you'd suffer in a UK recession but such a world would probably see sterling fall, so foreign currency would spread your risk.
The same thinking warns us of the limits of diversifying among equities. Most of these co-move to some extent over time with the general market and so would suffer if investors become loath to take market risk. This is especially the case for funds: as a matter of brute maths, baskets of stocks reduce exposure to shares' idiosyncratic risks but in doing so increase exposure to market risk.
You don't need detailed maths to diversify. For many of us, an intuitive awareness of correlations - coupled with a realisation that we know pitifully little about the future - would be an improvement.
