Prediction is very difficult, especially if it’s about the future.
In the annals of modern portfolio theory, commodities deserve an award for the biggest diversification head fake of all time.
By 2006, commodities looked like the ideal diversifier. Academic studies demonstrated that commodities had consistently generated equity-like returns over decades, but with basically no correlation to other asset classes. Performance over the preceding decade supported this, and allocators extrapolated these return characteristics out into the future. Mean variance optimizers loved it as the new efficient frontier jumped to the upper left.
Now fast forward to early 2016. Over the past decade, commodities have dropped 10.6% per annum and a dollar invested then is worth 33 cents today. Even worse, correlations to equities spiked to over 0.8 during the financial crisis and its aftermath. A 15% commodity allocation to a traditional stock-bond portfolio would have cost 30-50% of performance. With the benefit of hindsight, the whole diversification premise should have been thrown out the window.
The commodity head fake highlights three interesting issues with Modern Portfolio Theory. First, most models assume that different asset classes will have similar risk-adjusted returns over long periods of time. Sure, an asset class might underperform this year or next, but over ten years we’d expect reversion to the mean. This is grounded in the belief that markets are largely efficient: after all, if one asset class is going to outperform another (on a risk-adjusted basis), rational investors will buy the former and sell the latter until prices adjust accordingly.
The underperformance of commodities seriously undermines this assumption. On one hand, it is almost numerically impossible for commodities to do this badly over the coming decade; on the other, they would have to return over 20% per annum over the coming decade to catch up to the rest of the stock-bond portfolio. This is cold comfort to investors who’ve suffered through the past ten years, and underscores the mathematical near-impossibility of recovering from a nearly 70% drawdown over the first decade of a twenty-year window.
The next ten years are sure to have their own roster of seismic shifts in the global investing landscape.
Second, the future is expected to look a lot like the past. Astute insights about the past thirty years are of little practical value if the world keeps changing. This is the Achilles’ heel of most financial academic research (see our study on the Value factor here). In the two decades covered above, we’ve seen explosive technological change, the rise of China, several bubbles and busts, the introduction and near collapse of the Euro, unprecedented monetary policies, proliferation of ETFs and UCITs funds, sovereign debt crises, exponential growth of alternative investments, etc., etc. The next ten years are sure to have their own roster of seismic shifts in the global investing landscape.
Finally, maybe commodities shouldn’t be viewed as a normal “asset class.” Commodities are more than just financial assets, and there are many players who do things for non-economic reasons, which can distort prices and returns. For instance, some commodity exporting nations appear to have seriously overinvested to expand supply – in part due to cheap global capital – for social and political reasons. Excess capital investment has driven down returns for years and caught financial investors in the undertow.
Investing is a prediction business. When we construct a portfolio, we implicitly make a long list of assumptions about the future. Models are very useful in stress testing those assumptions – e.g. Monte Carlo simulations. But the models are only as good as the inputs, and this is where it gets tricky. The probability that commodities would underperform by this magnitude over a decade is something like 1 in 20 – highly unlikely, but not statistically impossible. However, add in the change in correlation due to fundamental economic reasons, and it’s clear that this outcome was beyond the scope of prediction a decade ago.