Investing in commodities (oil, gas, metals etc.) has gone from being a niche activity to a mainstream portfolio choice in the last fifteen years. This was driven by the long period of fast growth of commodity prices, which made a number of people very rich. The wider argument for commodities investment though has been that commodity prices are uncorrelated or even negatively correlated with equity returns. If true, this is a strong argument for including some commodity exposure in an institutional or personal portfolio. A recent paper from the Bank of International Settlements suggests it is not.
Put your eggs in lots of baskets
Diversification is the only free lunch in investing (*). Combining assets with the same return but less than perfect correlation of returns gives you a portfolio that is better than holding only one asset. When teaching this, we often start with the extreme case of inverse correlation, such as ice creams and umbrellas. Assuming that ice cream sales boom when it’s sunny and collapse in the rain and umbrella sales do the reverse, it’s easy to see that holding a portfolio of both gives you a steady return. You get profits whatever the weather and so reduce the variation in returns, which we assume all investors are keen to avoid (this is the meaning of risk aversion).
But the argument is much more powerful than this. So long as any asset is less than perfectly correlated with an existing portfolio, then its inclusion will add something to that portfolio by improving the trade off achieved between return and risk (measured by the standard deviation or volatility of those returns). This arises from the way that correlation works. It’s a non-linear function so any imperfectly correlated positive return adds something useful to a portfolio.
The perfectly diversified portfolio is therefore to hold a bit of everything, every possible asset. Many assets aren’t traded so in practice institutional portfolios were normally limited to public equities and bonds. The ideal public equity portfolio in this theory is the whole market, or equivalently the market index, assuming it’s correctly calculated and can be cheaply bought. Since the advent of low cost passive funds and ETFs, it is quite easy and cheap to get full diversification for most major equity markets.
The case for commodities
A case can be made that commodity prices are not just imperfectly correlated but negatively correlated with equity returns. Share prices should reflect the discounted value of future dividends of the companies that issued them. A rise in commodity prices would tend to decrease profits and future dividends, and so reduce share prices, depending on the intensity of use of the commodities in the particular company production process. But since virtually all companies use some commodities in their cost structures (lighting, heating and power supplies) then this seems plausible. So including commodities in your portfolio ought to improve its efficiency – the risk/return combination you can achieve.
That argument rests on commodity prices being driven by exogenous (external and unrelated) supply factors such as a new discovery of oil or a politically-motivated restriction on copper exports.
But what if commodity prices are driven primarily by changes in demand instead? This is what has happened in the last decade. The sharp increases in many commodity prices, especially oil, up until recently were mainly a result of strong demand from the rapidly growing emerging economies, chiefly China. But high economic demand would tend to boost share prices too. So the negative correlation might turn positive.
The BIS analysis shows that the long standing negative correlation that used to exist in the period up until the early 2000s turned closer to zero in the 2000s and then went positive after the financial crisis. The increased flow of funds by investors into commodities may have itself increased the correlation but the main cause seems to be the less tight supply markets and the consequent greater importance of global demand for both equity markets and commodity prices.
It is an awkward fact about economic data that the underlying relationships between them are not constant (they are “non-stationary“). Empirical economics is not therefore about identifying the unchanging “true” structure of the economy, because that structure evolves over time and may even change in a way that reflects our knowledge of it (“reflexivity“). The BIS authors try to cope with this by using a Dynamic Conditional Correlation approach (DCC), which was proposed by the Nobel winning econometrician Robert Engle. DCC is a multivariate GARCH model in which correlations are time-varying according to an autoregressive specification. That may not mean much to non-econometricians but the point is that it captures the changes in correlation over time, unlike a static correlation measure.
The authors also confirm that there is a clear break in the trend at the time of the Lehman bankruptcy in 2008. Since then, common shocks (news, whether good or bad, about prospects for the global economy) have driven both commodities and equities. This largely undermines the diversification benefit of holding commodities in a portfolio.
Beware of hidden correlation
True diversification is hard to find, partly because so many economic activities are related or have common causes. One of the best investment decisions I made was to buy wine (for portfolio, not for drinking, though the backstop of consumption puts a lower bound under the returns). But the underlying cause of rising prices of good wine, especially the fancy French brands, is rising global consumption by newly rich people in emerging economies. So that is closely correlated with emerging market GDP growth.
But it is also well known (or should be) that equity returns are very poorly correlated with a country’s GDP growth. The atrocious performance of my Chinese shares is a constant reminder to me of that.
(*) This isn’t true actually, because once everybody else diversifies their portfolios the market return on any asset should no longer give any benefit for diversification – it’s priced into market returns. You should only expect to earn a return above the risk free rate for taking non-diversifiable risk, which is the underlying systematic risk of the economy which can’t be diversified away. To the extent that different economies are less than perfectly correlated you can diversify geographically but that return has also largely been priced in, now that it is relatively easy to hold global portfolios. And the economies have become more correlated in any case owing to globalisation.
Another academic research report with similar findings is available here.
A case can