A lot of folks are scratching their heads about the recent high correlation between stocks and oil, and it has inspired some high profile musing. Ben Bernanke posits that both stocks and oil are reacting to slowing global growth. Howard Marks recently told my Bloomberg<GO> colleagues on television that any correlation between stocks and oil is proof that investors simply don’t understand the relationship between the two.
But the focus on the recent waltz of stocks and oil misses a larger concern: a multi-decade shift in correlations across risk assets that threatens to crack the foundation of portfolio theory.
In the abstract, a high correlation between stocks and oil -- and commodities generally -- should be the exception, not the rule. Lower commodity prices translate into lower costs and higher earnings for all but commodity producers, and higher earnings translate into higher stock prices. An obvious exception is a slow-growth environment that threatens both commodity prices and sales, but that would be cyclical by definition and any resulting spike in correlation between stocks and commodities should be fleeting.
At first blush, the long-term correlation between stocks and commodities appears to confirm that the two are not highly correlated. The correlation between the S&P 500 Index and the S&P GSCI Total Return Index has been 0.17 since 1988 (the first year for which data is available for all indexes referenced in this column). A closer look, however, reveals a different reality: The correlation between stocks and commodities has trended higher for two decades, and the correlation today would be unrecognizable to an observer two decades ago.
You can visualize this affinity by looking at changes in the ten-year trailing correlation between stocks and commodities, which has shifted from a negative 0.2 in 1997 to 0.5 today. This move of 0.7 in the correlation over two decades is not consistent with the theory -- however intuitively appealing -- that stocks and commodities don't dance in lockstep except on rare occasions.
These heightened correlations aren't limited to U.S. stocks and commodities. The same upward trend can be seen in correlations across risk assets. Just take a look at the ten-year trailing correlations in 1997 for the asset classes in the table below.
And then have a look at the same ten-year trailing correlations today.
In every case the correlations are meaningfully higher today. And in every case the trend higher has been two decades in the making.
So what does all of this mean and what does it have to do with portfolio theory?
Ever since Harry Markowitz introduced Modern Portfolio Theory (MPT) in the 1950s, investors have relied on correlations to build diversified portfolios of uncorrelated -- or less than highly correlated -- assets in order to minimize volatility and diversify risk. The implicit assumption in the model, of course, is that correlations are stable, and that historical correlations are indicative of future correlations.
I doubt that anyone today still believes that correlations are entirely stable.
The potential for chaos was laid bare during the financial crisis in 2008 when correlations famously spiked across risk assets and diversification consequently failed to insulate portfolios when it was most needed.
That experience popularized the notion that correlations break down during periods of market stress, and some investors have modified Markowitz’s original MPT by creating two sets of correlations – one for sunny days and another for rainy days.
But even a modified approach to MPT may not be enough. Just as with stocks and commodities, the upward drift of correlations across other risk assets for two decades belies the notion that correlations are stable except during occasional market storms. Affinities have heightened even in calmer market moments.
A shrewd and mature takeaway from this is that the future course of correlations will be anything but predictable -- and it’s not clear that MPT will help market participants find a roadmap for navigating that reality.
This doesn't mean that diversification is useless. You don’t need fancy models to see that owning more assets is safer than owning fewer, even if those assets don’t zig and zag the way you hoped. Nor does it mean that valuations are doomed to uniformity across risk assets. Indeed, despite rising correlations, the current environment offers a variety of valuation plays.
But it does mean that it’s time to rethink correlations when examining markets and making investment decisions. Or you can hope for a return to past patterns and brace yourself for the consequences of that kind of magical thinking.
This column does not necessarily reflect the opinion of Bloomberg LP and its owners.
To contact the author of this story:
Nir Kaissar in New York at firstname.lastname@example.org
To contact the editor responsible for this story:
Timothy L. O'Brien at email@example.com