Correlation Update

The extreme in finance and economics is by definition rare, and that makes it valuable for study.

The crisis of late is no exception. It’s one thing to analyze markets when everything is running smoothly, but sunny days don’t offer much, if any insight about what to expect during hurricanes.

On the matter of diversification benefits, or lack thereof, one might wonder how the volatility and selling have impacted correlations among the major asset classes. With that in mind, we ran the numbers for trailing 36-month correlations through September 30, 2008, which delivers the chart below.

101308a-thumb.GIF

As usual, correlations are a mixed bag, at least as we calculate them. (Our definition of correlations here is based on the trailing 36 months of monthly total returns. In all cases above, the correlations are in relation to U.S. stocks, as defined by the Russell 3000. The assumption is that investors are looking for opportunities to diversify their equity holdings, which tends to be the dominant risk asset.)

Unsurprisingly, stocks the world over have moved in tighter lockstep recently. In financial panics, all equities look the same, which is to say that investors want now, today, this minute. No wonder, then, that correlations have gone up from already high levels between U.S. stocks (Russell 3000) and stocks in mature foreign markets (MSCI EAFE) and emerging markets (MSCI EM).

The Russell 3000/EAFE correlation as of last month rose to 0.85, the highest in nearly three years. The Russell 3000/EM correlation also climbed, touching 0.73 in September, the highest since December 2006. (1.0 indicates perfect positive correlation, 0.0 is no correlation, and -1.0 is perfect negative correlation.)

The biggest change in terms of posting sharply higher correlations with U.S. stocks comes from high-yield bonds. Just 2-1/2 years ago, the correlation between the Russell 3000 and the iBoxx Liquid High Yield Index was a mere 0.34, meaning that the junk bonds over the previous three years at that point had been exhibiting a large degree of independence from U.S. stocks in terms of monthly performance. But as the chart above illustrates, that independence has faded quite a bit, with the correlation between the two asset classes jumping to 0.85 for the 36 months through last month. That’s virtually identical to the correlation between U.S. and foreign developed market stocks. The reason? Investors have been dumping junk bonds just as vigorously as stocks.

REITs, by contrast, have been going the opposite way. Correlations between U.S. stocks and REITs have been falling recently. The underlying cause: REITs have managed to lose significantly less money than domestic equities in recent months. Last month, for example, the Russell 3000 shed 9.4% while Dow Jones Wilshire REIT index lost a modest 0.4%.

Meanwhile, domestic and foreign bonds, along with commodities, are moving more in line with U.S. stocks of late, courtesy of the non-discriminating urge to sell anything and everything in recent months. Even so, these three asset classes still post low and negative correlations with domestic equities, suggesting that the traditionally potent diversification power of bonds and commodities is still largely intact if somewhat bruised.

Keep in mind that correlations are always in flux. Even so, correlations are a bit more predictable than returns. It’s a safe bet that that correlations between Russell 3000 and MSCI EAFE will remain in the upper range of the chart above, for instance. Bonds, by contrast, are likely to remain in the lower portion. Therein lies the basis for thinking about portfolio design.

Remember, too, that there are many ways to measure correlation in terms of trailing history. We use 36-month rolling correlations on these pages, in part to maintain consistency. The argument for looking at longer periods–5 years, 10 years or even longer–is a good one, since it cuts out much of the short-term noise and so it provides a more robust sample of what the “true” correlations are. In fact, we do just that, and much more, in our proprietary research. Then again, 10-year correlations are a slow-moving animal and so recent history has little impact. In order to see how the tide is shifting, then, we look at shorter term measures, too.

Overall, whenever two asset classes post correlations below 1.0, there’s a benefit to holding both, although the benefit may be relatively slim in pure correlation terms alone. History suggests that we also consider the fundamental nature of asset classes as they relate to one another. Selling (buying) one simply because it’s posting a higher (lower) correlation in recent months compared to another holding is probably short-sighted.

Indeed, there are many reasons why we can choose to hold an asset class, or avoid it. Current dividends, interest rates, expectations about earnings and performance, along with an investor’s risk tolerance are relevant variables too. But correlations are worth watching, if only to stay abreast of extreme moments, which may signal a change in trend is coming. When commodities were posting sharply higher correlations with stocks (relative to the past) back in 2006, for instance, that was a sign of a potential shift in the market relationships, as we suggested at the time.

To be sure, commodities are still valuable for diversifying equity risk. In fact, all the major asset classes always deserve serious consideration when building strategic portfolios, at least initially. Meanwhile, as the chart above reminds, the value of diversification waxes and wanes over time.


Originally published at The Capital Spectator on Oct 13, 2008 and reproduced here with the author’s permission.