Asset Correlations and Risk Management

Albert Einstein described compound interest as the ninth wonder of the world. Given his other contributions to science and society, one would probably have a pretty hard time taking the other side of that argument. If, however, they are accepting suggestions for the 10th, I would like to see “correlation” get some consideration.

The pandemic nature of the current economic crunch has raised awareness of just how interrelated everything is and as the world waited to see whether it would survive the various events that have characterized this almost 2 year old crisis of confidence we have witnessed markets long thought to be disassociated moving in lock step.

Mohamed El-Erian, co-head of just about everything at Pimco describes the situation from 30,000 feet:

“’You were increasingly seeing a breakdown’ of perceived relationships between asset classes, ‘and that was way before the latest phase in the markets, which accentuated the problems'.”

To back up his statement, M.E.E. had the boys in the back run some numbers and here is what he found. Between 1991 and 1994 the correlations between the S&P 500 and high-yield bonds was ~0.2-0.3; international stocks ~0.3-0.4; REITs ~0.3 and was negligible in commodities.

By early 2008 those numbers looked like: ~0.7-0.8 for high yield bonds; ~0.7-0.8 for international stocks; ~0.6-0.7 for REITs and slightly negative ~-0.2 to -0.3 for commodities,.

During the late 2008 meltdown those numbers got even higher with the S&P losing 37%, the MSCI major European, Asian and Australian index down 45%, emerging markets down 55%, REITs down 37% and the Dow Jones AIG Commodities dropping 37%.

Pimco’s head of analytics, Vineer Bhansali, offers an interesting insight here as a result of his analysis of the inter-market relationships saying that “when people start buying an asset, the act of them diversifying ultimately makes that asset less of a diversifier."

Ibbotson Associates, a leading authority on asset allocation, has looked at the numbers as well and come up with its own set of fun facts. From the period since 1973 when U.S. stocks rose more than 6% in a single month, European stocks rose 4.5% and Japanese stocks by 3%. When the U.S. markets fell more than 4.5% in a month, European stocks would fall 3.87% and Japanese stocks 2.97%. According to Ibbotson, REITs tend to go up about half as much as stocks in the good months but down about 2/3rds as much in the bad months.

As a result of his work, Michele Gambera, Ibbotson’s chief economist, says “There have been reasons to question diversification, no doubt about that. We are learning that there are a lot of implicit stock-market bets in a lot of asset classes.”

While the fact that the relationship between U.S. equities and some of the other asset classes has grown closer appears to be surprising to some in the industry, the negative correlation between CDS levels and equity prices has been examined and documented for quite some time now.

I have been watching this relationship first hand since late 2003 and it is this relationship upon which the CEC Strategy is based. If there has been a change in the CDS/equity relationship it would be that it has moved from the credit spreads being somewhat of a leading indicator of stock price movement to more of a coincident indicator. This has not affected the performance of the strategy as risk management procedures established early on required confirmation across the relationship to become evident before positions were established.

Enjoy the week.

Comments

[Living4Dividends]

Nice article - thanks - I like anything that comes from Ibbotson

2009 Jul 14 01:39 PM

[Carneades]

This article had what many other articles about correlation lack: nuance. There is an irrational tendency to suggest that the concept of correlation has been proven wrong or otherwise irrelevant by the events of the past two years. A more rigorous assessment would suggest this is not the case.

2009 Jul 14 11:25 PM

[DC Financial]

This article cites some important and common misunderstandings about correlation between assets, and asset classes. It is important for investors to understand a correlation's coefficient is not a stagnant number.

Correlation is a backward-looking number representing how two assets (or classes of assets) have BEEN correlated in the past. The first mistake most will make is thinking they are getting a larger statistical sample size when taking a longer time-line into their calculations, and that this larger sample size will accomplish the same thing as it would in other areas of statistics (though one could argue about sample size in general, but that is for another day). In reality, when taking a larger sample size, one is simply exchanging one backward-looking number for another. The first one being reprsentative of its own time, and the second the same of its time. Both numbers not neccessisarily being true of the current time.

Another problem with correlation, as this article shows, is that during hard economic times, correlations rush to 1--which defeats (or at least partially) defeasts the whole point of correlation, as during good times, one is sacrificing return with correlation.

2009 Jul 15 10:08 AM