SAN FRANCISCO - The conventional wisdom is that a low correlation between asset classes is good. But the negative correlation between stocks and bonds during the past two years has been downright ugly for most pension funds, asserts Frederick E. Dopfel, a managing director and senior strategist with Barclays Global Investors' client advisory group.
In a recent paper - "The Good, the Bad and the Ugly" - published by the San Francisco-based manager, Mr. Dopfel said the stock-bond correlation averaged 0.36 for the 27-year period ended Dec. 31, 2002, then diverged widely in 2001 and 2002, tumbling to -0.4 and -0.7, respectively.
The only other years with negative correlation were 1987 and 1988. Over a 77-year period, the last major run of negatives occurred from 1955 to 1965, when the average correlation was -0.23.
The implications of correlations falling to zero from 0.35 are striking. (Currently, most consultants peg the stock-bond correlation between 0.3 and 0.5 in their asset allocation studies.)
In an asset-only world, if the correlations fell to zero, an investor with a 65% stock/35% bond asset mix would see expected returns stay even and risk drop by 80 basis points. That's the good part.
In an asset-liability world, however, dropping correlations are bad.
The portfolio of assets most closely matching the typical pension plan's risk exposures would be roughly 20% equities and 80% bonds. (Younger plans would have higher equity exposures and more mature plans might have lower stock holdings.)
For an extremely well-funded plan, the result of dropping the correlation to zero is the same as in the asset-only world - an 80-basis-point reduction in risk. But for a plan that is 115% funded, surplus risk rises by 80 basis points. For a 100% funded plan, surplus risk increases by 100 points.
Here's the ugly part: For an 85% funded plan - closer to today's norm - surplus risk rises by 120 basis points, meaning companies would have to raise contributions.
The question remains whether investors think the stock-bond correlation will remain lower than its historic norm. If investors expect the correlation will remain low for a short period and then revert to its normal level, they can take tactical shifts in their asset mix, Mr. Dopfel wrote.
But if pension executives believe the correlation will remain low over the long term, they could strategically reallocate to asset classes with higher estimated correlations to their liabilities, such as TIPS, he wrote.
Mr. Dopfel's main point is straightforward: "Plan sponsors should pay more attention to surplus risk" and to the assumptions used in modeling asset allocation, he said in an interview.