March 3, 2009
The values are the difference of the average excess downside correlations minus the average excess upside correlations. A positive score is undesirable, as it represents greater correlation on the downside than on the upside. The score for US equities versus world ex-US equities ranks among the worst, at +58%.
Page offers a number of observations. Market neutral funds have been promoted by the hedge fund industry for their lack of correlation to US markets. However, the lack of correlation among these funds occurs in up markets and not in down markets, giving market neutral funds one of the worst diversification scores against US equities. Even though the beta of market neutral funds is theoretically zero, when crises occur they exhibit significant market correlation.
The failure of style diversification is shown by the poor (+51%) score for value versus growth, and the failure of size diversification is evident in the +53% score for large versus small/mid-cap (“smid”) stocks.
US equities and US bonds have a low correlation asymmetry, meaning that the average upside and downside correlations are similar.. This finding is not new and has been reported previously. Page said that fixed income asymmetries have weakened recently, at least partially because of a lack of liquidity in the bond markets. As credit markets have seized, fixed income markets have become increasingly correlated to equity markets.
Decoupling was achievable through mortgage-backed securities, although Page noted that this decoupling disappeared starting in 2007 along with the bursting of the housing bubble.
Even those sectors traditionally hailed as great diversifiers against US equities score poorly by these measures. Commodities, based on the S&P GSCI Commodity Total Return index scored +57.5% (using data from 1979-2009). Real estate was no better. Global REITs, based on FTSE EPRA/NAREIT Global Dollar index, scored +54.4%, and the FTSE EPRA/NAREIT US Dollar index scored +14.4% (using data from 1990-2009). These scores are computed against the Russell 3000 index for US equities.
Full-scale optimization
Page, Kritzman, and Chua say a new approach is called for to construct portfolios to provide true diversification benefits. Traditional optimization, using mean-variance algorithms, incorporates only the average correlation between asset classes, and does not incorporate asymmetries between up and down markets. Mean-variance artificially smoothes risk across up and down markets, and does not account for the shifting of risk dynamics over time.
To illustrate this, consider the asset allocations produced by mean variance techniques under three different sets of assumptions:
Optimal Portfolios |
US |
World Ex-US |
Cash |
Unconditional Correlation |
25% |
72% |
3% |
Down-Down Correlation |
14% |
77% |
9% |
Up-Up Correlation |
32% |
68% |
0% |
Display article as PDF for printing.
Would you like to send this article to a friend?
Remember, if you have a question or comment, send it to .
