In the aftermath of the recent financial crisis, the benefits of diversification have been called into question because of the rise in correlations between various asset classes. The authors argue that the diversification benefits were not achieved because most investors were diversified on an asset class basis rather than a factor-style basis. They examine the effectiveness of factor diversification relative to asset class diversification and conclude that factor diversification is more effective because it produced superior risk-adjusted performance for investors during the financial crisis.

The authors conclude that factor diversification offers a better solution for investors whose portfolio risks are dominated by stock market directionality. Their study expands on previous studies of risk parity investing by exploring the impact on diversification of both static and dynamic factor components. They suggest that investors engaging in factor diversification will benefit from risk reduction because lower correlations between factors relative to asset classes will result in lower drawdowns, better hedging, and better tail performance in bear markets. They observe that a factor-diversified portfolio, compared with an asset class–diversified portfolio, has a significantly higher Sharpe ratio and a lower volatility, which they attribute to a higher average Sharpe ratio among the constituents and lower correlation between the components. Even though dynamic strategies entail higher trading fees, the diversification benefit will still be achieved because the increase in the Sharpe ratio is significant enough to account for the trading cost differential. In addition, adding a nominal allocation of style premium to an asset class–dominated diversified portfolio could improve portfolio performance, risk, and correlation.

This study would be of interest to portfolio managers who question the merits of traditional asset class diversification, especially after the recent financial crisis.

In exploring the effectiveness of dynamic factor diversification relative to asset class diversification, the authors create two portfolios: an asset class–diversified portfolio and a factor-diversified portfolio. The asset class–diversified portfolio comprises an equally weighted allocation to U.S. stocks, non-U.S. developed market equities, global government bonds, global nongovernment bonds, and a basket of emerging market stocks, small-cap stocks, commodities futures, and property. The long–short factor-diversified portfolio includes five equally weighted components, beginning with four style premium components: global value stock style, global momentum stock style, carry style (composed of currency and fixed-income strategies), and trend style (composed of long–short strategies using forwards and futures). The last two styles are representative of liquid macro-asset trading using forwards and futures for equity, fixed income, currencies, and commodities. The fifth and final component of the factor portfolio is U.S. large-cap equity, which serves as a proxy for the equity premium factor. Performance, risk, and correlation statistics, such as the Sharpe ratio, maximum drawdown, return, volatility, and pairwise correlation, are derived for each portfolio with data for the period of 1973–2010.

The authors also document the performance and risk statistics for popular U.S. asset classes (U.S. equity, Treasury bonds, and corporate bonds) and factor premiums (size, value, and momentum) for the period of 1927–2010. Such documentation validates the argument that the factor premiums have been positive for a substantial period of time and might be expected to continue to be so. The argument for factor diversification is based on such expectations.

A lot of questions have been raised about why diversification strategies failed in the recent financial crisis. The authors do a good job of addressing this concern. The methodology, data, and time frame they use appear reasonable. It would be interesting to see how the results of the study would hold up if non-U.S. factors were used.