Beware of Changing Correlations!

March 09, 2012 | Greg Leonberger, FSA, EA, MAAA, FCA, Director of Research, Managing Partner

This week’s chart shows the dynamic nature of correlations between asset classes by comparing correlations amongst traditional asset classes over 20-year¹ and 5-year² historical periods.³ The chart above shows how much these correlations have all increased when comparing the 5-year figures to the longer dated 20-year period. What does this mean for investors? We see two main takeaways:

1. For those that rely on mean-variance optimization programs for determining their asset allocation, it is imperative to understand the exact timeframe reflected in the correlations used as inputs, as different time periods will yield not only different correlations but critically, different portfolio structures.
2. The correlation amongst traditional asset classes has increased in the last five years, thus it is more difficult to truly create a “diversified” portfolio that offers protection from large draw downs in the equity markets. This was never more apparent than during the financial crisis of 2008-2009.

As now outlined in both this chart and “Correlation Doesn’t Tell the Whole Story”, correlations can be helpful in conducting asset allocation studies, but they also feature some notable shortcomings that should be well understood by those who rely on them for portfolio decisions.

¹ March 1992 – February 2012
² March 2007 – February 2012
³ Indices used for analysis were Russell 1000, Russell 2000, MSCI EAFE, MSCI Emerging Markets, and BarCap Aggregate