A recent column from Bloomberg Gadfly discusses increasing correlations of asset classes. The correlation coefficient of periodic returns is a measure of the extent to which these returns move in the same direction. Contrary to a common misconception, a high correlation does not imply that the two assets or classes are identical. Ideally, to diversify a portfolio, long-term correlations among portfolio components should low.
This correlation shift has a major impact on portfolio construction following the Modern Portfolio Theory. The article uses ten-trailing-year correlations of various indices:
While a point-in-time analysis of ten-year correlations between indices is instructive, it is of little practical value to investors. Luckily, Alpholio™ has just introduced a new Multi-Correlation service, which provides an interactive analysis of rolling correlations.
In a typical analysis, monthly returns are used because they are less “noisy” than weekly or daily returns. A span of 36 months (three years) of returns is usually sufficient to approximate the long-term correlation and, at the same time, to nimbly react to rapid correlation changes. A rolling-period approach provides insights on how the correlation coefficient evolved over time. It can also facilitate calculation of useful statistics. Finally, instead of artificial indices that cannot be bought or sold, Alpholio™ uses ETFs.
To demonstrate the Multi-Correlation service in action, here is a chart of rolling correlations between each of several analyzed ETFs and one reference ETF:
In the above chart, the analyzed ETFs are the iShares Core S&P 500 ETF (IVV), iShares MSCI EAFE ETF (EFA), iShares MSCI Emerging Markets ETF (EEM), and iShares Cohen & Steers REIT ETF (ICF). The reference ETF is the iShares S&P GSCI Commodity-Indexed Trust (GSG).
Please note that the youngest of these ETFs (GSG) determines the common time frame of this analysis: the first full month of GSG returns was August 2006, so the first 36-month rolling return became available at the end of July 2009. The Multi-Correlation service determines the longest possible analysis interval automatically.
The following table contains statics of rolling correlations between each analyzed ETF and the reference ETF:
By default, the statistics are ordered in the ascending order of median value, but can be reordered in any ascending/descending order by clicking on the header of the respective column. In conjunction with the chart, these statistics show that each of the equity ETFs had a substantial (above 0.6) correlation to commodities, while the REIT ETF had the lowest correlation. The correlation of all four analyzed ETFs to the reference ETF declined from the second half of 2014 onward, with the REIT ETF showing the strongest decoupling. The Forecast statistic is the expected value of the rolling correlation one month forward, or in February 2016 in this example.
We hope that the Multi-Correlation service will become a useful tool for investors who want to construct well-diversified portfolios. If you would like to use this and other Alpholio™ services, please register on our website.