David Rosenberg, chief economist for Gluskin Sheff (Toronto), had an interesting graph the other day showing how well the yield of the U.S. 10-year Treasury bond tracked with the Nikkei 225 index (Japan).
This is an excellent example to show how widely correlations can vary between different time periods. The data has been analyzed year by year since (and including) 2004 through August 26, 2010. The results are displayed in the following graph:
There are several things to note:
1. Even though the correlation over the entire time period is borderline good, there have been years when the correlation has been weak, poor and even negative.
2. The average of the correlations for the component time periods is much different than the correlation over the entire time period.
3. The strongest correlations have occurred in three of the last four years.
4. The strongest correlation has occurred thus far in 2010.
The behavior of this correlation over the last seven years is a good example of why correlations must be tracked if one is depending on them for risk reduction in asset allocation. As many EMH (efficient market hypothesis) and MPT (modern portfolio theory) practitioners can attest, setting up a portfolio based on a pattern of correlation and letting it ride can sometimes produce very unexpected results.
Note: The author’s definitions of correlation quality are defined here.