Kat Questions the Use of Correlation Stats in Comp

Looking at return distributions, Mr. Kat found that normal distributions were indeed rare, and more and more econometric research is treating distributions as such. Just using those distributions in the calculation of a correlation coefficient, such as 0.53, may lead an investor to believe there is a higher correlation than what exists in reality, the paper argues.

Giving examples of situations where correlations can been skewed, Mr. Kat concluded in his research that investors need other methods to measure the relationships between asset classes in extreme markets. “Given the complexity of real-life (return) distributions, however, such methods are likely to be a lot more complicated than our old friend the correlation coefficient,” the report stated.

In previous research, Mr. Kat has pointed out the need for diversification and the need for a larger allocation to hedge funds. He has used correlation figures as part of that argument. In a research study, he found that the median return of a portfolio of 20 hedge funds has the lower standard deviation (a statistical measure of risk), but at the same time has a higher correlation with the S&P 500. Meanwhile for a basket of five hedge funds the standard deviation is slightly higher, but this figure is reached at a lower allocation to hedge funds resulting in less of a correlation to equities.