TY - JOUR T1 - The Dangers of Using Correlation to Measure Dependence JF - The Journal of Alternative Investments SP - 54 LP - 58 DO - 10.3905/jai.2003.319091 VL - 6 IS - 2 AU - Harry M. Kat Y1 - 2003/09/30 UR - https://pm-research.com/content/6/2/54.abstract N2 - Correlation is without doubt the single most important parameter in modern portfolio theory, where it is used to measure the dependence between the returns on different assets or asset classes. However, the returns on most assets and asset classes are not exactly normally distributed and tend to exhibit a relatively high probability of a large loss (known formally as “negative skewness”) and/or a relatively high probability of extreme outcomes (known as “excess kurtosis”). In such cases, correlation is not a good measure of dependence and may actually be seriously misleading. Although it appears quite surprising at first sight, at least part of the finding that the correlation between hedge fund returns and stock market returns is higher in down than in up markets for example can be attributed purely to technicalities. A normal distribution with a constant correlation coefficient will exhibit this sort of behavior. In this Academic's Corner, the author illustrates these issues in detail. ER -