Increasing correlations or just fat tails?

Increasing correlation during turbulent market conditions implies a reduction in portfolio diversification benefits. We investigate the robustness of recent empirical results that indicate a breakdown in the correlation structure by deriving theoretical truncated and exceedance correlations using alternative distributional assumptions. Analytical results show that the increase in conditional correlation could be a result of assuming conditional normality for the return distribution. When assuming a popular alternative distribution - the bivariate Student-t(r) - we find significantly less support for an increase in conditional correlation and conclude that this is due to the presence of fat tails when assuming normality in the return distribution. (C) 2007 Elsevier B.V. All rights reserved.