Weighted Average Correlation Matrices Method for Correlation Stress Testing and Sensitivity Analysis

Stress testing entails pushing risk parameters toward more extreme levels and exploring the impact on portfolio value. But it is not trivial to perturb a correlation matrix while maintaining its necessary properties. All entries on the diagonal must be 1.0; all off-diagonal entries must lie in the interval [–1.0,1.0]; and the matrix has to be symmetric and positive (semi) definite. In this article, Numpacharoen presents a remarkably simple approach for modifying a correlation matrix that maintains its required properties. He proves that the weighted average of two proper correlation matrices will also be a proper correlation matrix, which points the way toward easy procedures for stress testing all, or a subset, of the correlations among a set of securities.