Beyond Correlation: Extreme Co-movements Between Financial Assets

摘要: This paper investigates the potential for extreme co-movements between financial assets by directly testing underlying dependence structure. In particular, a t-dependence structure, derived from Student t-distribution, is used as proxy to test this extremal behavior. Tests in three different markets (equities, currencies, and commodities) indicate that are statistically significant. Moreover, "correlation-based" Gaussian multivariate Normal distribution, rejected with negligible error probability when tested against alternative. The economic significance of these results illustrated via examples: across G5 equity markets; portfolio value-at-risk calculations; and, pricing credit derivatives. Using likelihood ratio-based methods, we show presence significant asset commodities), well international (G5) markets. addition, ratio indicates structure not supported on basis observed co-movements. consequences several examples.