JOINT THRESHOLD EXCEEDANCES OF STOCK INDEX RETURNS IN BULL AND BEAR PERIODS.

The investigation of extreme returns on financial assets is motivated, among others, by financial risk assessment. We investigate joint, distributions of threshold excesses of daily returns on several stock indices, threshold meaning the 10%, respectively 90%, quantile of the return distribution. In a first step, the generalized Pareto distribution is fitted to a return series, then the logistic dependence function is used to couple pairs of threshold excess distributions. One of (mr goals is to compare the behaviour of exceedances in bull and bear periods mid relate it to the degree of dependency found in their joint distributions. Since there is a positive dependency among international stock markets, ignoring the dependency among threshold exceedances can be expected to underestimate the risk of high joint losses. It is shown that the degree of underestimation is worse for bear than for bull periods. [ABSTRACT FROM AUTHOR]