Peaks-over-threshold stability of multivariate generalized Pareto distributions (Q2482625)

It is well-known that the family of univariate generalized Pareto distributions (GPD) is characterized by their peaks-over-threshold stability. This paper extends this result to multivariate GPDs. As an immediate application one obtains, for example, that the excess distribution of a linear portfolio \(\sum_{i\leq d}a_iU_i\) with risks \(U_i\) and positive weights \(a_i\), \(i\leq d\), is independent of the weights, if \((U_1,\dots,U_d)\) follows a multivariate GPD with identical univariate polynomial or Pareto margins. This implies, for instance, that the expected shortfall as a measure of risk fails in this case.