Bounds for functions of dependent risks (Q854282)

For an \(n\)-variate real function \(\psi\) and a random vector \(X:=(X_1,\dots,X_n)\) the problem of finding the best possible lower bound on the distribution function of \(\psi(X)\) is studied when the marginal distributions of the individual risks \(X_i\) are given and the structure of dependence of \(X\) is partially or completely unknown. The problem is solved when the portfolio is 2-dimensional, the function \(\psi\) is non-decreasing in each coordinate and also if some information on the dependence structure of the portfolio is provided. When no information on the copula of the random vector is given, the authors provide a new bound which is proved to be better than the standard one used in the literature.