Preservation of the location independent risk order under convolution (Q2492183)

Let \(\leq_{\text{lir}}\) denote the location independent riskier stochastic order. Let \((X_i,Y_i)\), \(i=1,2,\dots,n\), be independent pairs of random variables. The authors show that if \(X_i\leq_{\text{lir}}Y_i\) for each \(i\), and if \(X_i\) and \(Y_i\) have log-concave density or probability functions, then \(\sum_{i=1}^nX_i\leq_{\text{lir}}\sum_{i=1}^nY_i\). They also show that if \(X_i\leq_{\text{disp}}Y_i\) for each \(i\) (where \(\leq_{\text{disp}}\) denotes the dispersive order), and if \(X_i\) and \(Y_i\) have log-concave distribution functions, then, again, \(\sum_{i=1}^nX_i\leq_{\text{lir}}\sum_{i=1}^nY_i\). Similar results hold for the excess wealth (or, equivalently, right spread) order.