A comparison between homogeneous and heterogeneous portfolios. (Q1413283)

One of the goals in the actuarial literature is to compare the riskiness of several portfolios. Suppose that \(N\) insurance businesses are exposed to \(N\) different independent risks \(X_1,\dots, X_N\) where \(X_i\) has the distribution function \(F_i\). Let us consider three portfolios: A heterogeneous portfolio and two different homogeneous portfolios. The heterogeneous portfolio is formed when the business \(j\) insures risk \(X_j\), \(HT=(X_1, \dots , X_N)\). Now, assume that each business \(j\) insures all the risks, so let \(X(i,j)\), \(i,j =1, \dots, N\), be independent random variables such that \(X(i,j)\) has the distribution \(F_i\). Consider a heterogeneous portfolio formed by a mixture of the individual risks \(X_1,\dots, X_N\) with probabilities \((1/N, \dots ,1/N )\), that is, \(HM1=(X(.,1), \dots , X(.,N))\) where \(X(.,j)=X(i,j)\) with probability \(1/N\). Finally, let a new heterogeneous portfolio be the average sum of all the risk, i.e., \(HM2=(\overline{X}_1, \dots, \overline{X}_N)\) where \(\overline{X}_j=\sum_{i=1}^NX(i,j)/N\). The author derives upper and lower bounds on the heterogeneous portfolio by the two types of homogeneous portfolios already described. Concretely, \[ E \psi (HM2) \leq E \psi(HT) \leq E \psi (HM1). \] The left inequality holds if \(\psi\) is a symmetric supermodular function that is convex in each argument separately while the right one is valid if \(\psi\) is a symmetric supermodular function.