Equilibrium recoveries in insurance markets with limited liability (Q2283131)

An insurer takes over a risk \(X = (X_k: 1 \le k \le n)\) from \(n\) homogeneous agents. A part \(\alpha\) of the wealth is invested in a risky asset. Then the wealth without the insurance claims becomes \(A(\alpha,\pi) = (W + n \pi)(1 + r + \alpha R)\), where \(\pi\) is the premium and \(R\) is the return over the riskless interest rate \(r\). Default occurs if \(\{ A(\alpha,\pi) < \bar X\}\) with \(\bar X = \sum_{k=1}^n X_k\). The compensation for the insured is \(f = (f_k(a,x))\), where \(f_k(a, x) = x_k\) if \(a \ge \bar x = \sum_{k=1}^n x_k\) and \(\sum_{k=1}^n f_k(a,x) = (1-\delta) a\) otherwise. Here \(\delta \in [0,1]\) denotes the recovery rate. A constraint is that \(f_k(a,x) \le x_k\). The insurer has the possibility to chose the constants \(\alpha\) and \(\pi\). The goal is to maximise the expected profit \(E[\{A(\alpha,\pi) - \bar X\}^+]\), where \(y^+ = \max\{y,0\}\) is the positive part of \(y\). The policy holders only take insurance if their expected utility \(E[u(w_0 - \pi - X_k +f_k(A(\alpha,\pi),X))]\) exceeds a level \(\underline u\). For example, \(\underline u = E[u(w_0 - X_k)]\) could be the expected utility without insurance. Here, it is assumed that all agents have the same initial wealth \(w_0\) and the same utility function \(u\). As an additional assumption, the risks \(X\) are conditionally exchangeable given \(R\). That is, the conditional distribution of \((X_1, \ldots, X_n) \mid R\) is invariant under a permutation of the index set \(\{1,\ldots, n\}\). Moreover, \(E[R] > 0\) is assumed, so investment is profitable. Not surprisingly, the insurer would like to choose \(\pi\) and \(\alpha\) as large a possible. If \(f_k < 0\) is allowed, the optimal solution is perfect pooling, \(f_k(a,x) = x_k - (\bar x - (1-\delta))/n\) if \(a < \bar x\). That is, every agent covers the same amount of the loss of the pool \(\bar x-a\). In this case, for some of the agents \(f_k < 0\) is possible. That is, an agent may have to pay more than his/her claim. If we further add the constraint \(f_k \ge 0\), then the optimal policy is \(f_k(a,x) = (x_k-\gamma)^+\) in the case \(a <\bar x\), where \(\gamma\) is chosen such that \(\sum_{k=1}^n f_k(a,x) =(1-\delta) a\). This means again, that each agent gets the same amount but does not have to cover more that the own loss.