Optimal proportional and excess-of-loss reinsurance for multiple classes of insurance business (Q6089414)

In this paper, the author considers a reinsurance strategy which fuses a proportional reinsurance and an excess-of-loss reinsurance in a continuous-time insurance risk model. The author considers the situation where there are multiple classes of insurance business and supposes that the claim numbers of the classes follow a multivariate Poisson process. The objective is to determine an optimal reinsurance strategy so that the expected utility on the terminal wealth is maximized. Using the Hamilton-Jacobi-Bellman (HJB) dynamic programming approach, the authors derive an HJB equation for the value function. A verification theorem is established, and sufficient conditions for the optimality are determined. In the case of an exponential utility, the author obtains an explicit solution to the optimal reinsurance problem via solving the respective static constrained optimization problem. Section 2 of the paper formulates the model. Specifically, the model dynamics described by a multivariate Poisson process for the multiple classes of insurance business are presented. The admissible strategies formed by combining a proportional reinsurance and an excess-of-loss reinsurance are provided. The objective criterion given by the expected utility on the terminal wealth and the respective value function are presented. Proposition 2.3 gives preliminary estimates for the value function. Section 3 provides the HJB equation for the value function as well as the verification theorem. Specifically, Theorem 3.1 shows that the value function is a viscosity solution of the HJB equation with an appropriate terminal condition. The verification theorem for the value function and the optimal reinsurance strategy is presented in Theorem 3.2. Section 4 considers the case of the exponential utility. Theorem 4.2 and Theorem 4.5 give the solutions to the optimal reinsurance problem, where the symmetric case is analysed in Theorem 4.5. Numerical examples are provided for models with two classes of insurance business and exponentially distributed claim sizes.