Mean-variance portfolio selection for a non-life insurance company (Q2472194)

This paper considers the problem of mean variance portfolio selection for a non-life insurance company. The insurance company may invest in a financial market which is modeled by a Levy version of the standard Black-Scholes-Merton financial market model, and is facing a risk process where the aggregate claims amount is modeled by a compound Cox process with stochastic intensity which is considered to be an Itô process driven by Brownian motion. The portfolio selection problem is modelled as a stochastic control problem, for the wealth process of the firm, which is now the solution of a Levy driven stochastic differential equation where the compound Cox process is included as a model for the claims. The control process is the amount of the wealth invested in the risky asset and the cost functional to be minimized is the linear combination expected distance of the wealth of the firm from some prescribed target, as well as the variance of wealth process. The constraint is to keep the expectation of wealth process at a terminal time equal to a prescribed value. This amounts to a quadratic linear control problem, which is an interesting generalization of the classical Markowitz and Merton portfolio problems in an insurance setting. The problem is also interesting from the mathematical point of view on account of the complexity of the stochastic processes modelling the financial market and the risk process. The authors derive a verification theorem, which connects the value function with the solution of a partial integrodifferential equation, and using a special ansatz as a candidate for the solution they obtain an analytic expression for optimal investment policy, in terms of a feedback law, involving the wealth process. The derivation of the analytic expression requires the specification of certain functions which are the solutions of some linear partial differential equations. The proofs of the results combine interesting mathematical techniques concerning the theory of Hamilton-Jacobi-Bellman equations, the theory of stochastic differential equations and probabilistic representation of the solution of PDEs. Based on this solution, they derive properties of the efficient frontier and the efficient portfolio. The results of the paper are interesting from the point of view of mathematics and are expected to have interesting applications in the field of portfolio management for insurance companies.