Optimal reinsurance and investment via stochastic projected gradient method based on Malliavin calculus

Yuta Otsuki,Shotaro Yagishita

2024-11-08

Abstract:This paper proposes a new approach using the stochastic projected gradient method and Malliavin calculus for optimal reinsurance and investment strategies. Unlike traditional methodologies, we aim to optimize static investment and reinsurance strategies by directly minimizing the ruin probability. Furthermore, we provide a convergence analysis of the stochastic projected gradient method for general constrained optimization problems whose objective function has Hölder continuous gradient. Numerical experiments show the effectiveness of our proposed method.

Mathematical Finance,Optimization and Control,Computational Finance

What problem does this paper attempt to address?

The problem that this paper attempts to solve is to optimize reinsurance and investment strategies by introducing a new method based on the stochastic projected gradient method and Malliavin calculus, in order to directly minimize the probability of ruin. Specifically, the paper proposes a static reinsurance and investment strategy optimization method. Different from traditional dynamic methods, this method is more practical in actual applications.

### Main problems and objectives of the paper

1. **Minimizing the probability of ruin**:
- The probability of ruin refers to the probability that an insurance company is unable to pay compensation due to insufficient funds at a certain point in the future. The goal of the paper is to minimize this probability by optimizing reinsurance and investment strategies.
- The specific definition of the probability of ruin is:
\[
P\left(\inf_{0 \leq t \leq T} U_t < 0\right) \quad \text{or} \quad P(U_T < 0)
\]
where \(U_t\) is the surplus process of the insurance company.

2. **Using static strategies**:
- Traditional methods such as the Hamilton - Jacobi - Bellman (HJB) method and the adjustment coefficient approach usually involve continuous - time stochastic processes, which are less realistic in practical operations. The method proposed in this paper adopts static strategies, that is, the decision variables are definite values rather than stochastic processes that change with time.

3. **Introducing Malliavin calculus and the stochastic projected gradient method**:
- In order to calculate the gradient of the objective function, the author uses Malliavin calculus to obtain an unbiased estimate and combines it with the stochastic projected gradient method for optimization.
- The advantage of this method is that it can handle asset models with geometric Brownian motion, which the adjustment coefficient method cannot do.

### Contributions of the paper

- A brand - new method is proposed to directly minimize the probability of ruin, rather than just optimizing its upper bound.
- Through Malliavin calculus and the stochastic projected gradient method, an effective way to calculate the gradient is provided, and the convergence of the algorithm is proved.
- For general constrained optimization problems, the convergence results are proved under the condition that the gradient of the objective function satisfies Hölder continuity.

### Summary

The core problem of this paper is to solve the problem of how non - life insurance companies minimize the risk of ruin through reinsurance and investment strategies in risk management by introducing new mathematical tools and optimization methods. Compared with traditional dynamic methods, the method proposed in this paper is more practical and can handle more complex asset models.