Optimal investment and reinsurance policies for the Cramér-Lundberg risk model under monotone mean-variance preference

Bohan Li, Junyi Guo, Linlin TianInt. J. Control 2024 · OpenReview

TL;DR

This paper tackles the complex problem of an insurance company optimizing its investment and reinsurance strategies under a "monotone mean-variance" (MMV) preference. By developing a novel framework that combines financial markets with jump processes in insurance, they derive explicit formulas for optimal strategies and the efficient frontier, offering a more robust approach than traditional mean-variance methods.

What Problem Does This Paper Address?

The paper addresses a critical challenge in financial and insurance mathematics: how to make optimal decisions when facing uncertainty about future wealth. Traditional "mean-variance" (MV) optimization, while popular, has a significant drawback: it can lead to irrational preferences where an investor might prefer a lower wealth outcome over a higher one. This is particularly problematic in dynamic settings like insurance, where the company needs to balance investments and reinsurance to ensure solvency and maximize returns. The authors introduce and apply the "monotone mean-variance" (MMV) criterion, which guarantees that preferences are always rational (i.e., more wealth is always preferred to less). They specifically apply this to the Cramér-Lundberg risk model, which accounts for the jump-like nature of insurance claims, a more realistic scenario than simpler models. The core problem is to find optimal investment (portfolio allocation) and reinsurance (retention level) strategies that maximize this MMV criterion for the insurance company's terminal wealth.

Methods

The authors employ a sophisticated combination of techniques:

  • Dynamic Programming and Hamilton-Jacobi-Bellman-Isaacs (HJBI) Equations: This is the cornerstone of their approach. They formulate the problem as a two-player zero-sum stochastic differential game (SDG) between the insurance company and an "imaginary player" representing the market. The HJBI equation is derived from the dynamic programming principle to solve this game.
  • Stochastic Differential Games (SDG): The problem is framed as a game where the insurance company (player one) aims to maximize the MMV criterion, while the market (player two) aims to minimize it. This framework allows for the analysis of strategic interactions.
  • Martingale Representation Theorem: A generalized version of this theorem is used to handle the complexities of the disturbed probability measure Q, which arises from the MMV criterion. This is crucial for connecting the MMV problem to a more tractable auxiliary problem.
  • Explicit Solution Derivation: After solving the HJBI equation, the authors derive closed-form solutions for the optimal investment and reinsurance strategies, as well as the value function (the maximized MMV criterion).

Datasets

Not discussed in the paper. The paper focuses on theoretical modeling and derivation, with a numerical example provided for illustration rather than dataset-driven evaluation.

Key Insights and Ideas

  • MMV Preference Solves Non-Monotonicity: The MMV criterion, unlike the standard MV criterion, ensures rational preferences, meaning more wealth is always preferred. This is crucial for realistic financial and insurance decision-making.
  • MMV Efficient Frontier Matches Classical MV: A surprising and important finding is that the efficient frontier (the set of optimal risk-return trade-offs) for the MMV problem is identical to that of the classical MV problem. This suggests that while the MMV criterion ensures rational preferences, it doesn't fundamentally alter the achievable risk-return trade-offs.
  • Jump Processes are Crucial: Incorporating jump processes in the insurance surplus model (Cramér-Lundberg) makes the problem more realistic and challenging, requiring advanced mathematical tools.
  • Zero-Sum Game Formulation is Powerful: Framing the problem as a zero-sum game between the insurer and the market provides a robust framework for deriving optimal strategies.

Key Results

  • Explicit Optimal Strategies: The paper provides explicit, closed-form formulas for the optimal investment portfolio and the optimal reinsurance retention level for the insurance company.
  • Explicit Value Function: The maximized value of the MMV criterion (the value function) is also derived in closed form.
  • Efficient Frontier Characterization: The paper characterizes the MMV efficient frontier, showing its equivalence to the classical MV efficient frontier.
  • Numerical Example: A numerical example demonstrates how the MMV criterion can lead to different, more rational outcomes compared to the classical MV criterion in specific scenarios, particularly when wealth outcomes fall outside the domain of monotonicity of the classical MV preference.

Limitations and Future Work

  • Model Assumptions: The paper relies on specific assumptions for the financial market (Black-Scholes model) and insurance model (Cramér-Lundberg with exponential claims). Extending these to more complex models (e.g., Cox processes, spectral negative Lévy processes) is a natural next step.
  • Coincidence of Strategies: While the efficient frontiers coincide, the paper notes that it's an open question whether the optimal strategies themselves always coincide for a broader class of jump processes. Future work could explore this further.
  • Computational Complexity: While closed-form solutions are derived, their practical implementation might still involve significant computation, especially for complex parameter settings.
  • Time-Inconsistency: The paper focuses on a time-consistent MMV criterion. Investigating time-inconsistent MMV problems could be a direction for future research.

Summary updated

BibTeX
@article{li2024optimal,
  author = {Bohan Li and Junyi Guo and Linlin Tian},
  title = {{Optimal investment and reinsurance policies for the Cramér-Lundberg risk model under monotone mean-variance preference}},
  journal = {Int. J. Control},
  year = {2024},
  url = {https://openreview.net/forum?id=OYulCO0YgF},
}
Figure