Abstract
Based on the mean-variance portfolio selection under multi-period criterion, this paper focuses on the study of the uncertain time horizon and the regime-switching market including the bankruptcy state, where the conditional distribution of exit time is followed by the market state. When the market enters the bankruptcy state, investors are assumed to get back part of the wealth from the bankrupt company, where $\delta$ refers to the retrieval rate. By introducing the Lagrange multiplier, we create an innovative expression for the wealth process and the iterative representation of the value function to obtain the analytical expression of the optimal strategy and the corresponding efficient frontier. Besides, some special cases and numerical examples are presented to demonstrate the effects of state-dependent exit probability and bankruptcy state on the investment strategy.
Introduction
Standard Mean-Variance (MV) portfolio selection, pioneered by Markowitz, typically assumes fixed time horizons and static market conditions. However, real-world financial environments are characterized by "regime switches"—abrupt shifts between bullish and bearish states—and the risk of company bankruptcy. Furthermore, an investor’s decision to exit the market is often not independent of the market's performance; "flight to safety" behaviors often trigger early exits during downturns. This paper, published in the Journal of Mathematical Finance, addresses these complexities by deriving optimal multi-period investment strategies for markets with state-dependent exit probabilities and a dedicated bankruptcy state.
Problem Formulation and Methodology
The authors model the market using a regime-switching Markov chain with $L$ states, specifically incorporating a Bankruptcy State. In this state, risky assets lose their primary value, but investors are modeled to recover a fraction ($\delta$) of their wealth, known as the retrieval rate.
A key innovation of this research is the treatment of exit time. Unlike previous models that treat exit time as a fixed constant or an independent random variable, this framework conditions the probability of exiting at time $t$ on the current market state $S_t$.
To solve the resulting optimization problem, which is non-separable due to the variance term in the objective function, the authors employ:
- Embedding Technique: Transforming the mean-variance problem into an auxiliary quadratic optimization problem.
- Lagrange Multipliers: Introducing a multiplier to manage the wealth process constraints.
- Dynamic Programming: Deriving Bellman equations to find the value function iteratively, resulting in a closed-form analytical expression for the optimal strategy.
Key Insights
The study reveals several critical dynamics in portfolio management:
- The "Safety Fuse" Effect: State-dependent exit probabilities act as a protective mechanism. If investors are more likely to exit during bearish or bankruptcy states, the model shows the efficient frontier moves downward, indicating that the ability to exit early reduces the overall variance of terminal wealth.
- Impact of Retrieval Rates: The retrieval rate ($\delta$) significantly influences strategy. Even a modest expected recovery from a bankrupt entity allows for more aggressive investment in risky assets during the periods preceding a potential crash.
- Model Robustness: The authors demonstrate that their framework is a general case of existing theories. By adjusting specific parameters, the model can replicate classic results from Markowitz, Li & Ng, and others.
Results and Discussion
The primary contribution is the derivation of explicit equations (Theorems 1 and 2) that provide the optimal allocation for risky assets based on current wealth and market state. Numerical simulations over a four-period horizon confirm that as the probability of bankruptcy increases, the efficient frontier shifts upward and becomes steeper, requiring investors to accept higher risk for the same expected return.
While the model provides a rigorous mathematical foundation, it currently assumes a single risky asset and a frictionless market (no transaction costs or taxes). Future research may expand this to multi-asset portfolios and incorporate real-world constraints like liabilities or time-lagged retrieval rates.
Conclusion
This research provides a more realistic framework for multi-period portfolio selection by integrating bankruptcy risks and state-dependent exit behaviors. By providing analytical solutions, it offers a robust tool for balancing risk and return in volatile, regime-switching environments.
Summary updated