On Utility Maximization under Multivariate Fake Stationary Affine Volterra Models

This paper solves Merton's portfolio optimization problem in a multivariate fake stationary affine Volterra setting by employing a martingale optimality principle and a Riccati backward stochastic differential equation to derive semi-closed form optimal strategies, overcoming the challenges posed by non-Markovian and non-semimartingale dynamics.

The Gist

Imagine you are the captain of a ship (your investment portfolio) trying to reach a distant island (your financial goals) as efficiently as possible. In the old days, sailors used simple maps where the ocean's currents were predictable and smooth, like a calm river. This paper, however, deals with a much wilder ocean: one where the waves are "rough," jagged, and unpredictable, and where the currents of different ships are tangled together in complex ways.

Here is a breakdown of what this paper does, using everyday analogies:

1. The Problem: Navigating a "Rough" Ocean

For a long time, financial models assumed that stock market volatility (how much prices jump around) moved smoothly, like a gentle breeze. But recent observations show that real market volatility is actually "rough." Think of it like the difference between a smooth silk sheet and a crumpled piece of paper. The crumpled paper has tiny, jagged folds that change instantly.

Furthermore, this paper looks at a multi-asset world. Imagine you aren't just sailing one boat, but a whole fleet of different ships. These ships don't just move on their own; they are linked. If one ship hits a storm, the others might get tossed around too (correlation). The author wants to figure out the best way to steer this entire fleet to maximize your wealth, even though the ocean is "rough" and the ships are tangled together.

2. The Obstacle: The Map Doesn't Work

In standard finance, captains use a "GPS" called the Hamilton-Jacobi-Bellman (HJB) equation. This GPS works great if the ocean is smooth and the future depends only on where you are right now (Markovian).

But in this "rough" world:

• The ocean has a memory (it's non-Markovian). What happened five minutes ago still affects the waves right now.
• The waves are so jagged that standard calculus breaks down (it's non-semimartingale).

Because of this, the old GPS (HJB) crashes. You can't just look at your current position to decide where to go; you have to remember the entire history of the waves.

3. The Solution: A "Magic Compass" (The Martingale Optimality Principle)

Since the old GPS failed, the author invents a new navigation tool. Instead of trying to predict the future path of every wave, they use a Martingale Optimality Principle.

Think of this as a Magic Compass that doesn't tell you where to go, but tells you if your current steering is "fair."

• The Strategy: The author constructs a special "value function" (a scorecard). If your steering strategy is perfect, this scorecard stays steady (it's a "martingale"). If your strategy is bad, the scorecard drifts downward (it's a "supermartingale").
• The Trick: To make this work with rough waves, they use a technique called Martingale Distortion. Imagine you are looking at the ocean through a special pair of glasses that smooths out the jagged edges just enough to see the path, without losing the reality of the roughness.

4. The Engine: The "Riccati-Volterra" Equation

To actually calculate the steering angle, the author solves a complex mathematical puzzle called a Riccati-Volterra equation.

• The Analogy: Imagine you are trying to bake a cake (the optimal strategy), but the recipe depends on the temperature of the oven right now AND the temperature history of the last hour.
• This equation is the "recipe" that tells you exactly how much of each asset to buy or sell at every moment. It takes into account:
  • How rough the volatility is (the crumpled paper).
  • How much risk you are willing to take (your appetite for danger).
  • How the different ships in your fleet are connected.

5. The Results: Two Types of Sailors

The paper solves this problem for two types of investors:

1. The Risk-Taker (Power Utility): Someone who is willing to take big swings for big gains. The paper shows them exactly how to tilt their portfolio to catch the biggest waves without capsizing.
2. The Conservative (Exponential Utility): Someone who is terrified of losing money. The paper shows them how to keep their fleet safe, even in the roughest storms, by hedging perfectly against the jagged waves.

6. The Simulation: Testing the Compass

Finally, the author runs computer simulations (Section 4). They create a fake "rough ocean" with two ships and test their new steering strategy.

• The Finding: They show that ignoring the "roughness" of the market leads to bad steering. By using their new "Magic Compass" (the optimal strategy derived from the Riccati-Volterra equation), investors can navigate these rough waters much more efficiently than with old methods.

Summary

In short, this paper is a new navigation manual for a chaotic, jagged financial ocean. It admits that the old maps don't work because the market is rough and interconnected. Instead, it provides a sophisticated, mathematically rigorous "Magic Compass" that helps investors steer their multi-asset portfolios to the best possible outcome, whether they are aggressive risk-takers or cautious savers. It proves that even in a messy, unpredictable world, there is a precise, optimal way to sail.

Technical Summary

Here is a detailed technical summary of the paper "On Utility Maximization under Multivariate Fake Stationary Affine Volterra Models" by Emmanuel Gnabeyeu.

1. Problem Statement

The paper addresses Merton's portfolio optimization problem (maximizing expected utility of terminal wealth) within a complex financial market environment characterized by:

• Multivariate Setting: A market with $d$ risky assets and one risk-free asset.
• Rough Volatility: The volatility processes are modeled using Volterra stochastic differential equations rather than standard Markovian diffusions. This captures the "roughness" of volatility paths observed empirically (Hurst parameter $H < 0.5$).
• Fake Stationarity: The model employs a specific class of fake stationary affine Volterra models. Unlike standard affine models where stationarity is an asymptotic property, these models are constructed to exhibit constant mean and variance over time (stationary regimes) despite being non-Markovian.
• Non-Markovian and Non-Semimartingale Nature: The underlying volatility processes are neither Markovian nor semimartingales. This renders classical stochastic control methods based on the Hamilton-Jacobi-Bellman (HJB) partial differential equation inapplicable.
• Correlation Structures: The model accounts for:
  • Leverage effects (correlation between asset returns and their own volatility).
  • Inter-asset correlations.
  • Both degenerate (identical correlation across assets) and general (heterogeneous correlation) cases.

The objective is to derive explicit optimal investment strategies for investors with Power Utility (CRRA) and Exponential Utility (CARA) preferences.

2. Methodology

Due to the failure of the HJB approach, the author employs a combination of the Martingale Optimality Principle and a Verification Argument.

A. Model Framework

• Volatility Dynamics: The volatility vector $V_t$ follows a scaled Volterra square-root process driven by a convolution kernel $K$ (e.g., fractional kernels $t^{\alpha-1}$).
• Fake Stationarity: The model introduces a "stabilizer" function $\varsigma(t)$ and specific initial conditions to ensure the process has constant mean and variance, allowing for consistent modeling across short and long maturities.
• Market Dynamics: Asset prices follow geometric Brownian motion with stochastic volatility, where the market price of risk is linear in the volatility.

B. Solution Strategy

1. Martingale Distortion Ansatz (Degenerate Case):

   • For the case where correlations are degenerate (identical across assets), the author adapts the martingale distortion transformation (inspired by Zariphopoulou, Fouque & Hu, and Han & Wong).
   • This involves changing the probability measure and constructing an auxiliary process $\Gamma_t$ (an exponential-affine functional of the volatility) to separate the wealth process from the volatility dynamics.
   • The value function is decomposed as $J_t = U(X_t) \cdot \Gamma_t$.

2. Riccati-Volterra Equations:

   • The auxiliary process $\Gamma_t$ is linked to the solution $\psi$ of a time-dependent, inhomogeneous Riccati-Volterra equation.
   • This equation is a functional integral equation involving the kernel $K$ and the model parameters.
   • The existence and uniqueness of the solution $\psi$ are established under specific assumptions on the kernel and parameters.

3. Verification Argument (General Case):

   • For the general correlation case (where correlations differ across assets), the martingale distortion ansatz fails.
   • The author employs a verification theorem approach. They construct a candidate value function and an optimal strategy based on the solution $\psi$ of the generalized Riccati-Volterra equation.
   • Using Itô's formula and properties of stochastic exponentials (Doléans-Dade), they prove that the candidate strategy yields a martingale for the value process, while any other admissible strategy yields a supermartingale. This confirms optimality without requiring the correlation structure to be degenerate.

3. Key Contributions

1. Multivariate Extension: The paper extends utility maximization results from univariate rough volatility models to a multivariate setting, addressing the critical practical need for multi-asset portfolio allocation with correlated risks.
2. Fake Stationary Framework: It introduces and utilizes fake stationary affine Volterra models, providing a unified framework that captures both short-term roughness and long-term stationarity, enabling robust fitting of the entire term structure of volatility.
3. General Correlation Handling: Unlike previous works that restricted solutions to degenerate correlation structures, this paper provides a solution for arbitrary correlation matrices using a verification argument, significantly broadening the applicability of the results.
4. Semi-Closed Form Solutions: The optimal strategies are derived in semi-closed form. They depend on the solution to a time-dependent multivariate Riccati-Volterra equation, which can be solved efficiently using numerical methods (e.g., fractional Adams-Bashforth-Moulton).
5. Dual Utility Results: The paper provides explicit optimal strategies for both Power Utility (CRRA) and Exponential Utility (CARA), covering different investor risk profiles.

4. Key Results

• Optimal Strategy Form: The optimal portfolio strategy $\alpha^*_t$ (amount invested in risky assets) takes the form:
  $$\alpha^*_t = \frac{1}{1-\gamma} (\lambda_t + \Sigma \Lambda_t) \quad \text{(Power Utility)}$$
  $$\alpha^*_t = \frac{1}{\gamma} e^{-\int_t^T r(s)ds} (\lambda_t + \Sigma \Lambda_t) \quad \text{(Exponential Utility)}$$
  Where:
  • $\lambda_t$ is the market price of risk.
  • $\Sigma$ represents the correlation structure.
  • $\Lambda_t$ is a hedging term proportional to $\nu \varsigma(t) \psi(T-t) \sqrt{V_t}$.
  • $\psi$ is the solution to the Riccati-Volterra equation.
• Value Function: The value function is expressed explicitly in terms of the initial wealth, the risk-free rate, and the initial value of the auxiliary process $\Gamma_0$, which depends on the solution $\psi$.
• Numerical Validation: The paper includes numerical experiments on a two-dimensional fake stationary rough Heston model.
  • Results illustrate how the Hurst parameter (roughness) and the stabilizing function affect the optimal allocation.
  • The simulations confirm that rough volatility significantly impacts hedging demands and portfolio rebalancing compared to classical models.

5. Significance

• Bridging Theory and Practice: The paper bridges the gap between the theoretical advances in rough volatility modeling (which have largely focused on option pricing) and practical portfolio management.
• Overcoming Non-Markovianity: It demonstrates a robust methodology for solving stochastic control problems in non-Markovian environments where standard PDE techniques fail, offering a template for future research in fractional and Volterra-based finance.
• Realistic Market Modeling: By incorporating multivariate correlations and fake stationarity, the model offers a more realistic representation of financial markets, particularly for managing portfolios with multiple correlated assets and capturing the "rough" nature of volatility observed in high-frequency data.
• Computational Feasibility: The reduction of the control problem to a Riccati-Volterra equation makes the solution computationally tractable, allowing practitioners to implement these strategies using standard numerical integration schemes.

In summary, this paper provides a rigorous mathematical framework and explicit solutions for optimal portfolio selection in a sophisticated, multivariate, rough volatility environment, overcoming the limitations of classical Markovian models and degenerate correlation assumptions.