Nash equilibria for relative investors via no-arbitrage arguments

Basic Information

• Paper ID: 2111.02310
• Title: Nash equilibria for relative investors via no-arbitrage arguments
• Authors: Nicole Bäuerle, Tamara Göll
• Classification: math.OC (Mathematical Optimization and Control)
• Publication Date: September 13, 2022 (arXiv v4)
• Paper Link: https://arxiv.org/abs/2111.02310

Abstract

In a common arbitrage-free semimartingale financial market, this paper addresses the problem of determining Nash equilibrium investment strategies for n investors who seek to maximize the expected utility of their relative wealth. The utility functions considered here can be quite general. By leveraging the linearity of stochastic integrals and classical pricing theory, the authors express all Nash equilibrium investment strategies in terms of optimal strategies from classical single-agent expected utility problems. The corresponding mean-field problem is solved in the same manner. The paper provides applications to four specific financial markets and compares results with those in the literature.

Research Background and Motivation

1. Problem Definition: The core problem addressed in this research is how to determine Nash equilibrium investment strategies in a multi-investor competitive environment, where each investor cares not only about their absolute wealth but also about their wealth performance relative to other investors.
2. Problem Significance:
   • In practice, investors often exhibit relative wealth concerns, such as performance competition among fund managers
   • Relative investment problems hold important significance in behavioral finance
   • Multi-agent game theory has substantial application value in financial markets
3. Limitations of Existing Approaches:
   • Previous research has been largely restricted to specific market models (e.g., geometric Brownian motion)
   • Utility function forms are limited (primarily CRRA or CARA types)
   • The uniqueness of Nash equilibrium remains an open question in the literature
   • Lack of unified treatment for general semimartingale markets
4. Research Motivation:
   • Establish a unified framework applicable to general arbitrage-free semimartingale markets
   • Handle more general classes of utility functions
   • Resolve the Nash equilibrium uniqueness problem
   • Provide concise solutions based on classical pricing theory

Core Contributions

1. Proposed a novel approach based on arbitrage-free pricing theory: By leveraging the linearity of stochastic integrals, the multi-agent relative investment problem is transformed into a classical single-agent problem.
2. Established a unified framework applicable to general semimartingale markets: Extended existing literature to allow for jumps in stock price processes.
3. Resolved the Nash equilibrium uniqueness problem: Under conditions where the classical problem has a unique solution, proved the uniqueness of Nash equilibrium.
4. Provided general formulas for explicit solutions: Derived closed-form expressions for Nash equilibrium strategies (Formula 3.3).
5. Extended to mean-field games: Successfully handled the limiting case where the number of investors approaches infinity.
6. Proved equivalence between Nash equilibrium and Pareto optimality: Demonstrated that individual optimality coincides with social optimality in this setting.

Methodology Details

Problem Formulation

Consider n investors in a common arbitrage-free semimartingale financial market, where each investor i's objective is:

$\sup_{\phi^i \in A} \mathbb{E}\left[U_i\left(X^{i,\phi^i}_T - \theta_i \bar{X}^{-i,\phi}_T\right)\right]$

where:

• $X^{i,\phi^i}_T = x^i_0 + (\phi^i \cdot S)_T$ is investor i's terminal wealth
• $\bar{X}^{-i,\phi}_T = \frac{1}{n}\sum_{j \neq i} X^{j,\phi^j}_T$ is the average wealth of other investors
• $\theta_i \in [0,1]$ is the competition weight parameter
• $U_i$ is the utility function

Model Architecture

Step One: Pricing Decomposition Using linearity properties, decompose investor i's problem into:

• Hedging component: investing part of initial capital to hedge competitors' wealth $\theta_i \bar{X}^{-i,\phi}_T$
• Optimization component: using remaining capital $\tilde{x}^i_0 = x^i_0 - \theta_i \bar{x}^{-i}_0$ for classical utility maximization

Step Two: Classical Problem Solution Each investor solves the simplified classical problem: $\sup_{\psi^i \in A} \mathbb{E}\left[U_i\left(Y^{i,\psi^i}_T\right)\right]$ where $Y^{i,\psi^i}_T = \tilde{x}^i_0 + (\psi^i \cdot S)_T$

Step Three: Linear System Solution Solve the linear relationship: $\psi^{i,*}_k(t) = \phi^i_k(t) - \frac{\theta_i}{n}\sum_{j \neq i} \phi^j_k(t)$ to obtain Nash equilibrium strategies.

Technical Innovations

1. Pricing Theory Application: Innovatively applies classical asset pricing theory to multi-agent game problems, simplifying problem structure through hedging concepts.
2. Linear Decomposition Technique: Leverages the linearity of wealth processes to decompose complex multi-agent problems into independent single-agent problems.
3. General Utility Function Treatment: Imposes no restrictions on utility function form, enabling handling of non-standard utilities such as Cumulative Prospect Theory (CPT).
4. Unified Analytical Framework: Provides a unified solution applicable to various market models.

Experimental Setup

Application Scenarios

The paper provides four concrete applications:

1. Lévy Markets: General Lévy processes with jumps, using exponential utility functions
2. Heston Stochastic Volatility Model: Considering stochastic volatility, using power utility functions
3. Cumulative Prospect Theory (CPT): CPT utility in Black-Scholes markets
4. Cox-Ross-Rubinstein Discrete-Time Model: Binomial model using exponential utility

Comparison Methods

Comparison with existing results in the literature, particularly:

• Results from Lacker and Zariphopoulou (2019) in Black-Scholes markets
• Existing solutions for other specific market models

Parameter Settings

• Competition weights $\theta_i \in [0,1]$
• Initial capital constraints: $x^i_0 - \frac{\theta_i}{n}\sum_{j \neq i} x^j_0 > 0$ (for positive domain utility functions)
• Various utility function parameters (risk aversion coefficients, etc.)

Experimental Results

Main Results

Theorem 3.2 (Main Result): If the classical problem (3.1) has a unique optimal strategy $\psi^{i,*}$ for all i, then there exists a unique Nash equilibrium:

$\phi^i_k(t) = \frac{n}{n+\theta_i}\psi^{i,*}_k(t) + \frac{\theta_i}{(n+\theta_i)(1-\hat{\theta})}\sum_{j=1}^n \frac{n}{n+\theta_j}\psi^{j,*}_k(t)$

where $\hat{\theta} = \sum_{i=1}^n \frac{\theta_i}{n+\theta_i} < 1$.

Specific Application Results

1. Black-Scholes Market:
   • Optimal investment amount: $\pi^{i,*} = C_i \cdot (\sigma\sigma^T)^{-1}\mu$
   • Constant factor: $C_i = \frac{n\delta_i}{n+\theta_i} + \frac{\theta_i}{(n+\theta_i)(1-\hat{\theta})}\sum_{j=1}^n \frac{n\delta_j}{n+\theta_j}$
   • Results coincide with special cases of Lacker-Zariphopoulou
2. Heston Model:
   • Optimal strategy has deterministic functional form: $\frac{\psi^{i,*}(t)S(t)}{Y^{i,*}_t} = \delta_i\lambda + f_i(t)$
   • Nash equilibrium maintains the same structure
3. CPT Application:
   • First application of CPT in relative investment problems
   • Validates method's applicability to non-standard utility functions

Mean-Field Results

Theorem 5.3: As $n \to \infty$, the mean-field equilibrium is: $\phi^*_k(t) = \psi^*_k(t) + \frac{\theta}{1-\bar{\theta}}\mathbb{E}[\psi^*_k(t)|F_T]$

where $\bar{\theta} = \mathbb{E}[\theta] < 1$.

Experimental Findings

1. Monotonicity: Larger competition weight $\theta_i$ leads to more aggressive risk-taking by investors
2. Consistency: Nash equilibrium coincides with Pareto optimal strategies
3. Stability: Mean-field limit provides approximations for large-scale games
4. Universality: Method applies to jump-diffusions, stochastic volatility, and various market models

Related Work

Main Research Directions

1. Early Work:
   • Basak and Makarov (2015): Continuous-time model with two agents
   • Espinosa and Touzi (2015): Existence and uniqueness results for n agents
2. Extended Research:
   • Lacker and Zariphopoulou (2019): Related Black-Scholes markets
   • Extensions to consumption-investment problems, forward utilities, incomplete information, etc.
3. Market Model Extensions:
   • Stochastic volatility models (CIR)
   • Random coefficient Brownian motions
   • Itô diffusion markets

Advantages of This Paper

1. More General Market Models: Semimartingale framework encompasses jump processes
2. Broader Utility Functions: Not restricted to CRRA/CARA types
3. Resolves Uniqueness Problem: Provides sufficient conditions for uniqueness
4. Unified Methodology: General approach based on pricing theory

Conclusions and Discussion

Main Conclusions

1. Method Effectiveness: The approach based on arbitrage-free pricing theory effectively solves multi-agent relative investment problems
2. Uniqueness Results: Under conditions where the classical problem has a unique solution, Nash equilibrium exists uniquely
3. Structural Simplicity: Nash equilibrium can be expressed as linear combinations of classical single-agent solutions
4. Mean-Field Convergence: The limiting behavior of large-scale games is characterized

Limitations

1. Classical Problem Dependence: Method effectiveness depends on solvability of the classical single-agent problem
2. Parameter Constraints: Competition weights must satisfy certain constraints to ensure problem validity
3. Market Completeness Assumption: Some results may depend on market completeness assumptions
4. Computational Complexity: For complex utility functions, solving the classical problem may remain difficult

Future Directions

1. Incomplete Market Extensions: Study relative investment problems in incomplete markets
2. Heterogeneous Information: Consider cases where investors have different information sets
3. Dynamic Games: Investigate dynamic evolution of strategic interactions
4. Empirical Applications: Apply theoretical results to actual financial data

In-Depth Evaluation

Strengths

1. Strong Theoretical Innovation:
   • Cleverly applies pricing theory to game-theoretic problems
   • Provides unified framework for handling general utility functions
   • Resolves open problems in the literature (uniqueness)
2. Significant Methodological Contributions:
   • Linear decomposition idea is simple and elegant
   • Avoids complex fixed-point calculations
   • Extensible to mean-field settings
3. Broad Application Range:
   • Applicable to various market models
   • Supports multiple utility function types
   • Encompasses discrete and continuous-time settings
4. Mathematical Rigor:
   • Complete and clear proofs
   • Reasonable technical assumptions
   • General results

Shortcomings

1. Lack of Empirical Validation:
   • Absence of verification with actual data
   • Parameter estimation methods not addressed
   • Practical applicability remains to be tested
2. Computational Complexity:
   • For complex utility functions, classical problem solving remains difficult
   • Numerical implementation challenges for large-scale systems
3. Economic Interpretation:
   • Economic meaning of some technical assumptions unclear
   • Economic intuition of equilibrium strategies could be further elaborated
4. Robustness Analysis:
   • Lacks robustness analysis for model misspecification
   • Parameter perturbation sensitivity insufficiently discussed

Impact

1. Academic Contribution:
   • Provides important methodological contributions to relative investment theory
   • May inspire more game-theoretic research based on pricing theory
   • Advances mean-field game theory
2. Practical Value:
   • Provides theoretical foundation for competitive analysis in fund management
   • Can be used to design incentive mechanisms and risk management strategies
3. Reproducibility:
   • Theoretical results are clear and easily verifiable
   • Application examples are concrete and implementable

Applicable Scenarios

1. Fund Management: Analyze performance competition among fund managers
2. Pension Fund Management: Relative performance evaluation of multiple managers
3. Investment Banking: Resource allocation and incentive design among departments
4. Regulatory Policy: Understanding market competition's impact on investment behavior

References

This paper cites 28 important references, primarily including:

• Lacker, D. and Zariphopoulou, T. (2019). Mean field and n-agent games for optimal investment under relative performance criteria. Mathematical Finance 29(4), 1003-1038.
• Espinosa, G.-E. and Touzi, N. (2015). Optimal investment under relative performance concerns. Mathematical Finance 25(2), 221-257.
• Basak, S. and Makarov, D. (2015). Competition among portfolio managers and asset specialization.

These references form the important foundation for research on relative investment problems, upon which this paper makes significant theoretical and methodological improvements.