Modeling Sub-Rational Human Investors in Financial Markets: Limited or Biased

Core Concepts

The authors introduce a flexible model incorporating aspects of human sub-rationality using reinforcement learning to analyze the impact on financial markets.

Abstract

The study explores human decision-making deviations from rationality due to limited information and psychological biases. It introduces models for bounded rationality, myopic behavior, prospect theory, and optimism/pessimism in trading strategies. The research uses market simulations to train RL agents and evaluate their impact on market dynamics.
The content discusses the challenges of modeling human behavior in financial markets due to computational limitations and psychological biases. It presents various models like bounded rationality, myopic behavior, prospect theory, and optimism/pessimism with examples illustrating their impact on trading decisions. The study highlights the importance of understanding human sub-rationality for better insights into investor behavior.

Stats

"Our model is trained using a high-fidelity multi-agent market simulator."
"Our experiments reveal that bounded-rational and prospect-biased human behaviors improve liquidity but diminish price efficiency."
"Loss aversion results in both risk averse behavior in gains and risk seeking behavior in losses."

Quotes

"The Bellman equation relates the value of the current state s to the one step reward and the value at the next state s′."
"Prospect biased investors tend to be risk-seeking when facing choices between losses."

Key Insights Distilled From

Limited or Biased

by Penghang Liu... at arxiv.org 03-12-2024

https://arxiv.org/pdf/2210.08569.pdf

Deeper Inquiries

How can behavioral finance studies enhance traditional economic models?

Behavioral finance studies can enhance traditional economic models by incorporating insights from psychology and other social sciences to provide a more realistic depiction of human decision-making. Traditional economic models often assume that individuals are perfectly rational, which is not always the case in real-life scenarios. By integrating findings from behavioral finance, such as cognitive biases and limited information processing capabilities, into economic models, we can better explain deviations from optimal decision-making observed in financial markets. This enriched understanding allows for more accurate predictions of market behavior and improved policy recommendations.

What are the implications of modeling human sub-rationality on investment strategies?

Modeling human sub-rationality in investment strategies can have significant implications on portfolio management and trading decisions. By considering factors such as bounded rationality, myopic behavior, prospect theory biases, optimism/pessimism, etc., investors can tailor their strategies to account for these deviations from perfect rationality. For example:

Bounded Rationality: Investors may adopt simpler decision rules or rely on heuristics rather than complex optimization techniques.
Myopic Behavior: Strategies may focus more on short-term gains/losses rather than long-term objectives.
Prospect Theory Biases: Portfolio allocations may be adjusted based on risk aversion tendencies towards gains or losses.
Optimism/Pessimism: Investment decisions could be influenced by overestimation or underestimation of future outcomes.
Understanding how these aspects impact investor behavior allows for the development of more robust and adaptive investment strategies that align with individual preferences and psychological biases.

How can reinforcement learning adapt to different aspects of human bias in financial markets?

Reinforcement learning (RL) offers a flexible framework to model various aspects of human bias in financial markets by adjusting reward functions and policies based on specific biases identified. Here's how RL can adapt to different types of bias:

Bounded Rationality: Introduce randomness or noise into action selection using Boltzmann softmax operator with adjustable parameters.
Myopic Behavior: Modify discount factors in the Bellman equation to prioritize immediate rewards over long-term gains.
Prospect Theory Biases: Adjust reward functions using concave/convex utility functions for gains/losses along with non-linear probability weighting schemes.
Optimism/Pessimism: Incorporate biased expectations into reward calculations leading to risk-seeking/risk-averse behaviors depending on optimistic/pessimistic outlooks.

By training RL agents within simulated environments that mimic these biases accurately, it becomes possible to develop adaptive trading strategies that reflect real-world investor behaviors influenced by various forms of cognitive bias present in financial markets.