Dynamic market modeling with heterogeneous agents: Applications in diverse markets

Abstract

Financial markets, with a staggering market capitalization of 94 trillion US dollars in 2020, hold utmost importance in finance and economics. To comprehend their dynamics and price behavior, numerous models have been proposed, such as the Capital Asset Pricing Model (CAPM) and Arbitrage Pricing Theory (APT). However, these traditional approaches often fall short in capturing extreme market situations and the complexities of crashing market crises. In response to this limitation, agent-based modeling (ABM) has emerged as a paradigm shift, offering more realistic assumptions and allowing for emergent behavior due to interactions among heterogeneous agents. This study utilizes ABM to investigate financial market microstructure and stock market behavior through the creation of two distinct market models. The first model, an artificial stock market, aims to replicate real market price features and explore the impact of various trading strategies. Incorporating agents with realistic trading strategies, the model reveals insights into the competition among different strategies, highlighting the dominance of computationally powerful agents. Moreover, the presence of noise traders is observed to contribute to market liquidity and movement. Overall, the ABM approach successfully reproduces stylized facts observed in real markets and provides a bottom-up understanding of market dynamics. The second model is a hybrid limit order book simulator that combines historic NASDAQ messages and actual market orders to create a more realistic trading mechanism. By examining the impact of order size, market volatility, market capitalization, and stock trading volume, the simulator demonstrates a close resemblance between simulated and actual market prices. Notably, larger order sizes lead to greater price impact, while higher volatility stocks exhibit more significant price impacts. Additionally, market capitalization and stock trading volume are shown to influence bid-ask spreads. Although the study has some limitations, such as the need for further exploration of agent heterogeneity in fundamental markets, it contributes valuable insights into financial market microstructure dynamics. The developed limit order book simulator offers a platform for testing stock market hypotheses and evaluating trading strategies, paving the way for future research in financial stock markets and optimal portfolio selection. The agent-based modeling approach and its ability to capture complex market interactions provide promising avenues for understanding financial markets at a granular level and making informed decisions in the world of finance.