ARTIFICIAL FINANCIAL MARKET: AN AGENT-BASED APPROACH TO MODELING THE INFLUENCE OF TRADERS CHARACTERISTICS ON EMERGENT MARKET PHENOMENA
Analytics
146 views ◎96 downloads ⇓
Abstract
Financial markets play a critical role in today's societies and are extremely important to the general health and efficiency of an economy. Solid understanding of the behavior of financial markets can benefit investors, governing organizations and in general, the whole economic growth. In financial markets, psychology and sociology of the traders have significant effects in giving rise to unique and unexpected (emergent) macroscopic properties, causing the classical economic approaches hardly effective to explain or reproduce them. Agent-based modeling (ABM) is a flexible methodology for simulating such complex systems and their behaviors. The literature using agent-based approach for analyzing patterns and phenomena observed in complex systems such as financial markets has grown into an important field of research in the recent years.This research proposes an agent-based model of stock market to study the behavior of market and traders. The model is created using a bottom-up approach, taking into consideration the effects of cognitive processes and behaviors of the traders (e.g. decision-making, interpretation of public information and learning) on the emergent phenomena of financial markets. We use Genetic Algorithm (GA) to better explore the parameters space and find the best parameter set with respect to our optimization function, which is replicating some of the important statistical properties present in actual financial time series (stylized facts). This study suggests that local interactions, rational and irrational decision-making approaches and heterogeneity, which has been incorporated into different aspects of agent design, are among the key elements in modeling financial markets. In this model, all agents fall under three general trading systems; fundamentalist, optimist, and pessimist, while having different beliefs and reaction intensities within each category. To evaluate the effectiveness and validity of our approach, a series of statistical analysis was conducted to test the artificial data with respect to a benchmark provided by the Bank of America (BAC) stock over a sufficiently long period of time. The results revealed that the model was able to reproduce and explain some of the most important stylized facts observed in actual financial time series and was consistent with empirical observations. Using this validated model, we estimate how and in what direction different market phenomena and states such as herding, volume, market volatility, number of traders in different categories and bullish/bearish markets, influence each other. Herding is an emergent property of financial markets, often leading to the creation of speculative bubbles. Bubbles inevitably make markets unstable and prone to major crashes, hence it is crucial to understand the origins and driving forces of herd behavior. The results from model generated time series suggest that herd behavior may cause or intensify volatility in the market, but not the other way around. Further, there is strong evidence of a bi-directional causal relationship between market volatility and trading volume in both model generated and BAC time series, implying that past values of trading volume in the market can help predict the current level of volatility, and vice versa. These investigations can help enhance the understanding of dynamics and structure of financial markets and paves the way for further explorations on causes and effects of the market and traders characteristics and behavior.Using our proposed validated framework, we design and develop the necessary infrastructure that allows us to better explore and understand the relationships between traders behavioral factors and global market properties.