Stock Market Simulation based on Interaction of Heterogeneous Learning Agents

The main object of this study is the stock market, seen as a complex system constituted of basic elements (securities, trading infrastructure and atomistic heterogeneous investors) and process flows (forecasting, investment decision making, trade execution, maintenance of financial records, etc.). More specifically, we explore the possibilities to use specific simulation models as tools for generating economically interesting and adequate systemic behaviour, analysing properties of the actual stock market and explaining certain market phenomena. We propose two specific artificial stock market (ASM) models, based on inductive, competitive reinforcement-learning behaviour of atomistic heterogeneous agents in highly uncertain environment. These models are used as a structured framework for the analysis of market self-regulation abilities, efficiency, determinants of price formation and investors’ attitude to risk, emergent properties of the financial market and possible determinants of asset price bubbles. One model is calibrated to actual financial data.One of distinctive features of the proposed modelling approach is a strong emphasis on economic behaviour of individual agents. In the proposed models boundedly rational agents base their decisions on economic considerations, such as estimation of discounted earnings and comparison of returns on different investment strategies, and pursue forward-looking behaviour in highly uncertain environment. Agents’ individual adaptation... [to full text]