An investigation of the behaviour of financial markets using agent-based computational models

Abstract

This thesis aims to investigate the behaviour of financial markets by using agent-based computational models. By using a special adaptive form of the Strongly Typed Genetic Programming (STGP)- based learning algorithm and real historical data of stocks, indices and currency pairs I analysed various stylized facts of financial returns, market efficiency and stock market forecasts. This thesis also sought to discuss the following: 1) The appearance of herding in financial markets and the behavioural foundations of stylised facts of financial returns; 2) The implications of trader cognitive abilities for stock market properties; 3) The relationship between market efficiency and market adaptability; 4) The development of profitable stock market forecasts and the price-volume relationship; 5) High frequency trading, technical analysis and market efficiency. The main findings and contributions suggest that: 1) The magnitude of herding behaviour does not contribute to the mispricing of assets in the long run; 2) Individual rationality and market structure are equally important in market performance; 3) Stock market dynamics are better explained by the evolutionary process associated with the Adaptive Market Hypothesis; 4) The STGP technique significantly outperforms traditional forecasting methods such as Box-Jenkins and Holt-Winters; 5) The dynamic relationship between price and volume revealed inconclusive forecasting picture; 6) There is no definite answers as to whether high frequency trading is harmful or beneficial to market efficiency.