Calibrating an Adaptive Farmer-Joshi Agent-Based Model for Financial Markets
This report is published in fulfilment of an honours degree in statistics. The project explores the calibration and simulation of the Farmer and Joshi (2002) agent-based model of financial markets using the method of moments along with a genetic algorithm and a Nelder-Mead with threshold accepting algorithm. The model is used for understanding daily trading decisions made from closing auction to closing auction in equity markets, as it attempts to model financial market behaviour without the inclusion of agent adaptation. However, our attempt at calibrating the model has limited success in replicating important stylized facts observed in financial markets, similar to what has been found in other calibration experiments of the model. This leads us to extend the Farmer-Joshi model to include agent adaptation using a Brock-Hommes (1998) approach to strategy fitness based on trading strategy proﬁtability. The adaptive Farmer-Joshi model allows trading agents to switch between strategies, favouring strategies that have been more profitable over some period of time determined by a free-parameter determining the profit monitoring time-horizon.