An Agent-Based Model to Study Informational Cascades in Financial Markets

The purpose of this paper is to build an agent-based model which is adapted to study the information cascades in financial markets. Thus, we design and implement a model with asynchronous continuous-time management, populated by heterogeneous traders, connected by an interaction network with directed and weighted edges, which allows them to observe and learn about the actions of their peers. The proposed model relaxes unrealistic assumptions of previous analytical models, particularly, homogeneity of traders and access to all previous decisions (full network). Therefore, the proposed model allows to study the impact of different factors in the emergence and magnitude of information cascades: mainly, the impact of the heterogeneity of traders behaviours and characteristics of the interaction network. Moreover, unlike analytical models which focus only on how information cascades occur, the proposed model allows studying the impact of information cascades on the price dynamics. Then, we use the model to perform a series of experiments to investigate the impact of various factors in the formation and magnitude of information cascades. The obtained results show in particular that informational cascades only occur when the market is dominated by an overwhelming majority of traders who have significant uncertainty in their own signals. This shows the benefit of designing policies to reduce uncertainty among investors to avoid informational cascades, whether by financial regulatory authorities or by companies with high degrees of information uncertainty.