Agent based modeling of single asset markets

Abstract

Understanding of the human behaviors has been a life long attempt along the history of science. This attempt has been accelerated along with the increasing speed of the computers which helps researchers to develop modern and more efficient modeling techniques. One of these recently developed approaches for modeling of the human behaviors is the simulation modeling approach. This thesis addresses the issue of the modeling of human behaviors (or decisions) by the application of the newly emerged simulation modeling technique called agent-based modeling. Agent-based modeling technique is a bottom-up simulation modeling approach through which human decisions are modeled and the global resultant dynamics emerged from the interactions of these humans, namely agents, possessing different decision mechanisms are tried to be understood. In this thesis, we apply agent-based modeling technique in order to model two different single asset market models and to grasp the fundamental parameters of these models that correspond to the global human behaviors which are represented by some statistical facts that are derived from the statistical analysis of real stock market asset returns. In order to accomplish this task, first, models are described deeply and model parameters (decisions) that are expected to be crucial are examined through the sensitivity analysis considering those global statistical facts. The first model introduced is a simple single asset market model in which there is only one type of trader with different decision parameters. Second model described is again a single asset market model with a more realistic, but more sophisticated trader setting, in which there are two general trader types. According to the investigation performed on these models, we could be able to reach important results on how humans behave while they are trading a single asset in a financial market.