Agent-Based Econometric Modeling

The purpose of this project is to build a new econometric paradigm: the agent-based econometrics. Compared with the homogeneous agents and rational expectation hypothesis which are the basic assumptions of many well established economic models, the agent-based econometric modeling has two main characteristics, namely, the heterogeneous agents and bounded rationality, respectively, which are more realistic in real world. The advantage of the agent-based econometric modeling is the use of macro data to study the heterogeneous behaviors among agents. This project attempts to establish a full process of agent-based econometric modeling. It contains the model setting, identification, estimation, and hypothesis testing, respectively. By proper design of the agent-based model, the estimation results of this project can link to the results of the behavior economic and the behavior finance literature. The identification and estimation are two important issues in agent-based econometric modeling because that the model is more complex than the model under homogeneous agents and rational expectation assumptions. That means that the commonly used maximum likelihood method and least square method could not able to handle this problem. This project considers the simulation-based econometric methods as a tool for model estimation. Finally, we apply the agent-based econometrics to both order-driven and dealers markets, respectively, which are the two main trading mechanisms in financial markets.

Project ID:PF9902-0540
External Project ID:NSC98-2410-H182-010-MY3