Development of an artificial market model based on a field study

An artificial market model for analyzing economic phenomena has been developed based on the results of a field study. In the construction of the model, the forms of the virtual dealers' utility function and learning methods were derived from field data. The results of simulating the artificial market were compared with the field data. First, the temporal movements of agents' forecasts in the simulation were compared with those of actual dealers as given by survey data. Next, the factors that the virtual dealers regarded as significant in forecasting were categorized. The categories were compared to those of actual dealers and found to be similar. Finally, the temporal changes in virtual dealers' opinions were compared with data obtained through interviews with actual dealers. The proposed model was found to enable repeated and detailed analysis of dealer thinking under various conditions in real time. (C) 2003 Elsevier Inc. All rights reserved.