An Arti cial Market Analysis of Exchange Rate Dynamics

In this study, we propose an exchange market model with arti cial adaptive agents (AGEDASI TOF) and conduct a computer simulation as a study on the arti cial market. First, AGEDASI TOF is compared with a random walk model and a linear regression model in out-of-sample forecasts. The results indicate that AGEDASI TOF outperforms the other models over all forecast horizons. Next, we analyze an exchange rate bubble in the real world using AGEDASI TOF. From the results, we conclude that the bubble starts by support of fundamental factors, grows by bandwagon expectations, stops by coincidence of all agents' expectations, and collapses by regressive expectations. This study shows that the arti cial market approach is e ective not only for qualitative analysis but also for quantitative analysis of economic systems in the real world.