Title: Self-fulfilling Ising Model of Financial Markets

Abstract: We study a dynamical Ising model of agents' opinions (buy or sell) with coupling coefficients reassessed continuously in time according to how past external news (magnetic field) have explained realized market returns. By combining herding, the impact of external news and private information, we test within the same model the hypothesis that agents are rational versus irrational. We find that the stylized facts of financial markets are reproduced only when agents are over-confident and mis-attribute the success of news to predict return to herding effects, thereby providing positive feedbacks leading to the model functioning close to the critical point.