Modeling the Stock Market prior to large crashes

Abstract

We propose that the minimal requirements for a model of stock market price fluctuations should comprise time asymmetry, robustness with respect to connectivity between agents, ``bounded rationality'' and a probabilistic description. We also compare extensively two previously proposed models of log-periodic behavior of the stock market index prior to a large crash. We find that the model which follows the above requirements outperforms the other with a high statistical significance.