Asset Pricing with Bayesian Learning and Signal Distortion

Jun Zhang

2004-01-01

Abstract:Assuming asymmetric information, heterogeneous beliefs among agents, as well as Bayesian learning from the performance of stock markets, this paper demonstrates that the stock prices would have a sustained deviation from its fundamental values a la the conventional expected present value (EPV) approach of asset pricing. More specifically, the EPV approach would underprice the stocks in a booming market and overprice them in a bearish one. In ad- dition, the deviation is biggest when there is a widespread rumor in the market rather than the announcement of the truth. The intuition is based on the idea that current beliefs of the stock's quality would affect the expectation of subsequent signals, which is ignored in the traditional asset pricing approach. An additional finding is that speculative manipulations are not profitable at least when all agents have the same accessibility to the signals. When we introduce signal distortions in the model, speculative manipulation is more profitable than the baseline model. As one of the contributions to the mainstream learning or herding literatures on financial markets, this paper is among the first few that endogenizes the price process.

Economics