Stock Market Volatility and Learning

We study a standard consumption based asset pricing model with rationally investing agents but allow agents' prior beliefs about price and dividend behavior to deviate slightly from rational expectations priors. Learning about stock price behavior then causes the model to become quantitatively consistent with a range of basic asset prizing 'puzzles': stock returns display momentum and mean reversion, asset prices become volatile, the price-dividend ratio displays persistence, long-horizon returns become predictable and a risk premium emerges. Comparing the moments of the model with those in the data using confidence bands from the method of simulated moments, we show that our findings are robust to different assumptions on the system of beliefs and other model features. We depart from previous studies of asset prices under learning in that agents form expectations about future stock prices using past price observations.