Heterogeneous Gain Learning and Long Swings in Asset Prices

This paper considers the impact of heterogeneous gain learning in an asset pricing model. A relatively stylized model is shown to generate persistent swings of asset prices from their fundamental values which replicates long range samples of U.S financial data. The detailed mechanisms of the learning models are then explored. Evidence suggests that agents' perceptions of risk and its dynamics and persistence are important in generating appropriate price/fundamental dynamics. Agents putting large amounts of weight on the recent past in their volatility models control a large fraction of wealth, and are important in perpetuating the volatility magnifying dynamics of the market.