forthcoming. Learning about risk and return: A simple model of bubbles and crashes. American Economic Journal: Macroeconomics

This paper demonstrates that an asset pricing model with least-squares learning can lead to bubbles and crashes as endogenous responses to the fun-damentals driving asset prices. When agents are risk-averse they generate fore-casts of the conditional variance of a stock’s return. Recursive updating of the conditional variance and expected return implies two mechanisms through which learning impacts stock prices: occasional shocks may lead agents to lower their risk estimate and increase their expected return, thereby triggering a bub-ble; along a bubble path recursive estimates of risk will increase and crash the bubble. JEL Classifications: G12; G14; D82; D83 Key Words: Risk, asset pricing, bubbles, adaptive learning. Thus, this vast increase in the market value of asset claims is in part the indirect result of investors accepting lower compensation for risk. Such an in-crease in market value is too often viewed by market participants as structural and permanent... Any onset of increased investor caution elevates risk premi-ums and, as a consequence, lowers asset values and promotes the liquidation of the debt that supported higher asset prices. This is the reason that history has not dealt kindly with the aftermath of protracted periods of low risk premiums. Alan Greenspan (2005). ∗We thank Bruce McGough for many valuable discussions, Ken Kasa and Bruce Preston fo