Market efficiency and learning in an endogenously unstable environment

\tTraders in this model of an asset market have the opportunity to conduct individual research to acquire a noisy signal of a security's future value, or they can employ least-squares learning in an attempt at extracting the private information of other traders through observing the price. For a fixed proportion of the traders using fundamental research, n, the model converges to a stable fixed point equilibrium. At the fixed point, the regression traders outperform the fundamental traders for all values of n > 0. The equilibrium suffers from a Grossman and Stiglitz (1980) type paradox of efficient markets. Endogenize n based on performance and the Grossman-Stiglitz paradox is alleviated. The model is characterized by an unstable fixed point. As the model converges towards the fixed point, the regression traders perform well. As n falls, the regression traders begin to have a substantial impact on the price, causing greater fluctuations in profits and in n. Inevitably, the actual n is significantly different than the value of n implicit in the regression traders' coefficient values, introducing error in the regression trader's forecast. This leads to substantial mispricing that results in losses to the regression traders. It also throws the model far from the fixed point, starting the convergence process over.(This abstract was borrowed from another version of this item.)