Learning to Agree: A New Perspective on Price Drift

This paper introduces statistical learning in an asset pricing model of differences of opinions. I show that the model converges to the unique rational-expectation equilibrium and furthermore I show that asset prices drift predictably in its neighborhood. Accordingly, the model offers a unifying perspective between two so-far mutually exclusive strands of the asset pricing literature. Learning preserves all the desirable features offered by the rational-expectations hypothesis (i.e. the traders use efficiently both the private and the public information available) while yet implying asset prices that drift predictably in the ex-ante sense of Banerjee et al. (2009).