| This paper uses a continuous time model with partially observable state variables to compare two alternative ways of filtering the hidden state. It assumes that decision makers infer the hidden state separately from their economic problem, by means of two alternative filtering procedures. The first filter, rational learning, extracts the unknown signal with a conventional nonlinear projection. The second procedure, learning with limited capacity, supposes that agents process information as if they were channels with limited capacity. I show how these filters affect asset pricing and optimal allocations in equilibrium. In particular, I show that the effect of the filters on risk exposure and average risk premium of the tangent potfolio is dependent on the mean of the filtered state. Learning with Limited Capacity: Implications in an Asset Pricing Model. |
