The Forward Premium Puzzle in a Model of Imperfect Information: Theory and Evidence

This paper studies the forward premium puzzle in an environment where private agents do not perfectly observe the shocks that drive monetary policy. Private agents optimally update their conditional expectations by means of the Kalman filter. The transition dynamics associated with Kalman filtering lead to fixed time-effects and conditional heteroskedasticity in the forward premium regression. I provide evidence for the presence of time-effects in the forward premium regression and find that the forward premium puzzle is significantly weakened. In particular, a 1 percent increase in the 1-month interest differential is expected to be accompanied by an additional 0.34 percent depreciation of the currency in the following month.