The Forward Premium Puzzle is one of the most prominent empirical anomalies in international finance. The puzzle is that the forward premium predicts future exchange rate movements but typically with a sign opposite to that implied by rational expectations. The existing research focuses on three possible explanations for the anomaly---a risk premium, an econometric misspecification, and non-rational expectations. The first and the second chapters of this dissertation introduce the puzzle and discuss the existing literature. The third chapter rules out risk premium and econometric misspecifications as stand-alone explanations for the puzzle, and argues that the assumption of non-rational expectations is necessary. The fourth chapter models the non-rationality in terms of Recursive Least Squares Learning on the part of participating agents. The key assumption is that risk neutral agents do not have perfect knowledge about the foreign exchange market, but attempt to learn the parameters underlying the stochastic process generating the exchange rate using constant-gain recursive least squares. When exchange rate data are generated from the model and empirical tests are performed, the results replicate the anomaly under plausible sets of parameter values. The model thus appears to provide a compelling explanation for the puzzle. Chapter four also provides further intuitive insight by analyzing the learning dynamics. The fifth chapter furnishes empirical results strengthening support for the model provided in chapter four. The parameters used for simulation are estimated empirically. The empirical features of the simulated data are compared with actual data. More general processes generating the exchange rate fundamentals are also estimated and tested for possible structural changes which would justify constant-gain learning. Overall, the evidence supports most features of this learning explanation of the Forward Premium Puzzle.
展开▼
