We propose a new explanation for the forward-premium and the delayed-overshooting puzzles. Both puzzles arise from a systematic under-reaction of short-term interest rate forecasts to current innovations. Accordingly, the forward premium is always a biased predictor of future depreciation; the bias can be so severe as to lead to negative coeffcients in the 'Fama' regression; delayed overshooting may or may not occur depending upon the persistence of interest rate innovations and the degree of under-reaction; lastly, for G-7 countries against the U.S., these puzzles can be rationalized for values of the model's parameters that match empirical estimates.

"When central banks set nominal interest rates according to an interest rate reaction function, such as the Taylor rule, and the exchange rate is priced by uncovered interest parity, the real exchange rate is determined by expected inflation differentials and output gap differentials. In this paper I examine the implications of these Taylor-rule fundamentals for real exchange rate determination in an environment where market participants are ignorant of the numerical values of the model's coefficients but attempt to acquire that information using least-squares learning rules. I find evidence that this simple learning environment provides a plausible framework for understanding real dollar--DM exchange rate dynamics from 1976 to 2003. The least-squares learning path for the real exchange rate implied by inflation and output gap data exhibits the real depreciation of the 70s, the great appreciation (1979.4-1985.1) and the subsequent great depreciation (1985.2-1991.1) observed in the data. An emphasis on Taylor-rule fundamentals may provide a resolution to the exchange rate disconnect puzzle"--National Bureau of Economic Research web site.

Abstract: My dissertation studies models of exchange rate determination that are standard in all respects except that market participants have incomplete knowledge about the economic structure, and employ adaptive learning rules to learn about the economic environment. My work on introducing model uncertainty into standard models is motivated by the well documented fact that when the underlying economic environment is known and is common knowledge to market participants, models under rational expectations cannot account for such basic features of the data as the relative volatility between exchange rate and fundamentals or the predictability of future exchange rate returns by the deviation of the exchange rate from the fundamentals. I partly response to this problem to use an alternative view of expectations, adaptive expectations, which can be a reasonable way to form expectations when the environment is excessively complex. I find that the model under adaptive expectations performs better than rational expectations in explaining why the forward exchange rate as a predictor of the future spot rate generates a large bias, why the maximal depreciation of a currency upon an unexpected monetary shock occurs some periods after the initial shock. I consider a standard monetary model where market participants learn about the economic structure using adaptive learning rules. While market participants are assumed to know the functional form of the stochastic process that drives the fundamentals, they do not know the parameter values which they assess by least squares learning. Market participants must also contend with unannounced regime shifts in the fundamental process. I compare the predictions of the model under adaptive learning to those generated under standard rational expectations and under adaptive expectations. I find that the model under adaptive learning dominates the alternative specifications of expectations in its ability to account for why the fundamentals predict exchange rate returns over long horizons, for generating exchange rate return volatility in excess of fundamentals volatility, and in generating persistent deviations of the exchange rate from the fundamentals. I conclude that the underlying model uncertainty goes far in helping to resolve some longstanding puzzles in the foreign exchange market.

This paper considers a version of the Dornbusch model of exchange rate dynamics which allows a nonlinear domestic demand for foreign assets function and imperfect substitutability between domestic and foreign interest bearing assets. Expectations of exchange rate changes are modelled as adaptive with perfect foresight being obtained as a limiting case. For sufficiently rapid speed of adjustment of expectations the model is able to generate cyclical behaviour of the exchange rate and expectations of its change. In the perfect foresight limit the cycles become relaxation cycles. To this underlying model of the fundamentals a white noise news process is added. Agents are assumed to attempt to learn about the system dynamics and the link between such learning and exchange rate volatility is studied. Two learning scenarios are considered. In the first scenario economic agents are regarded as a uniformly well-informed group of sophisticated traders. In the second scenario a group of naive traders coexist with the sophisticated traders. We find that both learning scenarios lead to increased volatility. However this effect increases in proportion to the weight of the naive traders.