Risk Premia in Forward Foreign Exchange Markets : A Comparison of Signal Extraction and Regression Methods

We investigate time varying risk premia in forward dollar/pound monthly exchange rates over the last two decades. We study this issue using a signal plus noise model and separately using regression techniques. Our models account for time varying volatility and non-normalities in the observed series. Our signal plus noise model fails to isolate a statistically significant risk premium component whereas our regression model does. We attribute the discrepancy in the results from the two methods to the low power of the signal plus noise model in discriminating between a time varying risk premium component and a serially uncorrelated spot exchange rate expectational error. An important reason for the low power of the signal plus noise model is its failure to use information on current period forward rates in extracting the risk premium