The Excess Returns Puzzle in Currency Markets: Clues on Moving Forward

This paper is designed to review the empirical literature on the excess returns puzzle: the difficulty encountered by standard risk premium models in accounting for relative returns in the foreign exchange market. Of particular interest are the studies using survey data to decompose ex post excess returns into an expected component - the risk premium - and a forecast error. On the whole, these studies have found evidence of violations of the rational expectations hypothesis (non-white noise forecast errors), and a time-varying risk premium. This suggests the need for an alternative specification of forecasting. The literature has left open however the question of whether the traditional models can account for movements in the premium as measured by survey. Although the traditional models have not been tested against survey data, there is reason from the outset to believe EUT provides a deficient foundation for a model of the risk premium. Experimental evidence is discussed showing that the predictions of EUT are grossly inconsistent with the behavior of actual subjects towards risky gambles. Lastly, the paper discusses alternative models of risk preferences drawing from the experimental findings on prospect theory, and ways in which their testing can be improved through use of the I(2) Cointegrated VAR model.