Rethinking What Survey Data has to Say about the Role of Risk and Irrationality in Currency Markets

A number of studies have used survey data on traders' exchange rate forecasts to examine the role of risk and non-REH forecasting in accounting for excess returns in currency markets. This work re-examines those results using an alternative estimation technique, the Cointegrated VAR, which allows for better examination of non-stationarity in a multivariate framework. The results demonstrate the importance of focusing on the persistence of deviations from any found relationships. Consistent with some later studies, clear evidence of a time-varying risk premium is found, and REH is rejected for all three exchange rate samples examined (BP/USD, DM/USD, and JY/USD). The results strongly draw into question though the interpretation that this represents obvious irrationality. The relationship between the forecast error and interest rate differential is found to be non-stationary at very high significance levels, implying that the correlations are spurious and unstable over time, and individuals are not, in fact, mis-forecasting in a fixed manner relative to interest rates.