Currency risk and imperfect knowledge: Cointegrated VAR analyses with survey data.

摘要

Much progress has been made in understanding excess returns in the foreign exchange market through the use of survey data on traders' exchange rate forecasts. On the whole, this literature, which is reviewed in chapter 1, has found that excess returns derive from both violations of the rational expectations hypothesis (non white-noise forecast errors) as well as a time-varying risk premium. What this literature has not done however is to determine whether any of the existing models of the risk premium can account for the time-varying risk premium found in survey data. The second and third chapters use the Cointegrated VAR model to test the Capital Asset Pricing Model (CAPM), the Consumption CAPM, and the Keynes-Imperfect Knowledge Economics (IKE) gap model, which relate the risk premium to the exchange rate's variance, covariance with consumption, and deviation from Purchasing Power Parity respectively. The strongest support is found for the Keynes-IKE gap model. The analysis of this model is then extended in chapter 4 to the I(2) CVAR framework, which is a unique empirical approach designed to account for data which undergoes persistent changes over time without the need for data transformations which cause a loss of information. The I(2) model also allows for more rigorous testing of the theory and a better examination of the dynamics between the exchange rate, expectations, prices, and interest rates. The Keynes-IKE gap model still performs quite well. Further, persistent changes are found for the real exchange rate in several instances, which is problematic for standard REH theory but fully compatible with the IKE theory.