A multivariate GARCH model of risk premia in foreign exchange markets

This paper investigates the existence of time-varying risk premia in deviations from uncovered interest parity based on the market capital asset pricing model. The empirical analysis is conducted using a broad data set of seven major currencies against the US dollar, and a world equity index in order to approximate the benchmark portfolio. The conditional covariance matrix of excess returns is modelled as a multivariate GARCH process. The results indicate significant conditional systematic risk. Estimated conditional beta coefficients are Very similar across currencies and behave uniformly over time. The explanatory power of the model is significantly higher compared to the constant beta CAPM specification. Furthermore, estimation results suggest that (1) expected excess returns are less volatile in foreign exchange markets compared to stock markets, and (2) including nominal dollar assets in international equity portfolios can reduce overall portfolio risk.