The intertemporal capital asset pricing model (ICAPM) has been widely used as a vehicle to explain the predictability of excess returns in forward foreign exchange markets. This approach ignores the heterogeneity in the assets actually traded in international asset markets. In this paper it is shown that the latent variable risk premium model is not rejected by the data when tested on a portfolio including equity and bonds in four currencies. However, application of a more general test rejects parameter stability, suggesting that the latent variable model is not as successful as the previous tests would imply.