Do foreign exchange return regressions convey useful information on return predictability?

This paper shows the possibility that the estimates from foreign exchange return regressions contain huge noise which makes it difficult to extract useful information about the predictability of foreign excess returns, in particular, if exchange rates are generated from a typical present value model with a near unity discount factor. The main reason is that the present value model induces a large bias in the estimation of the regressions accompanied by a high variability of the estimates. We also confirm that the volatility and persistence of both the spot return and the forward premium generated from the present value model are consistent with data.