Uncovered Interest Parity: Are Empirical Rejections of It Valid?

There is a vast empirical literature rejecting uncovered interest parity(UIP) on the basis of regressions of the actual exchange rate change against the forward premium/discount. In this paper, whilst we confirm the conventional regression analyses, we argue that they constitute only an indirect test of UIP and that there are serious econometric flaws in such regressions that make them an unreliable means of testing for UIP. Instead, we propose a two new profitability based tests of the UIP condition based on actual dollar returns of being in the domestic and foreign currency and we find evidence that in fact the UIP condition in fact seems to be holding for three of the four parities studied even though the conventional test would have rejected UIP in all four cases. Not only do our economically more meaningful profitability based tests lead us to accept the UIP condition for three of four currencies studied but they also seem to offer superior econometric properties compared to the conventional regression analyses.