Exchange Rates and UIP Violations at Short and Long Horizons

The much-studied Uncovered Interest Rate Parity (UIP) puzzle, the observation that exchange rates do not adjust sufficiently to offset interest rate differentials, is more complicated than commonly understood. It changes nature with the horizon. I confirm existing short-run evidence that high interest rate currencies depreciate less than predicted by the interest rate differential, but, building on Engel (2012), at longer horizons (4 to 7 years) I find a reverse puzzle: high interest rate currencies depreciate too much. Interestingly, the long-horizon excess depreciation leads exchange rates to converge to the UIP benchmark over the long-run. To address the changing nature of the puzzle, I propose a novel model, based on the mechanism of bond convenience yields, that can explain both the short and the long horizon UIP violations. I also provide direct empirical evidence that supports the mechanism.