A standard empirical finding in international finance is that countries with high nominal interest rates experience appreciations of their currencies, in contrast to predictions based on uncovered interest parity (UIP). However, tests of UIP have almost exclusively relied on data on short-term interest rates. In this paper, UIP is tested on long-term government bond yields. Since the presence of coupon payments induces a measurement error between the observed data and true returns, several different proxies for the latter are constructed. Furthermore, instrumental variable techniques are used. In contrast to the typical finding, the results are rather favorable to UIP. Copyright 2001 by Blackwell Publishing Ltd.