| Abstract: | We examine long-run relations implied by covered interest parity (CIP) in possibly cointegrated and nonstationary data series. Empirical evidence suggests that, ignoring market imperfections, CIP failed over January 6, 1984, through December 6, 1991, using weekly data from four major currencies relative to the U.S. dollar. The multivariate maximum likelihood vector autoregressive (VAR) methodology does not require data differencing and hence retains valuable information lost in previous research examining international market flows. Rejections are robust to both subperiod analysis and alternative interest rate series. Although test rejections are highly statistically significant, attainable economic profits appear small. Practitioners will find economic profits inconsequential relative to reasonable bounds on market frictions such as transaction costs. Nonetheless, the use of CIP to determine forward rates identically from interest rates and spot rates in academic studies is called into question. |
