Does the uncovered interest parity hold in short horizons?

Abstract

In this article, one of the contemporaneous monetary theories of exchange rate determination, namely uncovered interest parity (UIP), is examined. The UIP hypothesis assumes that if capital is perfectly mobile, then investors around the world will be indifferent between holding their portfolios in domestic or foreign securities because they obtain the same return from these assets. Based on a theoretical formulation, our ex post estimation results employing four developed countries exchange rates vis-à-vis US dollar indicate the failure of the UIP hypothesis using short-horizon interest differential and future spot exchange rate data in line with most empirical papers in the economics literature.

Notes

¹ A large literature also have emerged on why the slope coefficient in the regression of the change in the logarithm of the spot exchange rate on the forward premium is less than unity or even negative. As expressed in Beyaert et al. (2007), this would imply that investors in the foreign exchange market do not behave rationally since they would not take profit of predictable excess returns. See also, among many others, Froot and Thaler (1990), Lewis (1995) and Engel (1996) on this issue. However, some papers are able to give evidence in favour of the UIP theorem as well. See, for instance, Han (2004). Chinn and Meredith (2004) explain the contradiction between short- and long-term estimation results rejecting the validity of UIP theorem in the short-run but supporting it in long-horizon estimates in the sense that the long-horizon results differ sharply from the short-horizon results because the model's ‘fundamentals’ play a more important role in trying down exchange rate movements over longer horizons.

² Such as identical default risk, tax treatment, the absence of restrictions on foreign ownership, and negligible transaction costs.

³ All quoted interest rates are in percentages. Hence, following Sarantis (2006) the rates used in the estimation are measured by r = ln(1 + i/100). Besides, stock returns are measured by ln(Pt/Pt − 1), where P is the stock price index.

⁴ Our ex post estimation results not reported here reveal that the main findings obtained in this article do not sensitive to whether or not these latter instruments have been included into the model.

⁵ Following QMS (2004), the starting point of GMM estimation is a theoretical relation that the parameters should satisfy. The idea is to choose the parameter estimates so that the theoretical relation is satisfied as closely as possible. GMM estimation is based upon the assumption that the disturbances in the equations are uncorrelated with a set of instrumental variables. The theoretical relation is replaced by its sample counterpart and the estimates are chosen to minimize the weighted distance between the theoretical and actual values. GMM is a robust estimator in that, unlike maximum likelihood estimation, it does not require information of the exact distribution of the disturbances. For the GMM estimates reported in this article, we use the Newey and West (1987) weighting matrix, which ensures that the GMM estimates and their standard errors are robust to heteroskedasticity and autocorrelation of unknown form.

⁶ The smaller lagged values was also tried to be employed to the data. Although estimating similar results, the J-statistic expressed below to test the validity of overidentifying restrictions under the null hypothesis was rejected for small lagged values of the variables. But using 12-lags fitted to the statistical prerequisites.