The relationship between the interest rate and exchange rate under certain conditions

Despite the plausibility of Uncovered Interest Parity (UIP), it is generally rejected by empirical studies. This puzzling fact has been referred to the UIP puzzle. In fact, there is no reliable relationship between the interest rate differential and the exchange rate change that will guide the practitioners in the international financial markets. Given the absence of any proven theory about the general relationship between the two variables, this paper attempts to find some stylized facts or conditional regularity in the short run that may not hold always but still can be useful under specific conditions.

We gather monthly data on nominal interest rates and exchange rates of nine countries obtained from International Financial Statistics (IFS) of International Monetary Fund database. Our sample includes five high interest rate countries and four Safe Haven countries with low interest rates, and the sample period is from January 1993 to June 2012. We use 3 month nominal interest rates when calculating interest rate differentials, and examine 3 month percentage changes in the spot rate. The U.S. dollar is the base currency for all exchange rates that involve the dollar, and the cross rates are calculated as the ratio of two dollar based exchange rates.

In this paper, we first test the linear relationship between the interest rate differential and exchange rate change for the currency pairs with large interest rate differentials. We take account of the serial correlation in the residuals by estimating our covariance matrix with the Newey-West estimator. Since some previous research suggests that the interest rate differential may play different roles in foreign exchange rate determination in different exchange rate volatility regimes, we also use a nonlinear structural threshold regression model to allow different regimes, where the conditional volatilities are measured by a Generalized Autoregressive Conditional Heteroskedasticity (GARCH) model.

We find less evidence of the UIP puzzle in the currency pairs with large interest rate differentials, no matter whether the high interest rate country is emerging or developed, and we get positive estimates of Fama’s beta coefficients for about three fourths of those currency pairs. Also, we expect to see some stylized nonlinear regression facts in different regimes under different foreign exchange policies.