Over the course of three main chapters, my dissertation research investigates the relationship between exchange rates and monetary fundamentals under the present value model and provides validations of the present-value model for exchange rates with newly developed econometric techniques.
The first essay, “A Bootstrap Granger Causality Test between Exchange Rates and Fundamentals,” uses a bootstrap method to look for reliable evidence of the Granger causality relationship from exchange rate to fundamentals. This essay is in response to the finding of the statistically significant Granger causality relationship based on the asymptotic test procedure in the previous literature, which is taken as evidence for the present value model for the exchange rate. However, bootstrap test results show evidence against the finding in the previous work. Monte Carlo experiments suggest that the asymptotic Granger causality test in the previous study tends to spuriously reject the null hypothesis, and the evidence based on the asymptotic test for the present-value model is very likely due to small-sample problems. Thus, the evidence of the present-value model for exchange rates based on the bootstrap implementation is weak.
The second essay, “Exchange Rate Predictability under the Present Value Model,” uses the Monte Carlo simulations to explore exchange rate predictability under the present-value model so as to evaluate the implication of the present value model. This paper discusses how forces from fundamentals and the discount factor in the present-value model affect exchange rate change over different time horizons in the opposing directions. The simulation results show that highest rejection rates of the out-of-sample statistic are mostly found in the medium forecast horizon. This finding suggests that the force from the discount factor dominates the out-of-sample exchange rate change in the middle forecast horizon, and also implies that exchange rate predictability does exist in the present-value model.
The third essay, “k-Quarter Differences Exchange Rate Predictability with an Adjusted Out-of-Sample Test Statistic,” uses the real data to evaluate out-of-sample exchange rate predictability implication of the present-value model with an adjusted test statistic. The adjusted test statistic is designed to prevent the bias in the out-of-sample traditional statistic for the nested models. The results show that the exchange rate is predictable mostly at middle forecast horizon based on both the asymptotic distribution and the bootstrap distribution. The empirical findings are consistent with the implication of the present-value model found in the second essay, and therefore provide supporting evidence for the present-value model.