Predicting exchange rates using Taylor rule fundamentals: evidence from a portfolio optimisation framework

Abstract

The paper studies exchange rate predictability using Taylor rule fundamentals in an optimal portfolio framework.The study seeks to link exchange rate dynamics with capital flows. Profit-seeking economic agents are assumed to repatriate funds across borders in response to differentials in rates of return from risky assets of portfolios held. We develop a uncovered portfolio return parity (UPRP) based exchange rate model in which changes in the short-term nominal exchange rate depend on the difference of optimal portfolio returns between two economies. In a two country economy where USA is taken as the foreign country we test the model in 5 countries namely South Africa, South Korea, Brazil, Mexico and Poland. The model is benchmarked against a UIP model and a Random walk model in order to establish whether the study’s extension enriches exchange rate prediction literature. We find that the main UPRP model outperforms the Random walk model in the 12 month horizon for 4 out of 5 countries using CW statistics. For the 1-month horizon the main model is outperformed by the Random walk model in 4 out 5 countries and for the 2-month and 3-month horizons the main model beats the Random walk using CW statistics. Theil’s U statistics also show that with the exception of South Korea, the main model beats the Random walk across all countries in the 3 and 12-month horizons. We conclude that out-of-sample exchange rate forecasting based on an optimal framework and Taylor rule fundamentals produces better nominal exchange rate forecasts relative to the Random walk model and UIP model