한국 외환시장의 균형환율 도출 및 환율예측에 관한 연구

This study examines the out-of-sample Won-Dollar exchange rate forecasts based on various economic models with monthly data from January 1999 to January 2018. Future exchange rate changes are predicted using the models and the fundamentals of PPP, UIP, and Taylor rule. In addition to traditional Taylor rule model, a general form of monetary policy reaction function is introduced in this paper. Monetary authorities, such as the Fed and BOK, might place different weights to positive and negative deviations of inflation or output gaps from their target values. In order to reflect the possibility of central banks’ asymmetric preferences or nonlinear Phillips curve, a general form of interest rate reaction function is derived and utilized for exchange rate prediction. The results show that PPP model and PPP/monetary fundamentals have better ability to predict future Won-Dollar exchange rates than the random walk model over the short horizons as well as the long horizons. However, UIP model and interest rate differential fundamentals have poor predictability over all horizons. Taylor rule model has significant ability to predict the direction of future exchange rate movements over the one-year horizon and has smaller adjusted mean squared prediction errors (MSPE) than random walk over the two-year horizon. Fundamentals of the nonlinear interest rate reaction function have higher predictive power in comparison with fundamentals of the Taylor rule when it comes to the Clark and West statistics.