| Summary: | 碩士 === 國立臺北大學 === 經濟學系 === 102 === This study serves a papers to adopt uncovered interest parity (hereafter UIP),purchasing power parity (hereafter PPP) and monetary fundamental (hereafter MF) three theory to construct models for forecasting foreign exchange rate.Using these models to predict foreign exchange rate of JPY to NTD and compare Accuracy in different models and horizons.We collect monthly data include the spot exchange rate of JPY to NTD, the risk-free interest rate, the consumer price index, Taiwan's money supply M1A, Japan money supply M1 from 1994 to 2013.All of the data are from Datastream.
In this paper,we combine the following three parts.First, using general linear model proposed by Corte, Tsiakas (2011) and BAO-JIE YOU(2013) to develope UIP,PPP and MF three models to predict the exchange rate in the 1-month horizon.Second, comparing in these three models in 1-,3-,6-,12-month horizons proposed by Nelson C. Mark (1995) and Nelson C. Mark, Doo-Yull Choib (1997).Third, adopting RMSE and MAE as the measurement method and let the random walk model as a benchmark to measure predictive capability proposed by Meese and Rogoff(1983).So this study uses in-sample observations to constuct the models and uses out-of-sample data to compare the accuracy in different models and horizons.
Empirical result shows that UIP model whether in 1-,3-,6-,12-month horizons are unable to predict the future exchange rate movements.
MF model can catch future exchange rate movements in 6-,12-month horizons and horizon increase with the explanatory power;regardless of any horizons PPP model has great explanatory power to the future exchange rate movements.Like the conclusion of Nelson C.Mark(1995),the models' constructed by PPP and MF explanatory power increase as the horizon increase.The inspection of predictive accuracy, it shows that UIP model is relatively ineffective but PPP model and MF model are relatively effective than randon walk model in every horizon.The accuracy increase as the horizon increase.That is why Nelson C. Mark(1995) says "White Noise" increase the error and decrease accuracy in the short horizon, but in the long horizon the foreign exchange rate can be successfully predicted.
|
|---|
