Testing the predictive power of various exchange rate models in forecasting the volatility of exchange

Abstract

This Thesis tests the predictive power of ARCH, GARCH and EGARCH models in forecasting exchange rate volatility of Canadian dollar, Euro, British Pound, Swiss Franc and Japanese Yen using the US dollar as the base currency. We investigate both in-sample and out-of-sample performance of the volatility models using loss functions. The study further examines if the best model for the in-sample forecast will emerge as the best model for the out-of-sample forecast. The study finds that the GARCH(1,1) model outperforms all the other volatility models during the in-sample period. However in terms of the out-of-sample performance of the volatility models, the results are inconclusive, even though the ARCH model performed better most of the time than the complex models. The study concludes that the simple models should be given special consideration in terms of forecasting. Our results are robust to research on exchange rate volatility forecasting.