Comparative Analysis of Exchange Rate Forecasting Models:

Abstract

This dissertation critically evaluates the effectiveness of the random walk model for daily ex- change rate forecasting against some standard models, including ARIMA, GARCH, and a more advanced model, which is SETAR. The study aims to assess the performance of these models in exchange rate modeling and to compare their performance in out-of-sample forecasting across short, medium, and long-term horizons. The analysis utilizes the root mean square error in measuring the predictive accuracy of these models for daily exchange rate data for various currency pairs, such as GBP/SAR, GBP/KWD, and GBP/EUR. The evidence indicates that the RW model is very efficient at short-run daily exchange rate forecasting; most of the time, it performs as well as, if not better than, the more complicated counterparts. This is in line with the theoretical underpinning of the RW model: If no trends or patterns are observed, the current price can be considered the best predictor for fu- ture prices. However, for medium- and long-term forecasting of daily exchange rates, the study finds that models like ARIMA ensure better accuracy by capturing the underlying trends and seasonality. On the same note, the GARCH model is good at modeling time-varying volatility, while the SETAR model has a special edge in the capturing of multi-linear dynamics and regime shifts.