Assessing the forecastibility of ESTAR Models: Evidence from ringgit/yen Rate

The purpose of this paper is to contribute to the debate on the relevance of nonlinear forecasts in the financial markets. To that end, this study forecasts the yen-based ringgit by using the Exponential Smooth Transition Autoregressive (ESTAR) model. When formally assessed for forecast accuracy, th...

Full description

Bibliographic Details
Main Authors: Liew, Venus Khim-Sen, Lau, Evan, Ahmad Zubaidi, Baharumshah
Format: Article
Language:English
Published: IUP Publications 2007
Subjects:
Online Access:http://ir.unimas.my/id/eprint/18622/
http://ir.unimas.my/id/eprint/18622/7/Assessing%20the%20Forecastibility%20of%20ESTAR%20Models%20%28abstract%29.pdf
Description
Summary:The purpose of this paper is to contribute to the debate on the relevance of nonlinear forecasts in the financial markets. To that end, this study forecasts the yen-based ringgit by using the Exponential Smooth Transition Autoregressive (ESTAR) model. When formally assessed for forecast accuracy, the results reveal that the ESTAR out-of-sample predictors statistically outperform both the linear AR and random walk models at standard significant levels. The hypothesis of equal forecasting accuracy between ESTAR models and the random walk model is formally rejected based on the Fisher sign test. This paper offers some evidence on the ability to forecast exchange rates using nonlinear methods. Hence, we conclude that linear models are not always the optimal for forecasting exchange rate as there is some forecast accuracy that can be gained by considering the nonlinearity inherent in the exchange rate.