Is Sign-Only Forecast of Exchange Rate Easier Than Forecasting Both Sign and Size?

Journal of International Trade & Commerce, Vol. 13, No. 5, pp. 25-49, October 2017

25 Pages Posted: 8 Nov 2017

See all articles by Hae-Sun Park

Hae-Sun Park

Konkuk University - Division of Business Administration and Economics

Jong-Byung Jun

Suffolk University - Department of Economics

Kyeong-Won Yoo

Sangmyung University - Department of Economics and Finance

Date Written: October 24, 2017

Abstract

Forecasting exchange rate movements is extremely difficult. While the usual forecast requires determining the size and sign of change, we investigate if the direction of change alone is easier to forecast. The accuracy rate of monthly forecasts based on an economic model is compared with random guessing that has the expected accuracy rate of 50%. Diebold-Marino test is used to see if the advantage of the economic model is statistically significant. We find that even a simple linear model based on UIP can predict the direction of change better than random guessing, even if the model’s performance is worse than random walk in terms of root mean squared error (RMSE) which is a widely used measure of the average size of the forecast errors, Binary response variable models such as logit and probit do not seem to improve the accuracy of directional forecasts. We also find that the forecast accuracy is quite sensitive to the in-sample size of rolling regressions and the treatment of the observations with no change. These findings indicate that sign-only forecasting may be a reliable tool for qualitative decision-making of a foreign exchange market dealer or an exporter interested in hedging exchange risk.

Keywords: Binary Response Variable Model, Exchange Rate, Forecast, Simple Linear Model

JEL Classification: C22, C52, F31, F37

Suggested Citation

Park, Hae-Sun and Jun, Jong-Byung and Yoo, Kyeong-Won, Is Sign-Only Forecast of Exchange Rate Easier Than Forecasting Both Sign and Size? (October 24, 2017). Journal of International Trade & Commerce, Vol. 13, No. 5, pp. 25-49, October 2017. Available at SSRN: https://ssrn.com/abstract=3066507

Hae-Sun Park (Contact Author)

Konkuk University - Division of Business Administration and Economics ( email )

Seoul, 27478
Korea, Republic of (South Korea)

Jong-Byung Jun

Suffolk University - Department of Economics ( email )

8 Ashburton Place
Boston, MA 02108
United States

Kyeong-Won Yoo

Sangmyung University - Department of Economics and Finance ( email )

Seoul, 03016
Korea, Republic of (South Korea)

Register to save articles to
your library

Register

Paper statistics

Downloads
23
Abstract Views
160
PlumX Metrics