Forecasting exchange rates : an empirical investigation of advanced, emerging and frontier market economies

The purpose of this research is to investigate the application of different forecasting methods to predict the exchange rates of advanced, emerging and frontier market economies. To date, research on forecasting exchange rates has tended to focus mostly on advanced economies. Little attention has been paid on emerging and frontier market currencies and this research fills a major gap in the literature. Data are drawn from International Financial Statistics, monthly publications by the International Monetary Fund. Monthly data pertaining to 49 countries from 1972 M1 up to and including 2007 M12 are used for model derivation. The remaining observations i.e. 2008 M1 to 2010 M4 are held back for the purpose of out-of-sample forecast evaluation. The Lee and Strazicich (2003) unit root test was applied to examine the presence or otherwise of endogenous structural breaks. Three times series models, namely volatility, exponential smoothing, Naïve 1 plus a causal cointegration via ARDL (autoregressive distributive lags) model are used. Two- three- and four-way combinations of these four models are generated in an attempt to increase forecasting performance. The forecasting accuracy of all models is assessed via MAPE (mean absolute percentage error). Studies of forecasting exchange rates have used a variety of measures to assess forecasting accuracy. However, the MAPE is one of the most commonly used measures of error magnitude. This accuracy criterion has the advantage of being measured in unit-free terms. Granger Causality analyses are carried out to shed some light on the causal relationships between macroeconomic variables and exchange rate dynamics. The results show that single volatility models outperform other time series and a causal model in many of the emerging and frontier markets. These findings also provide additional evidence on leverage effects of advanced, emerging and frontier currencies exchange rates. Although statistically based forecast combination methods have not had much application in the field of exchange rate modelling, the results of this study show that such combinations often perform better than a single model for exchange rate prediction.

Name dc:type.qualificationname

phd

Level dc:type.qualificationlevel

doctoral

Grantor dc:publisher.institution

London Metropolitan University

Year dc:date.issued

2014

Author dc:creator

• Newaz, Mohammad Khaleq

• 330 Economics

Dc Identifier Grantnumber

N/A

OAI identifier oai:identifier

oai:repository.londonmet.ac.uk:699