Forecasting FOREX Volatility in Turbulent Times

The present study is an attempt to evaluate the predictability of the foreign exchange volatility in thirteen countries. The data covers the period of 2005-2009. To effectively forecast the volatility in the exchange rates, a GARCH model is used. The study compares the results between crisis period and a set of normal periods. The empirical results reveal that almost all countries except Thailand witnessed non-existence of volatility shocks at least once in a three year pre-crisis period but all the sample countries had volatility shocks in the crisis period of 2008-09. This apparently indicates that forecasting can be made at least for the next day given the high degree of volatility in the crisis period. The paper also reveals that exchange rates tend to have persistent conditional heteroskedasticity, and hence, could be predicted with one lag term