Modelling Volatility of Naira/US Dollar Exchange Rate Dynamics Using Conditional Heteroskedasticity Models with Non-Gaussian Errors

This study searches for optimal symmetric and asymmetric Conditional Heteroskedasticity (ARCH/GARCH) models that best fit and model volatility between Nigeria Naira and United States Dollar exchange rate dynamics in Nigeria using non-Gaussian errors. The study utilizes daily closing Naira/US Dollar exchange rate data from 12/11/2001 to 12/01/2017 making a total of 3665 observations. Symmetric ARCH and GARCH, as well as asymmetric EGARCH and TGARCH specifications were used to model the log return series in the presence of student-t innovations and Generalized Error distribution. Results show that symmetric ARCH (3) and basic GARCH (1,1) with student-t innovations as well as asymmetric EGARCH (1,1) with GED distribution and TGARCH (1,1) with student-t innovation were the best fitting models for the Naira/US Dollar exchange rate log return series. All the estimated models were found to be unstable and non-stationary indicating over persistence of volatility shock in the conditional variance. The asymmetric EGARCH (1,1) and TGARCH (1,1) models show supportive evidence for the existence of asymmetry and leverage effects suggesting that negative shocks produce more volatility in Nigerian foreign exchange market than positive shocks of the same magnitude. The study provides policy recommendations for traders and investors in Nigerian exchange market.