Modeling and Forecasting Exchange Rates Volatility Using Selected GARCH Models

Abstract

Modeling and forecasting foreign exchange rate volatility is a critical issue in various fields of finance today. Volatility of the Nigeria foreign exchange market was studied using the generalized autoregressive conditional heteroskedasticity (GARCH), exponential GARCH (EGARCH), threshold GARCH (TGARCH) and integrated GARCH
(IGARCH) models. In the study, the exchange rates of Nigeria Naira and four other country’s currency were compared. Based on daily data of the exchange rates of Nigeria Naira and respective US Dollar (USD), EURO (EUR), Pound Sterling (PDS) and the Japanese Yen (YEN), the tests for asymmetry and normality were evaluated using skewness, kurtosis and Jarque-Bera statistics. It was observed that the data confirm non-normality and asymmetry. This is evidence from the descriptive statistics including readings from the skewness, kurtosis and Jarque-Bera statistics. All the exchange rates were statistically significant at 5% level of significance except for the USD which is statistically insignificant. Also, in all the selected GARCH models, the AC, PAC and Q-Statistics show no presence of autocorrelation while volatility of the returns series and leverage effect are positively correlated at 5% confidence level except for the DYEN return series which is negative and significant. The model performance is assessed by checking the lowest Akaike Information Criterion (AIC) and it can be deduced that the IGARCH model is preferred for estimating daily return series for USD, PDS and EUR whereas TGARCH is seen to be appropriate for YEN during the study period.