Modeling and Forecasting Nigeria Exchange Rate Volatility Across Major Global Economic Blocs

This study models the volatility dynamics of weekly exchange rate returns for across major global economic blocs against the Nigerian naira using models from the generalized autoregressive conditional heteroscedasticity (GARCH) family. The currencies analyzed are the British pound sterling, CFA franc, euro, United States dollar, Japanese yen, West African Unit of Account, Saudi Arabian riyal, and Swiss franc. Both symmetric GARCH(1,1) and asymmetric EGARCH(1,1) specifications were employed to capture volatility clustering, persistence, and leverage effects. The estimated GARCH(1,1) models for the CFA franc and Japanese yen yielded variance equations of the form σ²t = 0.1997 + 0.8009ε²t-1+ 0.6564σ²t-1, for the CFA franc (α + β = 1.4573), indicating explosive volatility, and σ²t = 0.00004 + 0.2594ε²t-1 + 0.6079σ²t-1 for the Japanese yen (α + β = 0.8673), indicating high but mean-reverting persistence. For the remaining currencies, the preferred EGARCH(1,1) models produced statistically significant coefficients, with persistence parameters close to unity, such as 0.8916 (POUNDS), 0.9334 (EURO), 0.9277 (DOLLAR), 0.9838 (WAUA), 0.7464 (RIYAL), and 0.9231 (SWFRANC). The estimated leverage parameters were positive across all series, suggesting that positive shocks generate lower subsequent volatility than negative shocks of equal magnitude. The Akaike Information Criterion indicates that EGARCH(1,1) is more suitable for modeling weekly exchange rate volatility for the pound sterling, euro, U.S. dollar, WAUA, riyal, and Swiss franc, while GARCH(1,1) is preferred for the CFA franc and Japanese yen. Out-of-sample forecasts show that the conditional variances converge to their respective long-run levels for most currencies, except the yen. The results reveal substantial volatility persistence in Nigeria’s foreign exchange market and underscore the usefulness of asymmetric GARCH models for exchange rate risk assessment and forecasting.