Assessing The Effectiveness of the Garch Model in Forecasting Volatility of Headline Inflation in Namibia

This study assesses the effectiveness of various volatility models, the GARCH(1,1), TGARCH, and EGARCH in forecasting inflation volatility in Namibia, using data from January 2003 to December 2024. Our finding shows that headline inflation volatility in Namibia exhibits persistence and mean-reversion, which is a common feature of volatility. Employing multiple metrics and in-sample criteria such as AIC and Schwarz, the GARCH(1,1) model emerged as the preferred choice for in-sample estimation. However, diagnostic tests revealed persistent serial correlation in residuals across all models, indicating possible limitations of using univariate models to explain inflation volatility. Out-of-sample forecast evaluation from October to December 2024 identified the EGARCH model as the most accurate, with superior performance across RMSE, MAE, and MAPE metrics. Further, this study shows that recent inflation shocks significantly impact volatility, with little evidence of asymmetric effects from positive or negative shocks. The research highlights the importance of ongoing model monitoring and evaluation. Volatility models designed for short-term forecasting need to be adjusted so that their parameters accurately reflect the relevant dynamic structure of the variable predicted by that model. Further investigation into the serial correlation issue offers valuable insights about how this can be accurately captured, thereby designing a model for policymakers aiming to stabilise inflation and manage it effectively, and mitigate the risks.