Forecasting volatility in sub-Saharan Africa’s commodity markets

Abstract

Using spot prices from eighteen commodities traded by most Sub-Saharan African countries, this paper evaluates the out-of-sample volatility forecasting efficiency of seven models. The models evaluated included random walk, simple regression and five models from the ARCH family of models. Standard loss functions are used to examine the relative performance of the competiting models. The non-ARCH family of models consistently outperformed the ARCH family of models on all the evaluation criteria. Of the two non-ARCH family of models, the autoregressive model was superior. The results of the study suggest that government agencies in Sub-Saharan Africa that manage inflows from commodity markets can use autoregressive models in predicting volatility of inflows. Again, risk management strategies will be best served with autoregressive models.