Agricultural Commodity Price Volatility Models and Prediction Performance of GARCH Family Models

Abstract

Agricultural product price volatility leads to future retail price uncertainties for producers and consumers. This paper examines modeling and forecasting price volatility of four widely produced, highly household consumed and traded agricultural commodities, namely Teff, Wheat, Barley and Maize at Debre Markos of Ethiopia using time series retail price data from September 2010-August 2022. We compare the performance of GARCH family models against different error distributions for each price return and AIC, BIC,log likelihood and significant p-value were applied to identify the best fit GARCH family models and the results showed that ARCH(1,1)GED for Teff and Barley and EGARCH(1,1)STD for Wheat and Maize are appropriate models. Moreover, based on the sign and magnitude of the parameters (coefficients of residuals) there is asymmetry in the news, in which bad news has larger effect on volatility than good news for two cereals price returns. Furthermore, the forecasting performance of the models are evaluated using the root mean square error and mean absolute errors and four weeks ahead prediction performance in MAE are 0.0082, 0.0066, 0.0065 and 0.0144 for Teff, Wheat, Barley and Maize return models respectively which reveals that they perform well. They can be concluded that GARCH(1,1)GED for Teff and Barley and EGARCH(1,1)STD for Wheat and Maize are more accurate price return volatility forecast for risk management. Therefore, consumers and producers are recommended to use these models for future price predictions.