The role of sudden variance shifts in predicting volatility in bioenergy crop markets under structural breaks
Abstract
Forecasting bioenergy feedstock commodity volatility has received significant attention due to its importance in biofuel production and household consumption. Several extreme events, including the COVID-19 pandemic, have sparked interest in studying the role of structural breaks on volatility modeling and prediction in these markets. This study extensively examines the prediction performance of econometric models at multiple horizons using a rolling-window approach, with and without accommodating structural changes. We exploit the ICSS algorithm to determine the in-sample estimation windows to accommodate structural breaks. We extend the procedure beyond GARCH-class models. Also, the detected break information defines the regime dummies. The study innovatively evaluates the prediction performance of specific GARCH-class models by incorporating binary variables for sudden shifts in unconditional variance. Our findings reveal that accounting for the endogenously detected structural breaks through the dummy variables leads to considerable forecast accuracy gains.