Extreme weather events as the main driver of electricity price volatility in Italy: A GARCH-MIDAS approach with machine learning-based variable selection

This paper investigates the impact of extreme weather events on electricity price volatility in Italy, employing a novel combination of advanced econometric techniques and a robust variable selection process. First, we provide empirical evidence that extreme weather events significantly predict electricity price volatility. We compile a comprehensive set of economic and financial variables known in the literature to influence electricity price volatility and apply the Best Path Algorithm (BPA) for variable selection, identifying the most relevant predictors. A Granger causality analysis of the selected variables confirms that extreme weather events not only emerge as the primary factor driving volatility but also exhibit a clear unidirectional causal relationship. Second, we integrate weather variables into a GARCH-MIDAS model, to combine high-frequency electricity price data with low-frequency climate data, thereby capturing both short- and long-term volatility components. Additionally, we incorporate external shocks—such as the Russia–Ukraine war—as exogenous variables to account for broader geopolitical influences on the energy market. Our findings underscore the substantial predictive power of extreme weather events and external shocks on electricity price dynamics. This study enhances forecasting capabilities for policymakers and energy stakeholders, highlighting the urgent need for resilient energy market planning. Future research may extend this methodology to other regions and incorporate additional variables to further improve predictive accuracy.