Volatility of clean energy and natural gas, uncertainty indices, and global economic conditions

Existing studies rely on exogenous drivers to improve the accuracy of the volatility forecasting of the least polluting fossil fuels, natural gas. However, the academic literature lacks a comprehensive analysis on the factors driving the volatility of clean energy stock indices, although the transition to renewable energy sources has resulted in the emergence of the clean energy sector as a fast-growing sector very appealing to environmentally responsible investors. This paper examines the ability of five uncertainty indices and seven global economic conditions to predict the realized volatility of clean energy stock and natural gas markets. Using daily return data of four exchange-traded funds to track the performance of the global clean energy stock market and natural gas prices, we construct the monthly realized volatility and apply various methods, including shrinkage methods. We find that both uncertainty indices and global economic conditions successfully predict clean energy realized volatility. The shrinkage methods consistently outperform dimensionality reduction methods and combination forecast methods for clean energy and natural gas. Whether clean energy ETFs or natural gas are used, the predictive information extracted from global economic conditions outperforms uncertainty indices using shrinkage methods. This suggests that real economic activities rather than text-based measures of uncertainty should be considered when investors and policy-makers analyze the volatilities of clean energy and natural gas. We further investigate the forecasting power of uncertainty indices and economic conditions in terms of variable selection, market conditions, and seasonality effects. •This study mainly investigates the volatility of natural gas and clean energy stock.•Both uncertainty indices and global economic conditions successfully predict clean energy realized volatility.•The shrinkage methods consistently outperform dimensionality reduction methods and combination forecast methods.•Real economic activities rather than text-based measures of uncertainty should be considered.