ESG indicators and sectoral volatility: GARCH vs. hybrid machine learning

This study examines the influence of Environmental, Social, and Governance (ESG) factors on sectoral stock volatility using a combination of econometric and machine learning models. ESG indicators are incorporated into the mean equations of univariate and multivariate GARCH models to analyze sector-specific volatility dynamics and spillover effects. To capture time-varying correlations between sectors, we employ the Dynamic Conditional Correlation (DCC) MGARCH model, while a Long Short-Term Memory (LSTM) neural network is used to enhance return predictions by detecting nonlinear dependencies and long-range volatility patterns. Our findings indicate that ESG factors significantly affect sectoral volatility, with pronounced spillover effects in Financial and Industrial sectors. Although the hybrid MGARCH-DCC and LSTM framework provides deeper insights into ESG-driven volatility transmission, it does not outperform the MGARCH model in predictive accuracy. These results reaffirm the robustness of econometric approaches in modeling sectoral volatility while highlighting the growing importance of ESG factors in financial risk assessment. Investors and policymakers must recognize ESG-driven volatility spillovers when designing risk management strategies.