Forecasting realized volatility with spillover effects: Perspectives from graph neural networks

Abstract

We present a novel nonparametric methodology for modeling and forecasting multivariate realized volatilities using customized graph neural networks to incorporate spillover effects across stocks. The proposed model offers the benefits of incorporating spillover effects from multi-hop neighbors, capturing nonlinear relationships, and flexible training with different loss functions. The empirical findings suggest that incorporating spillover effects from multi-hop neighbors alone does not yield a clear advantage in terms of predictive accuracy. Furthermore, modeling nonlinear spillover effects enhances the forecasting accuracy of realized volatilities, particularly for short-term horizons of up to one week. More importantly, our results consistently indicate that training with the quasi-likelihood loss leads to substantial improvements in model performance compared to the commonly used mean squared error, primarily due to its superior handling of heteroskedasticity. A comprehensive series of empirical evaluations in alternative settings confirm the robustness of our results.