Forecasting global stock market volatility: The impact of volatility spillover index in spatial-temporal graph-based model

Abstract

The shocks on certain market spread to other markets due to the financial linkages of global economy, which is known as volatility spillover effect. In this study, we propose a volatility forecasting model for global market indices using the spatial-temporal graph neural network (GNN). The volatility spillover between markets are reflected in the model by estimating the linkage between markets, which is the input of GNN, using the volatility spillover index. An empirical analysis is conducted on eight representative global market indices. From the out-of-sample results, we found the following features. First, the proposed spatial-temporal GNN spillover model outperforms the benchmark models in short- and mid-term forecasting. Second, the forecasting accuracy highly depends on the inclusion of the market index with a high volatility spillover effect. Including S&P500, which contains the highest net spillover index, effectively helps to forecast the volatilities of other markets. Third, the investor can gain economic gain by using predicted volatility from proposed model in the mean-variance framework.