Volatility forecasting with an extended GARCH‐MIDAS approach

Abstract

This paper uses the generalized autoregressive conditional heteroscedasticity mixing data sampling (GARCH‐MIDAS) model to construct three types of extended models. Geopolitical risk uncertainty is included in the study as an introduced variable, and its impact on the Shanghai Stock Exchange (SSE) 50 index volatility is analyzed. The empirical analysis shows that the GARCH‐MIDAS‐RV‐EPU model with China's EPU is the best in predicting the volatility of China's stock market when the information of economic policy uncertainty (EPU) and geopolitical risk uncertainty (GPR) of other countries are included. When the common information model composed of China's economic policy uncertainty index and geopolitical uncertainty index is used to predict the volatility of the SSE, the model's prediction is better. Finally, when the model confidence set (MCS) and the interval length index that changes the forecast outside the sample are used to retest each conclusion, the results are very robust.