Volatility prediction model based on EPU-GPR-VIX multidimensional portfolio

Accurate volatility forecasting is crucial for risk management, asset allocation, and macroprudential policy formulation. In 2022, amid multiple unprecedented shocks, the annualized volatility of the CSI 300 Index exceeded 21.6%. However, traditional macroeconomic indicators tend to lag, and existing uncertainty measures often fail to fully capture volatility driven by sudden shocks. To address this, we develop a multi-dimensional uncertainty framework integrating economic policy uncertainty, geopolitical risks, and market panic, and incorporate it into a GARCH-MIDAS model for volatility forecasting. Empirically, the combined CNEPU-GPRA-VIX model achieves superior out-of-sample forecast accuracy. This result is robust across model confidence set tests and various rolling window settings. Our findings indicate that Chinese stock market volatility is jointly driven by domestic policy uncertainty, geopolitical tensions, and global risk sentiment. These insights offer quantitative support for systemic risk monitoring and asset allocation decisions.