Stock exchange volatility forecasting under market stress with MIDAS regression

This paper presents two different approaches of volatility forecasting. One is based on option‐implied volatility (IV), the other involves conducting time series methods using historical volatility. With that purpose, we study eight developed stock markets, offering implied volatility indexes for the 2008 financial crisis. We evaluated the 1 month out‐of‐sample volatility forecast performance of two statistical‐based models, Mixed Data Sampling (MIDAS) and GARCH, and compared the results with option‐implied volatility indexes. Our results suggest that MIDAS produce superior forecast performance compared to GARCH model and IV method. While options are not available for all assets, we believe that MIDAS model can be a sophisticated tool for researchers and analysts to forecast future volatility with its ability to process high‐frequency data.