Stock market volatility prediction: Evidence from a new bagging model

Abstract

The purpose of this study is to investigate which model can improve the precision of the categorical economic policy uncertainty indices in predicting volatility in the U.S. stock market. In this study, a new model is constructed by combining autoregressive model and bagging method. The empirical outcomes indicate that machine learning models outperform traditional forecasting models and that the new model constructed in this study has the best forecasting ability. We perform robustness tests using an alternative stock index, alternative forecasting windows, and different economic cycles. The results show that these findings are robust. We hope to provide new insights into the application of the bagging method in stock market volatility forecasting.