Improving stock market volatility forecasts with complete subset linear and quantile HAR models

作者：

Highlights：

• We design complete subset linear (CSLR) and quantile regression (CSQR) HAR models.

• Our approach is on the border of machine learning and standard econometric literature.

• Our sample covers four broad market indices: S&P 500, NIKKEI 225, STOXX 50, SSEC.

• CSLR and CSQR tend to outperform benchmark models: HAR-RV, HAR-SJ, HAR-SV, HAR-CJ.

摘要

•We design complete subset linear (CSLR) and quantile regression (CSQR) HAR models.•Our approach is on the border of machine learning and standard econometric literature.•Our sample covers four broad market indices: S&P 500, NIKKEI 225, STOXX 50, SSEC.•CSLR and CSQR tend to outperform benchmark models: HAR-RV, HAR-SJ, HAR-SV, HAR-CJ.

论文关键词：Volatility density,Complete subset regression,Forecasting,Quantile forecasts,HAR model,Stock market

论文评审过程：Received 15 June 2020, Revised 14 March 2021, Accepted 9 June 2021, Available online 18 June 2021, Version of Record 21 June 2021.

论文官网地址：https://doi.org/10.1016/j.eswa.2021.115416