基于已实现半协方差的投资组合优化

在传统的风险度量方法中,常见的协方差估计量并未区分资产收益的下侧风险和上侧收益,而一般的下偏矩估计量则存在非对称性和难以加总的缺点.本文引入已实现半协方差矩阵(RSCOV)作为风险度量进行波动率预测和投资组合研究.本文将RSCOV应用于两种常见的风险分散投资策略—风险平价(ERC)策略和全局方差最小(GMV)策略,并将机器学习中的在线加权集成(OWE)算法用于提升已实现波动率预测方法HAR-RV的样本外预测表现.通过研究发现,相比起已有的其他风险衡量方式,仅包含负向波动信息的下半RSCOV能够更好地被用于平衡组内各资产的风险贡献.基于A股市场2011-2018年的高频数据,本文通过实证研究发现,OWE-HARRV在月度预测步长下的效果优于HAR-RV,而下半RSCOV则能够使ERC策略以及GMV策略在保证一定平均收益的同时,降低了组合收益的极端损失.

Portfolio optimization based on realized semi-covariance

Among traditional volatility measurements, normal covariance estimators are not able to distinguish the downside risk and upside gains of asset return, while traditional lower partial moment estimators are asymmetric and impossible to sum up. Therefore, this paper introduces a new risk measurement called realized semi-covariance (RSCOV) to conduct volatility forecasting and portfolio optimization. Based on decomposition of realized covariance matrix, we test it on two common diversification investing strategies, equally-weighted risk contribution (ERC) strategy and global minimum variance (GMV) strategy. To perform forecasting, we adopt online weighted ensemble (OWE) algorithm in machine learning domain to boost the out-of-sample performance of HAR-RV. Compared to existing covariance or realized covariance, we find that realized downside semi-covariance matrix, that only contains information about negative volatility, can be used to better balance the risk contribution of assets in portfolio. Then, using high-frequency data of A share market spanning from 2011 to 2018, empirical result shows that our OWE-HAR-RV can outperform HAR-RV in monthly prediction. Lower RSCOV can be applied to ensure risk parity and minimum variance portfolio strategies to achieve better allocated asset weights and lower maximum loss while maintaining certain portfolio return.