序列稀疏自回归方法及其在美股做空数据分析上的应用  

Sequential Sparse Auto-Regression withApplication to Stock Short Interest Data in US

作　　者：刘静[1] 余琴 吴捷 李阳[1] LIU Jing;YU Qin;WU Jie;LI Yang(University of Science and Technology of China,Hefei 230026;Anhui University,Hefei 230601)

机构地区：[1]中国科学技术大学,安徽合肥230026 [2]安徽大学,安徽合肥230601

出　　处：《财贸研究》2024年第1期60-70,共11页Finance and Trade Research

基　　金：国家自然科学基金项目“基于自适应投影方法的高维复杂数据统计分析及应用”(12101584)。

摘　　要：采用序列稀疏回归的思路来处理向量自回归模型,并设计适用于大规模时间序列数据分析的序列稀疏自回归方法。研究表明:从因子角度刻画向量自回归模型可以有效地将稀疏矩阵估计问题分解成稀疏奇异向量的估计问题,从而极大地提高了计算效率。以1523家美股上市公司1973年1月—2014年12月的做空数据为例,利用此方法探索公司之间的大规模做空关联网络。研究发现:此方法可以有效地恢复股票做空份额(即某一公司的空头股份数量)与股票收益率之间隐藏的关联网络,对于股票风险溢价研究具有一定启发意义。This paper adopts the idea of sequential sparse regression to deal with the vector auto-regressive model and develop the sequential sparse auto-regression for large-scale time series data analysis.The study shows that characterizing the regression model from the factor point of view is capable of dissolving the sparse matrix estimation problem into the sparse singular vector estimation problem,thus greatly improving the computational efficiency.Using the stock short interest data consisting of 1523 US listed companies as an example,we intend to explore the large-scale short interest association network among the firms through our method.Further research finds out that the proposed method can effectively uncover the underlying correlation network between the stock short interest(i.e.the number of shares held short in a given firm)and the aggregate stock return,which has certain enlightening significance for the study of stock risk premium.

关 键 词：向量自回归模型 关联性网络 稀疏建模 股票做空份额 大数据分析 

分 类 号：F830[经济管理—金融学]