上市公司股权质押风险特征识别与测度  

Identifying and Quantifying Share Pledge Risks of Listed Companies

作　　者：杨柳勇[1] 岑盛楠 贾少卿 YANG Liuyong;CEN Shengnan;JIA Shaoqing(School of Economics,Zhejiang University;Ningyin Wealth Management LLC)

机构地区：[1]浙江大学经济学院 [2]宁银理财有限责任公司

出　　处：《金融市场研究》2023年第12期72-80,共9页Financial Market Research

摘　　要：我国经历了两次大规模股权质押风险爆发,事前风控的缺位是重要原因。本文以股权质押事件作为研究样本,以“触及平仓线或发生冻结”作为标签,使用特征工程构造质押需求、治理结构、公司财务、二级市场等多维度特征,使用机器学习中的XGBoost算法对质押风险进行预测和分析,关注大股东股权质押风险的事前防控。实证结果表明,XGBoost算法训练得到的股权质押风险识别模型具有良好的预测性能,在测试集中召回率最高可达74.44%。此外,特征重要性分析表明,在股东性质方面,境内非国有大股东的股权质押风险更高。在治理结构方面,大股东持股比率越高,股权质押风险越高,而其他大股东持股比率的提升能抑制这种风险。在公司财务方面,股权质押风险较高的上市公司具有更高的成本费用率、更高的负债率和更差的现金流情况。在二级市场方面,股权质押风险较高的上市公司具有较低的市销率和流动性水平。China has undergone two rounds of large-scale share pledging crises,and the absence of prior risk controls were important factors in both of these instances.This paper examines these events and looks at the demand for share pledge financing,as well as corporate governance,corporate finance practices and secondary market characteristics.It uses the XGBoost algorithm in machine learning to predict and analyze pledge risks,focusing on prior prevention and control of the share pledge risk of majority shareholders.Empirical results show that the share pledge risk classification model trained by XGBoost has a very good predictive performance,and the highest recall in the test set reaches 74.44%.An analysis of feature importance shows that,compared with other majority shareholders,share pledge risk is highest among domestic non-state shareholders.In terms of governance structure,the higher the shareholding ratio of the majority shareholders,the higher the risk of share pledges.Conversely,an increase in the shareholding ratio of other shareholders inhibits this risk.Companies with a higher share pledge risk have lower cost expense ratios,higher debt ratios,and worse cash flow.Companies with higher share pledge risk have lower price-tosales ratios and liquidity levels.

关 键 词：股权质押 大股东 特征工程 机器学习 

分 类 号：F832.51[经济管理—金融学]