公司和高管特征与上市公司违规行为——基于机器学习的经验证据
DOI:
作者:
作者单位:

作者简介:

通讯作者:

中图分类号:

基金项目:


Corporate and managerial individual characteristics and listed company violation: Evidence from a machine learning approach
Author:
Affiliation:

Fund Project:

  • 摘要
  • |
  • 图/表
  • |
  • 访问统计
  • |
  • 参考文献
  • |
  • 相似文献
  • |
  • 引证文献
  • |
  • 资源附件
  • |
  • 文章评论
    摘要:

    上市公司违规问题一直备受资本市场高度关注,对于其影响因素的单一维度变量因果关系研究固然关键,但构建有效的整体性预测模型研究同样具有重要意义.本研究基于内部治理视角选取重要的公司特征和高管个人特征构建上市公司违规行为预测模型,以2008年—2019年中国A股上市公司为样本,利用机器学习算法LightGBM和SHAP工具,研究两类特征对违规行为的预测能力,重要性排序及预测模式.研究结果表明:模型可以在一定程度上预测公司违规行为,公司特征比高管个人特征对预测产生的影响更大.其中,上市公司信息透明度越高、总资产净利率越大、资产负债率越低、高管团队持股比例越高、业绩波动性越小、分析师关注度越高,模型预测违规的倾向越低;高管年龄偏小、公司存在董事长与CEO两职合一情况时,模型预测违规的倾向增高.大部分特征均与违规行为呈现非线性关系,与传统理论和实证研究结论相一致.本研究从预测视角拓展我国公司高管特征研究,为监管部门和投资者提升监管和投资效率、企业完善内部治理机制提供经验证据.

    Abstract:

    Listed companies’violations have been an important issue that attracts the attention of the capital market. While studying the causal relationship between single-dimensional variables and this issue is crucial,constructing an effective holistic prediction model is also of great significance. This paper constructs a prediction model of listed companies’violations based on important company characteristics and managerial individual characteristics from the perspective of internal governance. Using a sample of Chinese A-share listed companies from 2008 to 2019,this study introduces two machine learning algorithms,LightGBM and SHAP,to examine the predictive ability,importance ranking,and prediction mode of the two types of characteristics on violation behaviors. The results show that the model can predict corporate violations to a certain extent,and corporate characteristics have a greater impact on the prediction than managerial individual characteristics. Specifically,higher information transparency of listed companies,higher net profit margin of total assets,lower asset-liability ratio,higher managerial shareholding ratio,lower performance volatility,and higher analyst attention are associated with a lower tendency for the model to predict violations. In addition,the model predicts an increased tendency for violations when executives are young and when the chairman and CEO roles are combined. Moreover,most corporate characteristics and managerial individual characteristics exhibit a non-linear relationship in predicting corporate violations,which is consistent with the findings of traditional theoretical and empirical studies. Overall,our study enriches the research on the characteristics of corporate executives in China from a predictive perspective and provides empirical evidence for regulatory authorities and investors to improve supervision and investment efficiency and for companies to optimize internal governance mechanisms.

    参考文献
    相似文献
    引证文献
引用本文

何瑛,任立祺,于文蕾,杜亚光.公司和高管特征与上市公司违规行为——基于机器学习的经验证据[J].管理科学学报,2024,(6):43~68

复制
分享
文章指标
  • 点击次数:
  • 下载次数:
  • HTML阅读次数:
  • 引用次数:
历史
  • 收稿日期:
  • 最后修改日期:
  • 录用日期:
  • 在线发布日期: 2024-07-28
  • 出版日期:
您是第位访问者
管理科学学报 ® 2024 版权所有
通讯地址:天津市南开区卫津路92号天津大学第25教学楼A座908室 邮编:300072
联系电话/传真:022-27403197 电子信箱:jmsc@tju.edu.cn