主管:中国科学院
主办:中国优选法统筹法与经济数学研究会
   中国科学院科技战略咨询研究院

中国管理科学 ›› 2025, Vol. 33 ›› Issue (8): 75-89.doi: 10.16381/j.cnki.issn1003-207x.2022.2294

• • 上一篇    下一篇

竞争网络与公司财务绩效

朱建新(), 刘可心, 曾能民, 吴雄   

  1. 哈尔滨工程大学经济管理学院,大数据与商务智能技术工业和信息化部重点实验室 (哈尔滨工程大学),黑龙江 哈尔滨 150001
  • 收稿日期:2022-10-25 修回日期:2024-01-10 出版日期:2025-08-25 发布日期:2025-09-10
  • 通讯作者: 朱建新 E-mail:zhjx@vip.163.com
  • 基金资助:
    国家社会科学基金重点项目(20AGL009)

Competitive Network and Financial Performance: Empirical Evidence Based on Explainable Random Forest

Jianxin Zhu(), Kexin Liu, Nengmin Zeng, Xiong Wu   

  1. School of Economics and Management,Harbin Engineering University,Key Laboratory of Big Data and Business Intelligence Technology (Harbin Engineering University),Ministry of Industry and Information Technology,Harbin 150001,China
  • Received:2022-10-25 Revised:2024-01-10 Online:2025-08-25 Published:2025-09-10
  • Contact: Jianxin Zhu E-mail:zhjx@vip.163.com

摘要:

竞争网络位置作为企业制定战略决策的关键因素,其对财务绩效影响的重要性评估是本文研究的重点。本文以我国2008-2019年的财产保险行业为研究样本构建竞争网络,采用基于可解释性随机森林的机器学习模型衡量竞争网络特征对企业财务绩效的重要性,并进一步分析竞争网络特征对拟合企业财务绩效模型的相对重要性排序以及因果关系。研究发现:(1)整体而言,竞争网络或者市场结构特征的缺失都会降低对企业财务绩效模型的拟合性能,其中竞争网络特征缺失带来的影响更为明显。(2)按照重要程度对竞争网络中的特征进行排序,并进一步采用SHAP解释模型探索竞争网络特征影响财务绩效的内在机理,结果表明,个体网规模、中间人次数和接近中心性这三个最为重要的特征对财务绩效均服从幂律分布。(3)进一步地,运用倾向得分匹配方法验证了变量间的强因果关系,证明该研究范式对传统计量模型的开发也具有重要的指导作用。本文的结论丰富了竞争网络与财务绩效相关领域的研究成果并对其提供依据,且为面向高维、非线性因素间的因果分析与验证提供了新的解决思路。

关键词: 竞争网络, 财务绩效, 随机森林, SHAP模型

Abstract:

The significance of the competitive network position as a key factor in firms' strategic decision-making and its impact on financial performance is the focus of this study. First, using China's property insurance industry from 2008 to 2019 as the research subject, a Random Forest model is constructed to investigate the predictability of competitive network characteristics and market structure characteristics. Next, the importance of competitive network and market structure is analyzed using the feature importance evaluation method of the Random Forest model. Finally, the nonlinear functional relationship between these important features and financial performance is revealed using SHAP values and further the causal relationship between key competitive network features and financial performance is verified using traditional econometric models (Propensity Score Matching). It also explores the differences between nonlinear and linear model results.The research findings show that (1) Overall, the absence of either competitive network or market structure features can reduce the fitting performance of the enterprise financial performance model, with the impact of missing competitive network features being more pronounced. (2) In terms of the ranking of feature importance, competitive network features account for approximately 30% of the importance to enterprise financial performance, significantly higher than the importance of market structure features. Key features within competitive network include: individual network size, intermediary frequency, and closeness centrality. (3) These three key features follow a power-law distribution with respect to enterprise financial performance, and the existence of a strong causal relationship between important competitive network features and financial performance was verified through propensity score matching. The conclusions of this paper enrich the research outcomes in the field of competitive networks and financial performance, providing support for them, and offer a new solution for causal analysis and verification between high-dimensional, nonlinear factors.

Key words: competitive network, financial performance, random forest, SHAP model

中图分类号: