《保险研究》20240402-《中国财产保险公司财务预警研究——基于随机森林财务诊断法》（段伊雪、彭雪梅、方匡南）

［摘   要］建立健全保险业财务预警机制可以为保险业审慎性监管提供工具，有助于维护金融安全。由于“偿二代”对保险公司财务风险的监测滞后，本文提出了随机森林财务诊断法，尝试提前识别财产保险公司潜在的财务风险并预警，并基于2016~2021年81家财产保险公司的数据进行检验。研究发现，随机森林财务诊断法的预警准确性显著高于其他模型，创新指标具有预警价值，多个指标对财务危机的贡献度呈现明显的“阈值效应”，且净资产收益率、每百元保费经营活动净现金流量和净资产变动率三者具有识别财务危机的“联合效应”。本文研究结果有助于监管机构在众多保险机构中快速筛选出问题公司，并能够助力财产保险业界实现对财务危机公司的“诊断+治疗”功能。

［关键词］财产保险公司；财务预警；随机森林模型；财务诊断法

［作者简介］段伊雪，西南财经大学金融学院博士研究生；彭雪梅，西南财经大学金融学院教授、博士生导师；方匡南，厦门大学经济学院教授、博士生导师。

A Study of Early Financial Warning for China Property & Casualty Insurance Company—Based on the Random Forest Financial Diagnosis Methodology

DUAN Yi-xue,PENG Xue-mei,FANG Kuang-nan

Abstract：The establishment and improvement of the insurance industry early financial warning mechanism can provide an instrument for prudential supervision of China’s insurance industry,which is conducive to financial security maintenance.Due to the delayed effect of the "C-ROSS" in monitoring potential financial risks arising in P&C insurance companies,the paper suggests to apply the random forest financial diagnosis methodology to study the early financial warning of the P&C insurance industry,and use the data of 81 P&C insurance companies from 2016 to 2021 to test its effect.The study shows that the methodology constructed in this paper can not only identify potential financial risks in advance but also be more accurate than other models,and the innovative indicators have early warning value.Multiple indicators presents an obvious "threshold effect",and return on equity,the net cash flow of operating activities per 100 yuan premium and change rate of net equity show a "joint effect" to identify financial distress.These results can help regulators promptly screen out problematic firms among a large number of insurance institutions and are of practical value in assisting the P&C insurance industry to realize "diagnosis + treatment" for financially distressed companies.

Key words：property & casualty insurance company;early financial warning;Random Forest model;financial diagnosis methodology