《保险研究》20220107-《贝叶斯网络挖掘健康对中老年人购买商业医疗保险的影响研究》（马丽娜）

［摘   要］本文利用2018年中国健康与养老追踪调查(CHARLS)数据，基于健康视角研究了我国中老年人购买商业医疗保险的影响因素和传导机制链。在量化自评健康、身体健康、心理健康、社会健康行为以及个人特征的总计23个指标中，基于AIC准则利用逐步回归技术仅筛选出11个对购买商业医疗保险存在显著影响的指标，借助偏相关系数评价了其重要性。进而，利用爬山算法、自助重采样技术和模型平均技术构建贝叶斯网络，挖掘出影响中老年人购买商业医疗保险决策行为的5个特征模式，涉及5个因素:年龄、社交类型数、教育水平、居住地类型和是否吸烟。研究发现，在影响购买商业医疗保险的决策行为方面，年龄、教育水平和社交类型数发挥中介效应，且对若干健康指标具有替代效应。相比于身体健康和心理健康，社会健康行为对于中老年人购买商业医疗保险产生更强的影响。保险公司应积极创新差异化的商业医疗保险产品。虽然基于本文的数据并未发现我国中老年人购买商业医疗保险存在逆向选择风险，但仍需防范产品创新诱发的逆向选择风险。政府和保险公司应加大对商业医疗保险的宣传力度，提升中老年人群对商业医疗保险的认知水平，提高他们通过商业医疗保险规避风险的意识。

［关键词］商业医疗保险；贝叶斯网络；Logit回归；健康；特征模式

［基金项目］国家自然科学基金青年科学基金项目“基于Markov模型的反向抵押贷款及其连结长期护理保险产品的定价研究”(项目编号:71401124)。

［作者简介］马丽娜，首都经济贸易大学金融学院副教授、硕士研究生导师，主要研究方向:保险精算、数据挖掘和复杂网络。

Bayesian Network Mining the Impact of Health on the Purchase of Commercial Medical Insurance for Middle-to-old-aged Persons

MA Li-na

Abstract：Based on the data of China Health and Retirement Longitudinal Study (CHARLS) in 2018,this paper studied the influencing factors and transmission mechanism chains of purchasing commercial medical insurance for middle-to-old-aged persons in China from the perspective of health.The results showed that among the 23 indexes of self-rated health,physical health,mental health,social health behavior and personal characteristics,only 11 indexes which had significant impacts on the purchase of commercial medical insurance were selected based on AIC criteria by using the stepwise regression,and their importance was evaluated by means of the partial correlation coefficient.Then,Bayesian networks were built with the hill-climbing algorithm,bootstrap resampling and model averaging technology.We dug out five characteristic patterns that affected the decision-making behavior of middle-to-old-aged persons to purchase commercial medical insurance,involving five factors:age,number of social types,education level,residence and smoking or not.While influencing the decision-making behavior of purchasing commercial medical insurance,age,education level and number of social types played an intermediary effect,and had a substitution effect on multiple health indicators.Compared with physical health and mental health,social behavior health had a stronger impact on the purchase of commercial medical insurance for the middle-to-old-aged persons.The insurers should actively innovate the differentiated commercial medical insurance products.Although based on our data,adverse selection risks aren′t found for middle-to-old-aged persons in purchasing commercial medical insurance,it is still necessary to guard against the adverse selection risks induced by product innovation.The government and insurer should strengthen the publicity of commercial medical insurance,improve the cognitive level of middle-to-old-aged persons on commercial medical insurance,and raise their awareness level of avoiding risks through commercial medical insurance.

Key words：commercial medical insurance;Bayesian network;Logit regression;Health;characteristic pattern