车辆保险索赔次数中类别风险因子的差异性分析——基于零膨胀泊松模型的Blinder-Oaxaca分解
Differential Analysis of Category Risk Factors in Vehicle Insurance Claim Frequencies——Blinder-Oaxaca Decomposition Based on the Zero-Inflated Poisson Model
郭念国 1周鹏2
作者信息
- 1. 河南工业大学理学院,河南 郑州 450001
- 2. 中央财经大学财经研究院,北京 100098
- 折叠
摘要
对车辆保险中影响索赔次数的类别风险因子差异性分析给出了一种统计方法.因为车辆保险费率厘定中,风险因子间往往具有一定的相关性,若继续使用广义线性模型拟合索赔次数分布,依据模型的估计参数解释类别风险因子对索赔次数的影响时就会产生一定的偏差.为解决此问题,依据Blinder-Oaxaca分解思想,基于类别风险因子对零膨胀泊松模型进行分解,并对一组车辆保险索赔次数数据进行建模分析,利用Bootstrap方法模拟类别风险因子差异并给出置信区间.与传统的零膨胀泊松模型拟合结果进行比较,结果表明,使用零膨胀泊松模型的Blinder-Oaxaca分解,更能反映出数据本身的信息差异,体现分类风险因子间的风险水平,从而为分类费率厘定技术提供一定的理论依据.
Abstract
In this paper,a statistical method is given for the analysis of the difference of category risk factors affecting the number of claims in vehicle insurance.Because there is often a certain correlation between risk factors in the vehicle rating,if the generalized linear model continues to be used to fit the distribution of the claim frequencies,a certain deviation will occur when explaining the influence of category risk factors on the claim frequencies according to the estimated parameters of the model.In order to solve this problem,this paper models the zero-inflated Poisson model based on the category risk factor under the Blinder-Oaxaca decomposition idea,models and analyzes a set of vehicle insurance claim data,and uses the Bootstrap method to simulate the difference of category risk factors and gives confidence intervals.Compared with the fitting results of the zero-inflated Poisson model,the results show that the Blinder-Oaxaca decomposition using the zero-inflated Poisson model can better reflect the information difference of the data itself and reflect the risk level between the classification risk factors,so as to provide a certain theoretical basis for the classification rate determination technology.
关键词
索赔次数/类别风险因子/零膨胀泊松模型/Blinder-Oaxaca分解Key words
claim frequencies/category risk factors/zero-inflated poisson model/Blinder-Oaxaca decomposition引用本文复制引用
基金项目
河南省联合基金(222103810059)
中国商业统计学会一般规划课题(2023STY56)
出版年
2024
NETL
NSTL
万方数据