基于分位数回归的交强险费率厘定研究

An Analysis of Ratemaking of Auto Compulsory Liability Insurance Using Quantile Regression

交强险具有多种保险责任的特点为其费率厘定带来新问题。本文通过索赔分类的思想对我国某地区交强险的保单数据进行分析,探讨用于分类费率厘定的因子是否能够真实反映被保险人在人身伤害或财产损失类索赔中呈现的风险特征。运用零调整的分位数回归,我们发现现有费率因子对于这两类索赔的影响有显著区别,并且交强险目前的费率厘定方式无法真实反映被保险人的风险水平。本文提出费率因子对于不同损失程度的索赔应具有一致性效应,并据此来评价费率厘定的合理性。本文建议将两类索赔分开厘定费率或加入新费率因子等方案可以改进目前的费率厘定方法。

The multiple coverages of auto compulsory liability insurance induces new problems in ratemaking. This study made an empirical analysis of China auto compulsory liability insurance claim data under the methodology of claim classification, to discuss whether rating factors used for risk classification could reflect the actual risk characteristics of insureds on body injury and property damage respectively. Using zero-adjusted quantile regression, we found that existing rating factors had significantly distinct influences on two types of claims. And current ratemaking couldn＇t reflect the actual risk level of insureds properly. The study proposed the principle that rating factors should have consistent effects on claims study suggested to separate two types of regardless of their level of losses to verify the rationality of ratemaking. The claims for ratemaking or add new rating factors to optimize the current ratemaking method.