Tweedie分布在车险费率厘定中的应用

发布时间：2018-05-26 08:46

本文选题：Tweedie分布 + 复合泊松-伽玛分布　；参考：《保险研究》2017年01期

【摘要】：在车险费率厘定中经常假设索赔频率与索赔强度分别服从泊松分布与伽玛分布,即假设总索赔服从复合泊松-伽玛分布。为了估计各风险类的纯保费(即总索赔均值),通常做法是对索赔频率与索赔强度分别建立广义线性模型(GLM),进而得到各风险类的索赔频率与索赔强度的均值,然后把两均值简单相乘即可;另一种做法利用复合泊松-伽玛分布是Tweedie分布的特例这一性质,直接对总索赔建立广义线性模型,进而也可以得到各风险类的总索赔均值。本文阐述了两种建模方法在处理车险费率厘定问题时的区别,通过对来自国外、国内的两组数据进行实证分析,比较了两种建模方法的优劣,并得到了一些初步结论。
[Abstract]:In the determination of vehicle insurance rate, it is often assumed that the claim frequency and claim strength are assumed to follow the Poisson distribution and the gamma distribution respectively, that is to say, the general claim suit is assumed to follow the compound Poisson gamma distribution. In order to estimate the net premium of each risk category (that is, the average value of the total claim), a generalized linear model is established for the claim frequency and the claim strength respectively, and then the average value of the claim frequency and the claim strength for each risk category is obtained. Then the two mean values can be simply multiplied; the other way is to establish a generalized linear model for the total claim by using the property that the compound Poisson gamma distribution is a special case of the Tweedie distribution and then to obtain the total claim mean of each risk category. This paper expounds the difference between the two modeling methods in dealing with the problem of vehicle insurance rate determination, compares the advantages and disadvantages of the two modeling methods through the empirical analysis of two groups of data from abroad and China, and obtains some preliminary conclusions.
【作者单位】：南开大学金融学院;
【基金】：国家自然科学基金(No.71271121)的资助
【分类号】：F224;F841.334

【相似文献】