面板数据下的线性混合模型及其在车险费率厘定中的应用
Linear Mixed Models for Panel Data and Their Applications in Auto Insurance Rate-making

精算师在进行车险净保费信度厘定时可采用关于面板数据的线性混合模型,本文采用每次交通事故平均损失额和事故发生频率作为车险净保费的计算指标。利用2008~2012年31个省、市、自治区5年的数据，建立面板数据下的线性混合模型，选取人均地区生产总值、每平方公里人口数、民用汽车拥有量作为解释变量，得到每次交通事故平均损失额和事故发生频率的估计模型，进而得到纯保费估计。这一研究可为车险费率市场化提供一定的理论支持和参考。
Linear mixed models for panel data and their applications in auto insurance pricing are discussed. Actuaries can use linear mixed models on panel data for credibility rate-making purposes. Using panel data (2008-2012) from 31 provinces, municipalities or autonomous regions of China for an empirical analysis, the present paper studies the long-term equilibrium between the average losses of traffic accident of each time, frequencies of accidents and per capita GDP, population density, the quantity of civilian vehicles by using the panel data model. The average losses of traffic accident of each time and the frequencies of accidents are selected as measures of net premium. The conclusions of this paper provide with theoretical support and reference for vehicle insurance rate-making marketization.