对于非寿险准备金估计,传统流量三角形进展年的选择一直饱受争议,过长时易因数据不足难以估计,过短时可能会出现零赔付的情况。引入操作时间可以解决这一问题,但直接计算的操作时间实际上被严重高估。为此基于信度思想,使用个体信息将保单分组,建立两步广义线性混合模型,首先引入进展时间估计操作时间,再使用操作时间估计指定时间段内的未决赔款准备金。使用某非寿险公司的车损数据进行实证分析,并将该模型与使用传统流量三角的广义线性模型进行对比,结果均表明该模型效果甚佳。
It is very difficult to determine the length of development year in traditional run - off triangles for non - life insurance reserve estimation. It is either too long to get enough data for estimation or too short to ensure each development year have a claim. The introduction of operational time is a solution to this problem. Unfortunately, the operational time is highly overestimated when it is directly computed. Base on the credibility theory, we grouped policies by individual information, and created a two -step GLMM model. Firstly, we introduced the development time to estimate the operational time, then we used the operational time to estimate the IBNR for a specific period. Finally, we conducted an empirical analysis of vehicle damage insurance of a non - life insurer and made a comparison between this model and the GLM using traditional run - off triangles. Both analyses indicated that this model produced a better result.