报告时间：2021年11月17日 19：00-20：00

报告地点：腾讯会议：501 277 112

报告人：杨亮（西南财经大学）

报告摘要：In this paper, we extend the idea of embedding the top-down method into quantile regressions to derive risk loadings in classification ratemaking. By restricting that the portfolio’s total risk premium should equal the sum of the risk premiums of each policy, we first implement the bootstrap method based on generalized linear models to calculate the total risk premium of the portfolio at the collective level, and then allocate it to the individual policy to estimate the quantile level in qunatile regressions. Three modeling frameworks are considered based around traditional quantile regression model, fully parametric quantile regression model, and quantile regression model with coefficient functions, which we develop specifically in this classification ratemaking setting. This approach also allows estimating risk loading parameters in various premium principles, e.g., expected value premium principle, standard deviation premium principle, Wang premium principle, and quantile premium principle proposed in Heras et al. (2018) and Baione and Biancalana (2019), to determinate risk premiums at the individual level. The empirical result shows that the risk premiums calculated by the method proposed can reasonably differentiate the heterogeneity of different risk classes.

报告人介绍：杨亮，西南财经大学精算系讲师，毕业于中国人民大学统计学院。现任研究领域包括：非寿险 费率厘定、准备金评估、巨灾保险、风险管理与评估、机器学习方法在车辆网大数据上的应用等。在《统 计研究》（2篇）、《数量经济技术经济研究》、《中国软科学》、《系统工程理论与实践》、《系统工程 学报》、《经济学家》、《Insurance: Mathematics and Economics》等国内外核心期刊发表论文十多篇。

邀请人：马学俊