摘要
我国财险企业的操作风险管理起步较晚,目前尚未建立损失事件库,损失数据的缺失导致量化分析的研究较少。通过搜集整理并分析近15年来财险企业的欺诈类操作风险损失事件,发现数据具有"高频低损"和"低频高损"的特征,因此采用两阶段损失分布法(PSD-LDA)进行拟合。考虑到操作风险事件的属性以及数据搜集的不完整性,运用复合Poisson-Geometric分布来拟合损失频率从而度量欺诈类操作风险损失,并计提了相应的经济资本来抵御非预期的操作风险损失。结果表明,相对于设定损失频率为齐次泊松分布,基于复合PG分布的PSD-LDA模型能更好的拟合财险企业欺诈类操作风险损失,为我国财险企业操作风险的度量和管理提供了新思路。
The operational risk management of P&C insurance companies had a late start in China,and the loss event database has not been established currently. The lack of loss data leads to scanty quantitative research. The fraudulent operational risk loss events of Chinese property insurance companies in the last 15 years was collected. Studies of these data revealed that it had the characteristics of "high frequency and low loss" and "low frequency and high loss". Therefore,a loss distribution approach based on piece wise-defined severity distribution (PSD- LDA) was used to calculate the fraudulent operational risk losses and economic capital to protect against unexpected operational risk-related losses in property insurance companies. Considering the attributes of operational risk events and the incompleteness of data collection, composite Poisson-Geometric distribution was proposed to fit the loss frequency. The results showed that the PSD-LDA model based on compound PG distribution could make a better fit for the loss of the fraudulent operational risks when compared with the Poisson distribution, which provides a new path for the measurement of the operational risk of Chinese insurance companies.
作者
陈迪红
刘冬梅
CHEN Dihong LIU Dongmei(Hunan University of Finance and Statistics,Hunan Changsha 410079)
出处
《保险研究》
CSSCI
北大核心
2017年第2期84-94,共11页
Insurance Studies
基金
中国人保财险灾害基金资助