Abstract: In recent years, with the frequent occurrence of major operational risk events in insurance companies, operational risk has become one of the important risks that cannot be ignored in the insurance industry. However, most of the researches ignores the correlation among risk events, which significantly affects the accuracy of operational risk measurement results. In this paper, we consider the correlation in the operational risk measurement model, and embed the correlation structures into the framework of the loss distribution approach, and study the measurement of operational risk on the basis of considering the frequency dependence, severity dependence, and loss dependence. Moreover, in view of the fact that the lack of operational risk data in the insurance industry seriously hinders the empirical research, we propose a data collection standard for operational risk in the Chinese insurance industry and collect 922 operational risk data of the Chinese insurance industry from 1995 to 2019 collected from public media. The empirical results based on this database show that: there are indeed complex correlations among operational risk events in the insurance industry, and ignoring these correlations may seriously underestimate the measurement value of operational risk; Considering different types of correlation will also lead to significant differences in measurement results. When measuring the operational risk of the insurance industry, we also need to consider the types of correlation among risk events carefully. The results of this paper can provide an important basis for the rational allocation of operational risk capital in the Chinese insurance industry.

Key words: operational risk, insurance industry, risk dependence, loss distribution approach, Copula

摘要： 近年来保险公司重大操作风险事件频发,操作风险已成为保险业最为重要的风险之一。而现有的保险业操作风险度量研究大多忽视了风险事件之间的关联关系,显著影响着操作风险度量结果的准确性。本文在保险业操作风险度量模型中引入相关性刻画方法,将相关结构嵌入损失分布法框架中,在考虑操作风险事件之间的频率相关、强度相关和损失相关三种类型相关性的基础上对操作风险进行度量研究。并且针对保险业操作风险数据缺失严重阻碍该领域实证研究的难点,提出了中国保险业操作风险数据收集的字段标准,从公开渠道收集了中国保险业1995年至2019年共922条操作风险数据。基于该数据的实证结果显示:保险业的操作风险事件之间确实存在着错综复杂的相关关系,忽略这些相关关系可能会严重低估操作风险的度量值;并且考虑不同类型的相关关系也会导致度量结果存在较大差异,在度量保险业操作风险时,需要慎重考虑风险事件之间的相关关系类型。本文的研究结果可以为我国保险业操作风险资本金的合理配备提供重要依据。

关键词: 操作风险, 保险公司, 风险相关性, 损失分布法, Copula