The Measurement Of Operational Risk Of The Property&Casualty Insurance Companies In China Based On PSD-LDA

With the promulgation of the "C-Ross" the regulation of China's insurance industry has entered the era of risk oriented.As one of the seven risks faced by the insurance industry,operational risk plays a more and more important role in the framework of risk management.However,the risk management of Chinese insurance industry started late,especially the P&C insurance companies,they have not yet established loss events database,the lack of loss data resulting in less quantitative research.Through the analysis of the operational risks that P&C companies facing and considering the results of the loss data collected in reality,the paper takes the fraudulent operational risk as study object,and quantifies the risk loss by LDA model.The fraudulent operational risk loss events of P&C insurance companies in the last 15 years was collected from the related websites.Studies of these data reveal that it has the characteristics of "high frequency and low loss" and "low frequency and high loss".So,the PSD-LDA model could be used to quantify the loss.Previous researches generally treated the loss data collected as the real risk events,so they usually used Poisson distribution to fit the loss frequency.However,they are not the same when considering the operational risk events that were unrecognized,identified but unexposed and exposed but missing.In consideration of the existence of these bias,the paper proposed the compound Poisson-Geometric distribution to fit the loss frequency.Empirical study reveals that the deviation between the gathering risk events and the real risk events do exist,and the deviation of the "high frequency and low loss" event is higher than that of the "low frequency and high loss" event.The PSD-LDA model based on the compound Poisson-Geometric distribution is used to measure the risk loss and the corresponding economic capital.And the paper also discussed the application of the quantitative model in operational risk management.The results show that the PSD-LDA model of compound PG distribution can make a better fit of fraudulent operational risk compared with Poisson distribution.This model not only provides a new idea about the measurement of operational risk for insurance company,but also can be used in the management of operational risk.