Bank Operational Risk Measurement Based On The Akaike Weighted Integration Model

With the globalization of the world economy,the financial products and the business of banking institutions becoming more and more complex.A series of operational risk has caused great losses to banks.As a result,banks and regulators have paid more and more attention to the measurement and management of operational risks in many countries.This paper focus on a core problem that the lack of loss data in the measurement of operational risk.For banks,it need to measure the operational risk in the case of missing data.For regulator,it have to know more about the imformation of the operational risk to reduce the impact of information asymmetry under the condition of laking of relevent information.It is a good way to combine the internal data and the external data to solve the problem,but it can be improved in measurement models and methods.For this reason,the paper studies the problem that how to combine the internal data and external data in operational risk measurement,and try to discuss the reasons,principles,and methods for integrating internal and external data when measuring operational risk were discussed.At last,the paper tries to establish a bank operational risk Akaike Weighted integrated model based on the Loss Distribution Method,Peak Over Threshold,and Akaike Weight Method and collects data on operational risk losses of the Chinese Commercial Banks for empirical analysis.The details are as follows:(1)Used the loss distribution meathod to measure the operational risk based on the loss data.It is found that Negative Binomial distribution is better than the Poisson distribution to fit the frequency of loss event and Lognormal distribution is the best distribution to fit the loss severity,followed by Burr distribution,Pareto distribution and so on,but it is not a good fit for the tail loss even used the Lognormal distribution.In addition,relative to the frequency distribution,the distribution of loss severity has a greater influence.(2)In order to better fit the tail loss of operational risk,the paper used the piecewise fitting method to fit the "high frequency and low loss" and "low frequency and high loss" by Loss Distribution method and Peak Over Threshold model.It is found that the Generalized Pareto distribution is better than the Lognormal distribution to fit the tail loss.(3)According to on the previous research,the paper use the Bühlmann-Straub model to combine the internal and external data.The result has show that it is very unstable and have the arge deviation because this meathod ignores the differences of internal and external data and is affected by the sample size easily.(4)In order to combine the internal and external data more reasonable,the paper build the operational risk model based on the POT and the Akaike Weight method.The result is very stable,it show that the model can describe the internal data and external data heterogeneity very well.