The Empirical Research On Loss Distribution Approach Of Banks' Operational Risk

For the banking area, the operational risk measurement is no longer a new field. Many domestic and foreign scientific scientists and practitioners have much research on this topic, mainly focused research on measurement tools. BaselⅡAccord Proposed that the operational risk measurement methods include Basic Indicator Approach, Standardized Approach, Advanced Measurement Approach. In the Advanced Measurement Approach, the bank can use its internal model to measure the Bank's own operational risk regulatory capital in special interval(one year)and level of 99.9% Confidence. Supervisior suggested that internationally active banks can build own econometric model through creating databases, theoretical models of research and statistical validation of. In this paper, we mainly research the loss distribution approach. In this paper, on the Advanced Measurement Approach in the loss distribution approach has been studied, as compared with loss distribution approach is higher than the other advanced metrology in complexity of the model, the accuracy of measurement results and the risk sensitivity.As an actuarial model, Loss distribution approach mainly includes the loss frequency distribution and the loss severity distribution. During the empirical study on the Loss distribution approach, Data mainly based on the internal fraud losses of China's four state-owned commercial banks in the 1999-2009. The article choosed the methods of a single sample K-S test and P-P plot to determine the optimal theoretical distribution of the loss frequency and loss severity, and based on the maximum likelihood to estimate the parameters. In calculating the total loss distribution function, firstly, the article used Monte Carlo simulation method to simulate the loss frequency and loss severity data; secondly, used the convolution method to calculate the total loss; finally, based on the total loss, calculate the capital that can cover the risk of internal fraud. In summing the 56 types of operational risk losses, introducing the Copula functions can eliminate correlation between the type of loss. Finally the conclusion on the operational risk loss distribution can meet the bank's actual loss distribution better.