Operational Risk Measuring Model For Commercial Banks In China And Its Empirical Research

Operational risk is one of the three risks of banking defined by Basel Committee, which is a kind of risk existing since banks established. While compared with credit risk and market risk, for a long time, people did not pay much attention to it. With development of financial industry, huge loss caused by operational risk occurred more and more frequently and diversifiddly. How to strengthen the measurement and supervision of the operational risk has become an important premise. However, in our country, the measurement and supervision of the operational risk is still lack of necessary policies and regulations, and the corresponding database of operational risk has not established. Moreover, the feedback mechanism of operational risk is incomplete. Many factors caused the data of commercial bank losses in our country are incomplete and not objective, which caused great obstacles to the accuracy of measurement. Above all, how to measure operational risk accurately is of great significance to the commercial bank in our country. In this paper, I selected two methods:one is Income Model, which is a typical method of "top-down"; the other one is Loss Distribution Approach, which is a typical method of "bottom-up". Based on the two methods, I made some improvement and proposed how to combine the two methods on measuring operational risk, and then gave some policy advice.This paper can be divide into six parts below.PartⅠ, introduction. In this part, I discussed the research background and significance, and then summarized the current situation on operational risk researching, finally, analyzed the main problem the domestic and foreign scholars facing.PartⅡ, the characteristics of the operational risk measurement and methods choosing. In this part, I gave operational risk different definitions in different measurement methods, and analyzed the difference between operational risk and credit risk and market risk. I introduced two kinds of typical classification, and then reviewed some classical measurement methods, including methods the Basel Committee advised and methods mainly used in academic circles. Finally, in consideration of the current situation on operational risk measurement, I picked out models suit to our country's commercial banks.PartⅢ, the distribution of operational risk events in our country. The data comes from the media and some related research achievements, which covered from 1994 to 2010. I characterized the distribution both from the point of view of time and classification, and then analyzed the underlying causes.PartⅣ, the measurement of operational risk using loss distribution model. In this part, I summarized the fundamental of loss distribution model, setting and fitting inspection, and then measured the operational risk using loss distribution model. In consideration of the problem of data cut-off, I revised the traditional model, and finally determined the value at risk under different confidence levels.PartⅤ, the measurement of operational risk using income model. In this part, I discussed the flaws of estimation method of traditional income model. In order to overcome the flaws, I introduced the panel data method to characterize the heterogeneous effect between the cross-section members. First, I summarized the related theory on income model and panel data method, and then estimated the varying-coefficient fixed effect model using cross-section Seemingly Unrelated Regression method. Finally, I completed the value at risk under different confidence levels.PartⅥ, conclusion and prospect. By comparing loss distribution model and income model, I proposed the principle on how to combine the two methods together, that is using loss distribution method approach with "top-down" in macro level and "bottom-up" in micro level; using income model approach with "bottom-up" in macro level and "top-down" in micro level. Finally, based on the measuring result, I proposed some policy advice on operational risk management. I also discussed the flaws of this article and problems remained to be further study.The innovation points of this paper reflect in the following aspectFirst, in the course of using loss distribution model, I summarized distribution forms more comprehensive, in order to fit the skewness and thick tail characteristics of loss distribution. These distributions included gamma and Rayleigh distribution, which are widely used in engineering, electric power, and other disciplines. This enlarged the range of choice of distribution and made the fitting more accurate. Besides, in the process of monte carlo simulation, based on the research achievement, I put forward a new method to adjust the result, in order to avoid the result's drastic fluctuations caused by amending the function's parameters, because the loss distribution method is very sensitive to the parameters.Second, on the part of income model, traditional models either make regression using each bank's time series data or mix many banks' time series data together, both of them did not take into account the correlation between different banks. This paper introduced the panel data method to characterize the heterogeneous effect between the cross-section members and estimated the varying-coefficient fixed effect model using cross-section Seemingly Unrelated Regression method, which made the parameter estimation more effective.Third, this paper proposed how to combine the two approaches to make the measurement more exactly.