OPERATIONAL RISK IN BANKS: FPD-PSD-OLDA MODEL AND OpVaR-CaR MEASURE

• Main Author: DWI PUTERA (NIM:10114040), GANDHIANO
• Format: Final Project
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/27449
• Institution: Institut Teknologi Bandung

Operational Risk in Banks is defined as the risk of bank’s loss resulting from inadequacy or failed internal processes, human error, technology information system, and/or the presence of external events environment. Operational Risk in Banks can be quantified using the AMA (Advanced Measurement Approach) method which allows banks to develop their own internal models, such as the OLDA (Operational Loss Distribution Approach) model. The model involves the concept of aggregate by having two important components such as frequency and severity components. The OLDA model will be developed into a model that includes the stochastic process effects on the frequency component and the range of HFLS-LFHS (High Frequency Low Severity – Low Frequency High Severity) risk types in the severity component. It is called as FPD-PSD-OLDA (Frequency Process Distribution and Piecewise-defined Severity Distribution Based on Operational Loss Distribution Approach) model. In this paper, OLDA and FPD-PSD-OLDA models are constructed based on four methods including convolution method, characteristic function, Panjer recursion, and FFT (Fast Fourier Transform). Based on the illustration, it is seen that Panjer recursion and FFT are often used is not applicable to the overall distribution of severity. In contrast, convolution methods regarded as classical methods are found to be flexibly capable of being developed for the overall distribution. OLDA and FPD-PSD-OLDA models are applied to one pair of business lines and types of risk events in the bank. The concept of model development for banks is done by summing the FPD-PSD-OLDA model for two pairs of business lines and the type of risk occurrence that is dependent on each other, specifically named as the Bivariate FPD-PSD-OLDA aggregate model. The model is constructed by showing the various sizes of the two pairs of dependencies that involve Copula representation. OLDA and FPD-PSD-OLDA models are also used to predict the measure of the Operational Risk in Banks for one pair of the business lines and the type of risk event called the OpVaR-CaR (Operational Value-at-Risk-Capital-at-Risk). OpVaR-CaR measure compared to various construction methods (Panjer recursion and FFT) with explicit approximation method (Bocker and Sprittula Approximation), shows that the OpVaR-CaR measure tends to insignificantly changed and can be controlled effectively. Basel II states that the maximum limit of the predicted measure of the Operational Risk in Banks is the sum/aggregate of each of the OpVaR-CaR components of the business line and the type of risk event pairs. Those limit value is used as a prediction reference of OpVaR-CaR Bivariate FPD-PSD-OLDA aggregate model to show how optimal the capital reserve that need to be prepared by the banks. To quantify the exceeding value of the maximum limit, then an OpVaR-CaR Diversification measure is defined. Based on the research, it also shown that diversification is not influenced by correlation factor and tail dependencies.