Operational Risk Management

Operational risk measurement and management is one of the emerging disciplines within banking risk management. Regulatory capital requirement for operational risk was introduced through Basel II guidelines. Advanced Measurement Approach (AMA) is one of the three approaches for regulatory capital computation for operational risk. The approach poses significant challenges in terms of operational loss data collection, scenario generation, development of a modelling framework, modelling of count and severity of losses, modelling of risk mitigation and risk aggregation.

Asymmetrix OpAMA Capital Engine provides integrated modules for loss data collection, scenario analysis, convolution and risk aggregation. AMA capital is computed based on Loss Distribution Approach and Scenario Based Approach. For LDA, multiple discrete and continuous distributions can be fitted to count and amount of losses respectively to identify the distribution that best describes underlying loss data. For SBA, scenarios can be generated using multiple methods. Convolution of frequency and severity distributions is accomplished through a Monte Carlo Simulation engine.

Asymmetrix OpAMA Capital Engine

Loss Data Collection (LDC)

LDC module provides a workflow-driven web-based platform for collection of loss data. Detailed data is collected about operational risk events, loss, recoveries, costs, controls, linkages with other events etc.

Scenario Analysis

Scenarios can be built using Percentile Method, Interval Approach as well as Individual Scenario Approach.

Parameter Estimation and Uncertainty

Discrete and continuous distributions are fitted to frequency and severity data respectively using multiple methods such as MLE, Moment Matching and PWM. Parameter Uncertainty is measured through parametric and non-parametric simulation to create confidence interval around the parameter values estimated.

Goodness of Fit

GOF tests such as Anderson Darling, KS, Chi-square, CVM alongwith graphic tests such as PP-plot, QQ-plot are available to identify best fit to underlying data. Information criteria such as AIC and BIC are available to identify best fit after penalising for model complexity.

Convolution and Aggregation Modelling

OpVaR computations using multiple methods such as Monte Carlo Simulation, Closed-form solution and Fast Fourier Transform, with and without incorporating insurance mitigation is supported. To understand reasonableness of capital estimates, empirical simulations are also performed. Aggregation Modelling is performed using correlation matrix and user selected copula that decides tail dependence (t copula and gaussian copula).

EVT Applicability

Applicability of EVT to loss data and threshold selection is done through multiple tests such as Hill Plot and Mean Excess Plot.

Products in this category

Related Products