In the wake of the financial crisis of 2008, regulatory pressure on financial institutions has increased significantly, particularly around the three components of risk: Credit Risk, Market Risk, and Operational Risk. In response, banks have continued to increase capabilities in quantifying each risk type. This brief overview will focus on financial modeling approaches used in Operational Risk.
Operational Risk is defined by the Basel II accord as “the risk of direct or indirect loss resulting from inadequate or failed internal processes, people and systems or from external events.”  Operational losses can occur for many reasons, and are often grouped into seven categories (in accordance with Basel II) which include Internal Fraud (EL1), External Fraud (EL2), Employment Practices and Workplace Safety (EL3), Clients, Products and Business Practices (EL4), Damage to Physical Assets (EL5), Business Disruption and System Failures (EL6), and Execution, Delivery and Process Management (EL7). Although each loss category is often considered separately, several methods exist for measuring and modeling the interactions between different loss types.
A standard technique for modeling operational losses is the Loss Distribution Approach (LDA) which is an accepted methodology under the Advanced Measurement Approach (AMA) proposed by Basel II. For LDA, frequency and severity are modeled by separate distributions with Monte Carlo techniques used to sample the distributions to perform operational Value at Risk (VaR) measurements. A very readable and highly informative introduction to the field of OpRisk (including the Loss Distribution Approach) is Marcelo Cruz’s “Modeling, measuring and hedging operational risk.”  This book provides information regarding both the qualitative and quantitative aspects of OpRisk measurement, and is written at a conversational level that is easily digestible and detailed enough for practical use.
OpRisk modeling is often based on multiple data types including internal loss data, external loss data, and structured scenario data obtained from subject matter experts. External loss data can be obtained from large databases, for example the ORX database which includes operational loss data contributed by numerous financial institutions across the globe.
LDA can be adapted to incorporate multiple data sources via a Bayesian approach. When following the Bayesian statistical approach, a prior estimate can be used to incorporate subjective information into the model in conjunction with the actual data. A rigorous and deep-diving reference with detailed information about LDA in a Bayesian framework is Pavel Shevchenko’s “Modelling Operational Risk Using Bayesian Inference.”  This book covers a number of additional useful modeling approaches in Operational Risk including credibility theory, Extreme Value Theory (EVT), and copula models for dependence between risk types.
A more recent collaboration of Cruz, Shevchenko and Gareth Peters is “Fundamental Aspects of Operational Risk and Insurance Analytics”, which is an up-to-date and exhaustive reference on all aspects of Operational Risk modeling.
The field of Operational Risk modeling continues to evolve, and it is important for banking institutions to be aware of the latest developments. Prescio Consulting brings advanced expertise in the field of Operational Risk to help your institution develop and validate models in ways that meet or exceed industry standards.
- Basel Committee on Banking Supervision, “Consultative Document Operational Risk”, 2001
- Basel Committee on Banking Supervision, “Operational Risk – Supervisory Guidelines for the Advanced Measurement Approaches”, 2011.
- Marcelo G. Cruz, “Modeling, measuring and hedging operational risk”, Wiley, 2002.
- Pavel V. Shevchenko, “Modelling Operational Risk Using Bayesian Inference”, Springer, 2011.
- Marcelo G. Cruz, Gareth W. Peters, Pavel V. Shevchenko, “Fundamental Aspects of Operational Risk and Insurance Analytics”, Wiley, 2015