Exploring Risk Exposure Methodologies
- 19 March 2021
Structured methods for operational risk analysis
In early 2020, ORX launched a new project to explore the use of structured scenarios for operational risk measurement, specifically structured scenarios based on Bayesian networks. Working with Elseware, our aim was to discover how financial institutions are using structured scenarios. ORX members can access a full report exploring our findings, while non-members can download a free summary report.
Why use structured scenarios?
Much of operational risk quantification relies on two techniques – loss distribution approaches (LDAs) and scenario analysis. Each has its own merits and areas of application: LDAs are useful for familiar risk types for which sufficient data is available, while scenarios can be used to assess new or emerging risks, or for which data may be sparse.
A third approach, which in some way tries to build a bridge between these two, are structured scenarios. The assumption behind these approaches is that the underlying exposure an institution has to a specific risk can be described by a set of factors. These variables drive the frequency and severity of loss events, and by identifying and understanding them, you can both measure the exposure to a risk and know how best to control and mitigate it.
Structured scenarios and Bayesian networks
Structured scenarios using Bayesian networks are built on risk drivers corresponding to both firm-specific and external variables describing the environment. They are developed using subject matter input to build structures that capture loss generating processes that reflect the organisational risk profile to specific risk types and scenarios.
As they are partly data-driven, they can be seen as building a bridge between conventional scenarios and purely quantitative approaches to measure operational risk exposure. They, therefore, promise to provide a more nuanced view of operational risk profiles. Nevertheless, while structured approaches have been in use for a while and are increasing in popularity, they have yet to become a widely adopted approach for operational risk quantification.
Bringing the industry together to answer two key questions
This project explored two key questions:
- Can the industry work together to create a factor-driven approach to measure operational risk that more adequately reflects the organisational risk profile?
- What can ORX do to support members that are using, or plan to implement, such an approach?
This meant exploring whether we could, together with a group of industry representatives, create a set of basic models or structures, which individual firms can use and adapt for their own purposes. The intention was to focus on two scenarios to explore the feasibility and interest in this kind of collaboration. In light of the 2020 Covid-19 outbreak, the two scenarios that were chosen were pandemic and vendor failure.
The project was run in collaboration with Elseware, a specialist consultancy with experience in structured scenarios based on Bayesian networks, who provided input and guidance during the project. It was also overseen by a steering group of our member institutions, who were responsible for the general direction of the project and provided input and feedback on the developed structure.
Download the summary report for more
The free summary report gives you information on a range of key areas, including:
- Visualising structured scenarios
- Why use structured scenarios?
- Bayesian networks for operational risk management
The report also presents two example structures that were developed during the study – one on Pandemic and one on Vendor Failure.