Lessons in stress testing
“Stress testing is like Max Verstappen in Formula 1: We are learning, but at the same time, we drive fast.” Our stress testing event on September 28th offered some remarkable, valuable insights.
The event started with a short introduction to stress testing by Jaap Karelse, partner at Zanders. He highlighted the importance of stress testing by emphasizing that stress test outcomes can restore confidence in the market and display control by banks.
Furthermore, banks can benefit from sophisticated internal stress testing as input for strategic decision-making and recovery planning. While there are clear benefits associated with stress testing, there are some difficult hurdles to overcome. For example, it remains challenging to identify stress scenarios, because it is notoriously difficult to predict the real-life outcomes of such a hypothetical scenario.
Stakeholders and data quality
Bram Kauer, senior advisor strategic risk, at ABN AMRO, then focused on the key success factors for a stress testing framework. He focused on collaboration and breaking the silos, clear governance, robust modeling, regulatory compliance and creating confidence and value for the bank.
Kauer said: “When a bank desires to reap the benefits of stress testing, it is important that the bank goes beyond the scope of a mere regulatory exercise. For example by using the stress test results to make the business plan more robust and facilitate smart capital allocation decisions.”
He added that stress testing is a regulatory requirement to check the feasibility of a contingency plan. “A critical part of stress testing is that the stakeholders, including the board, agree on the scenarios and methodology before conducting the actual test”, he said.
Another key success factor in stress testing, according to Kauer, is data quality, which can be enhanced by involving a central quality assurance team, including model validation, from an early stage. “The central quality assurance team can help improve the whole infrastructure for models, data and IT, such that the processing time can be reduced.” As regulatory stress test requirements will become more demanding, a solid infrastructure is essential.
Complexity and uncertainty
As founder of GloComNet, professor Lex Hoogduin provided some interesting insights as he presented an alternative view on complexity and uncertainty and their association with stress testing.
He discussed how the FAUC, a Framework for Acting under Uncertainty and Complexity, can help to create an environment in which models are optimally used.
”Most crises happen due to a lack of imagination”
This framework can enhance awareness and is designed to help build relevant and plausible scenarios. “It is important that scenarios are built together with a group of different stakeholders and have the approval and attention of senior management”, Hoogduin said.
“The FAUC can help give meaning and rationale behind stress test. Due to the fact that it is difficult to learn from low frequency events using statistics, narrative techniques can be powerful in scenario building and coping with uncertainty.”
One way to conduct narrative stress testing is to imagine what scenarios can ‘kill’ the bank. Hoogduin said: “By imagining any scenario (or final step) that can break the bank, one can create a storyline attached to this scenario and find out what was the event that triggered this downfall. This can provide insight on actions needed to prevent the scenario from materializing.”
Reverse testing and board roles
After the presentations, a roundtable discussion was facilitated, in which the latest developments in the field of stress testing were discussed. This interactive session focussed on scenario definition and integration of stress testing into already existing models.
One interesting comment was that quantifying risk factors is inherently subjective. Considering it is based on expert opinion, it relies heavily on historical data and our collective recent memories.
An alternative to gathering scenarios for stress testing is to use reverse stress testing. In a reverse stress test each risk factor is stressed to find vulnerabilities for banks. If these vulnerabilities are assessed as plausible, they will be covered and monitored.
”If a bank loses reputation, market value will drop regardless of the capital buffer”
The discussion also focused on the role the management board should play in scenario creation. The attendants agreed that the management board should have ownership over the complete stress testing exercise to ensure maximum commitment and follow-up actions.
Including the management board in scenario creation from the start is necessary for them to agree with the scenarios used. It invites the board to provide input, if necessary. Often, board members do have specific concerns regarding the business. Stress test scenarios become more valuable if these specific concerns are incorporated in the scenarios.
Systems, effects and metrics
Then the question was raised whether it’s useful to invest in integration of stress testing in existing systems.
Automation of internal stress testing has more added value than automation of the EBA stress tests, because of the higher frequency of conducted exercises. It is useful to set up a modular approach as components may change over time, like for instance the implementation of requirements for IRRBB and IFRS9.
Should banks focus on the effects on capital or on the effects on liquidity? Liquidity is a binding constraint, as banks can default due to the lack of liquidity, even if the capital buffer is still sufficient.
However, liquidity is currently a hidden risk, because financial institutions can now lend at the ECB as last resort. Nonetheless, banks still need to consider scenarios where this regime is removed and liquidity is once again an issue.
The attendants then discussed what metrics to use to address the outcomes of internal stress tests. These risk measures do not necessarily need to be the same as regulatory ratios. Especially when bank specific risks are analyzed, selecting risk measures based on these specific risks might be more sensible. Risk measures should be chosen that best monitor the risks set out in the risk appetite statement.