How to ensure AI and Machine Learning scores are fair and ethical
2018 Stress Test Results: Banks look safe under no deal Brexit, but are the models right?
17 December 2018
On 28th November 2018 the Bank of England (BOE) published the results for the 2018 Concurrent Stress Test run by the seven largest banks in the UK. Banks had to present their capital position across a 5-year forecasting horizon, under a stress scenario more severe than the 2008 financial crisis. While the results overall were positive, banks were issued a reminder to continue to up their game in terms of the models they use to stress test their portfolios, especially considering the prospect of a ‘disorderly’ Brexit.
UK Banks deemed strong enough to continue
The strong capitalisation of the UK financial system currently allows the major UK banks to maintain their Common Equity Tier 1 ratio (CET1) after the stress at a level double pre-crisis values. Liquidity is also sustained in the system, as after the financial crisis major UK banks have substantially reduced their dependence on wholesale funding and maintained significant reserves of high-quality liquid assets.
According to the BOE the shocks to macroeconomic variables modelled as part of this stress scenario broadly match the likely impact on the economy of a no deal Brexit and as such the results published provide a level of assurance that the UK financial system would be able to cope with a disorderly Brexit. Temporary volatility would be expected in the sterling markets, but UK banks are deemed strong enough to continue serving businesses and households, maintaining the credit supply to the real economy.
Model management needs to improve
The results are positive overall, but they were issued with a reminder to financial institutions to continue to up their game in terms of the models they use to stress test their portfolios. Alongside the 2018 stress-test results, the BOE published a document providing an assessment of the effectiveness of the participants’ model risk management (MRM) frameworks.
Mindful of the sizeable gap in the banking industry between model management for stress testing models and for other regulatory models, in April 2018 the Prudential Regulation Authority (PRA) released a Supervisory Statement (SS3/18) after consultation with the banking industry, delivering expectations on firms to improve the control and governance of the models used in stress testing activities. The BOE assessed banks’ stress testing model frameworks against those principles as part of the 2018 stress testing exercise.
MRM practices in stress testing need to improve to support greater credibility of these important results and to give confidence to the public that the banking system that failed them a decade ago is indeed now able to cope with extreme economic scenarios and real-life threats such as a no deal Brexit. Whilst the Old Lady of Threadneedle Street may assess the final outputs, it is the responsibility of the individual banks to understand the unique nature of their portfolios and to be able to measure performance under different circumstances, avoiding a repeat of the shockwaves that followed the Lehman Brothers’ fall 10 years ago.
Key areas of concern
The guidance that the Prudential Regulation Authority (PRA) gave were sensible model management practices centred around model definition and identification, model governance, implementation and validation. And whilst the PRA stated that banks have generally showed an increased awareness of the need of having in place an effective stress testing MRM structure, areas of concerns were highlighted:
- Progress varied across banks, with some banks needing to make substantial improvements to achieve the appropriate standard.
- A culture of stress testing is still not always fully embedded in some banks’ management and in particular the Boards of some institutions are still too detached from the stress testing models and do not have a good handle on model limitations. This undermines their potential for use as risk management tools underpinning important business strategy decisions.
- Expert-judgement driven adjustments to model outputs were not always justified by an appropriate level of empirical analysis.
The PRA is planning to provide feedback to individual banks about the specific areas requiring improvements and is considering including some of the qualitative review outcomes in next year’s publication of banks individual assessments. Additionally, the PRA has noted that findings from these institution-based assessments may be used to inform the setting of the bank-specific PRA buffer, making the need for banks to improve their stress testing modelling frameworks grounded in very real financial penalties.
While the 2018 results may look good for the UK banks involved, there is a stark reminder to all financial institutions to improve modelling, especially while Brexit uncertainty continues. That’s why during 2019 all lenders should pay more attention to stress tests than previous years.