Ben O’Brien, Risk Practice Director, Jaywing
The 2015 stress testing results of the UK’s seven largest banks and building societies have been revealed. The results show that there has been significant improvement in stress testing processes since the 2014 review, however the results identified considerable variation across banks.
During the 2015 stress test, the Bank of England (BoE) undertook a review of banks’ stress testing model frameworks – having found that they were weak during the 2014 exercise. For some banks their model management standards were found to have improved. But others needed to make considerable improvements – including implementing and embedding model management policies more fully.
Some banks lacked formal processes to approve stress testing models and had weak model governance. That’s why the BoE highlighted that banks are expected to invest significantly to implement and maintain robust model standards.
But how can banks improve their model standards and what should they do?
Stress testing adds an extra layer of complexity to normal modelling methods. For many credit risk modelling teams, this may be the first time that time-series methodologies have been used. A number of statistical issues particular to econometric modelling need to be addressed in any stress testing exercise, and the regulator is increasingly looking for institutions to be aware of these and show evidence of competency in handling them.
The nature of stress testing means that a combination of top-down vs. bottom-up modelling approaches is needed, and the methodology design needs to address the level of granularity employed at different stages of the modelling.
Consideration also needs to be given to the level of integration that can be achieved between different aspects of modelling. For example, in a dynamic stress test, assumptions on the volume and profile of future lending will inform both net interest margin projections and impairment forecasts. Therefore, central coordination of modelling activities across different business areas is vital to avoid contradictory assumptions being used.
While a top-down approach to modelling default rates may underpin the stress test forecast, it may still be necessary for this to be woven into bottom-up account-level models so that forecasts of complex financial credit risk metrics can be derived, such as IRB Risk Weighted Assets and Impairment.
Modelling is expected to cover all asset classes, but the methods are likely to vary according to the portfolio being assessed. The ability to model economic impacts on profit and risk drivers for different portfolios may depend on the nature of the specific portfolio, the products involved and the available data history. Therefore, it’s important for organisations to consider a variety of different modelling approaches to both revenue and losses.
One of the necessities for successful stress testing is the need to develop a wider framework across an institution that allows enterprise-wide stress testing to be carried out. The reliability of models need to be assessed through the exploration of different approaches. Moving forward, it’s recommended that stress testing models be subject internally to the same governance and control as credit risk models under the CRD IV, which will entail a significant amount of time being allowed for model validation and oversight when planning the development work.
As organisations move towards IFRS 9 compliance, it will be apparent to many that the core economic response modelling at the heart of stress testing is similar to that involved in IFRS 9 expected credit loss estimation. Increasingly, institutions are looking to align these two aspects of their business operations. Indeed, the same infrastructure that drives the economically driven factors in lifetime expected loss ought to be implemented for stress scenario loss calculations.
The Bank of England’s future stress testing expectations
While the BoE recognises that improvements take time, banks are expected to continue to invest significantly to implement model development standards, maintain robust model inventories and strengthen their independent model review frameworks.
An important objective of the concurrent stress testing framework is to support improvement in banks’ own risk management and capital planning capabilities. By strengthening banks’ own stress testing capabilities, banks’ will be able to better assess potential risks to their business, both as part of the concurrent stress tests and beyond. This should support resilience of individual institutions and the system as a whole.
In a way, The Bank of England announcement mirrors everyday life, whilst they recognise they themselves have to do more on stress testing modelling, they still want institutions to up their game – it is a case of do what we say, not what we do!