The evolution of stress testing in banking supervision

Introductory statement by Elizabeth McCaul, Member of the Supervisory Board of the ECB, at the Joint EIB-ESM-SRB Conference “Bank Resolution and the Common Backstop for the Single Resolution Fund: A new decisive step to completing the Banking Union”

Frankfurt am Main, 10 December 2021

Thank you for inviting me to this interesting conference. I am really honoured to contribute to this panel session on stress testing from a banking supervisors’ perspective. I will first argue that, over the past decade, stress testing in banking supervision has evolved from quantifying immediate capital shortages in distressed banks to analysing vulnerabilities in a broader sample of banks under different macroeconomic scenarios. Next, I will explain the various uses of stress testing in our banking supervision, while comparing the stress test approaches in Europe with those in the United States. Finally, I would like to highlight some new developments in the evolution of stress testing.

Stress testing since the Great Financial Crisis

During the Great Financial Crisis (GFC), stress testing became an important tool used by banking supervisors to identify and quantify immediate capital shortfalls in distressed banks. In order to address prevailing uncertainty about the quality of banks’ balance sheets, the European Banking Authority (EBA) in 2011 provided transparency by conducting the first EU-wide stress test exercise on the resilience of the banking sector, and then publishing detailed information about exposures, the main drivers behind the stress impact and the composition of capital on a bank-by-bank basis.

In 2014 the EU-wide stress test methodology became part of the ECB’s comprehensive assessment exercise and − when the Bank Recovery and Resolution Directive was transposed into national laws − the stress test outcomes became relevant for determining the maximum size of a precautionary recapitalisation. To reduce the burden of bank failures on taxpayers, the Bank Recovery and Resolution Directive prescribes that resolution costs are to be borne first and foremost by banks’ shareholders and creditors. State support is to be confined to solvent institutions; it is to be forward-looking in nature and is not to be used to offset losses that the institution has incurred or is likely to incur in the near future. In practice, this means that losses stemming from an asset quality review or from the baseline stress test scenario must be covered by private funds (such as shareholders’ funds or bail-inable debt). Precautionary recapitalisations approved by the European Commission for financial stability reasons are limited to the capital depletion incurred by the bank under the adverse stress test scenario. To date, three banks have been recapitalised under the precautionary recapitalisation framework.

With the establishment of the Single Supervisory Mechanism, ECB Banking Supervision extended the scope of the biennial EU-wide stress test to cover all banks under its direct supervision. In addition, the ECB introduced targeted stress tests in the alternating years, for example on interest rate risk in the banking book (2017) and on liquidity risk (2019). In 2022, more than 100 banks will participate in a climate risk stress test. These stress tests cover all banks and are no longer “pass-or-fail” tests for banks in distress.

Over the past decade, stress testing has evolved to facilitate analysis of vulnerabilities across all banks under alternative scenarios as part of the annual Supervisory Review and Evaluation Process, or SREP for short, in accordance with the Capital Requirements Directive. The change whereby stress testing became an integral part of the supervisory toolkit for all banks rather than solely an instrument to be deployed for distressed assets can also be understood against the backdrop of a strong improvement in the capital position of European banks over the past decade^[1]: the Common Equity Tier 1 ratio of significant institutions increased from 7% in 2007 to 13.5% in 2015, and it continued its upward trend even during the COVID-19 crisis, reaching 15.6% by the end of the second quarter of 2021.

The use of stress tests: capital depletion and stress capital buffers

In both Europe and the United States, the implementation of a stress capital buffer in banks’ “capital stack” has made stress testing an integral part of determining the capital requirements for all banks.

On the heels of the Great Financial Crisis, the US Federal Reserve developed a top-down approach to stress testing in its Comprehensive Capital Analysis and Review for large banks. Using its own supervisory models and based on inputs standardised across banks, the Fed calculates itself the capital impact of an adverse scenario, assuming that the bank’s balance sheet keeps growing and dividends and share buybacks continue according to the bank’s capital plan. If a bank fails the test by incurring a capital shortage in the most adverse scenario, the Fed imposes direct measures on the bank’s management to remediate the shortage. Typically, these measures require restrictions on planned capital distributions and an adjustment of the bank’s capital plan. The US authorities have thus created a direct and easily communicable link between their stress test outcomes and the consequences for the bank. The US stress test depletion was initially applied as a separate requirement in a parallel calculation to the capital stack, which consisted of the sum of Pillar 1 minimum requirements plus the macroprudential capital buffers. Since March 2020 however, the Fed has simplified the capital requirements calculation by abolishing the parallel calculation approach and has introduced a Stress Capital Buffer on top of the stack of the Pillar 1 minimum capital requirements.^[2]

In Europe, the stress test feeds into our SREP and the stress capital depletion is hence part of a “holistic” assessment of the banks’ resilience and continuity. The lag between publication of the stress test results and our SREP decision can blur the link between the stress test and the supervisory measures imposed on banks. Nevertheless, since the 2016 EU-wide stress test, we include in our SREP decisions guidance that is derived from the capital depletion under the adverse scenario. The Pillar 2 guidance is a non-binding capital requirement, but we have added it on top of the capital stack because we expect banks to reserve it as a buffer for times of stress. As a logical consequence of being a non-binding stress buffer, we allowed banks at the outbreak of the COVID-19 crisis in 2020 to dip into their guidance until the economic situation has normalised, hopefully by the end of 2022.

The use of qualitative findings from the stress test exercise

Unlike the US Comprehensive Capital Analysis and Review, the EU-wide and targeted ECB stress tests follow a (constrained) bottom-up approach using significant inputs from the banks in terms of risk exposure and other data. Moreover, European banks generate their stress test projections using their own internal models, the advantage being that the exercise itself requires banks to invest in their risk management capabilities. Another advantage is that the bottom-up approach provides a strong basis for banks to agree with the EBA’s detailed disclosure of more than 15,000 datapoints per bank in the EU-wide exercise. However, the bottom-up approach may lead banks to “game” the exercise and underestimate their vulnerabilities. Over the years, the ECB has addressed the potential for this behaviour by developing a comprehensive system of quality assurance to verify the underlying bank data and to challenge the banks’ projections against calculations based on the ECB’s own top-down models.

For example, the banks’ stress test submissions are scored against measurable criteria for accuracy and quality of information, and timeliness and response time to requests. The quality assurance iterations give the Joint Supervisory Teams an insight into the data quality and IT infrastructure of each bank. A poor performance will also contribute to the Joint Supervisory Team’s view on the reliability of a bank’s own stress projections in its Internal Capital Adequacy Assessment Process. Concretely, our data quality findings in the 2021 stress test exercise contributed to a lower score for risk data aggregation in the assessment of the risk governance for 11 out of 89 banks in the SREP, which ultimately supported a higher Pillar 2 capital requirement. The SREP decisions will also spell out qualitative measures requiring the bank to improve its data quality, risk data aggregation and internal stress test capabilities.

We also draw on stress test outcomes when planning our supervisory activities: for banks that perform poorly in the stress test, we may launch a deep dive or on-site investigation at the bank’s premises a year later. The results of our stress test have motivated on-site investigations in areas such as data and IT infrastructure and risk and capital management, as well as in more specific fields like credit risk or market risk.

Finally, the stress test scenarios developed for the EU-wide exercises have become an important benchmark in our banking supervision. In particular, the baseline and adverse stress test scenarios serve as yardsticks when analysing the bank’s own internal stress scenarios, which underpin their capital and multi-year business plans, for example. Supervisors can thus challenge overly optimistic bank scenarios when assessing the banks’ business strategies.

Looking ahead: new developments

While the character of stress testing has surely changed over the past decade, the key elements of the EU methodology have remained broadly the same. The EU-wide stress test of 2021 focused consistently on solvency and transparency, while maintaining the bottom-up approach as well as the “static balance sheet” assumption. Recent developments suggest that these key elements may see changes over the next decade too.

A first example of change is the Vulnerability Analysis (VA) performed by the ECB in 2020. To reduce the operational burden on banks and their vulnerability to the COVID-19 shock, we decided for the first time to run a top-down stress test across 86 banks using ECB models only. The VA was exceptional, but sparked thinking about a shift to a hybrid approach in future EU-wide stress tests. Some items of the test could be moved to a more centralised level, at which supervisors will draw up calculations using top-down models, while the banks would still look after other parts under a bottom-up approach. The main motivation for such a hybrid approach is twofold: first, moving certain risk areas to a top-down approach would allow us to remove some of the methodological constraints currently serving as a conservative backstop on the bottom-up projections. This would help increase the accuracy of results. Second, a partial top-down approach could reduce the resources used by both banks and supervisors. Overall, the ambition would be to improve the cost-benefit trade-off of the regular stress test exercises.

A second novel example is the methodology for the ECB climate risk stress test 2022, which will be a learning exercise for banks and supervisors alike. The aim is to identify vulnerabilities, best practices and challenges that banks are facing in relation to climate-related risk management. The focus is not on solvency and there will not be a direct capital impact through the guidance under Pillar 2. The test takes a modular approach, comprising qualitative questionnaires and quantitative stress test projections. Module 3 entails a bottom-up stress test targeting the transition and physical risks of climate change; banks will submit starting points and their own projections based on the ECB’s methodology and pre-defined scenarios. For transition risk, the stress test focuses on the potential impact under both short-term and long-term scenarios.

The short-term stress test is based on a three-year baseline scenario assuming a disorderly transition to a low-carbon economy and a static balance sheet. The latter assumption effectively freezes the portfolios of banks at the reference date (end-2021). This key element of the European stress test approach has often been criticised because it does not allow for any management action to reduce the stress impact. The long-term climate change stress test has a 30-year horizon ending in 2050 and applies, for the first time, a “dynamic balance sheet” assumption. This is necessary to shed light on banks’ strategic choices when confronted with the transition towards a more carbon-neutral economy. We will ask banks to outline their strategies under three different long-term scenarios, thereby allowing them to adjust their portfolios to changing circumstances.

Conclusion

Let me conclude. Over the past decade, stress testing in banking supervision has evolved from quantifying immediate capital shortages to analysing targeted vulnerabilities of banks under different macroeconomic scenarios. The stress test provides banking supervisors with a quantification of stress capital buffers, but we are also following up on more qualitative findings. To be sure, the EU-wide stress test exercise and the targeted ECB stress tests are characterised by learning-by-doing and are steadily improving. It also seems fair to say that since the Great Financial Crisis the key features have remained broadly the same, such as the focus on solvency and transparency, as well as the bottom-up methodology and the static balance sheet assumption. Looking ahead, although these key features will no doubt remain in place in the near and mid-term, I also foresee further changes in the purpose and character of the stress test exercises towards assessing emerging risks in a more dynamic and integrated manner.