Stress testing plays a major role in banks’ risk management. It is seen as a valuable complement to regular bank portfolio risk measurement in a normal environment. “Regular” portfolio risk measurement, e.g. in the area of lending, is based on, among other things, internal ratings and valuations of collateral. The growing importance of stress testing during the financial crisis reflects, on the one hand, the trouble regular tools had capturing risks adequately in some cases and, on the other, the crisis conditions themselves, demonstrated by, for example, the slump in GDP in 2009.
Stress testing definition
Stress testing is designed to identify environmental conditions that may cause heavy losses in the bank portfolios it looks at. It simulates the impact of hypothetical events on a bank portfolio of, for example, SME loans. So it answers the question “What would happen if …?” How badly would, for instance, a sharp economic slump affect SMEs’ credit standing and thus their rating? Stress testing hence does not deliver a forecast about a normal, anticipated development. It does not say whether the simulated situations will ever occur, when they could occur or with what (usually very low) degree of probability they could occur.
The methodology used in stress testing often involves scenario analyses. Scenario analyses with a limited number of scenarios distinguish between best case, base case, adverse case and worst case. Best case (most favourable development) and base case (expected development, planning) are obviously not stress-type scenarios. Base case developments are often used as reference scenarios to obtain an idea of the severity of a stress test’s impact through a deviation analysis geared to the loss in a stress scenario. The worst case scenario, i.e. the scenario that would generate the heaviest possible loss, is not a suitable stress testing scenario either. No bank can take precautions against the worst case – a bank run, for instance – on the strength of stress testing results if it wants to continue operating successfully in the marketplace. In other words, if the sky falls down on us, it is “game over” anyway. The possibility of disaster scenarios occurring has to be dealt with differently – in the event of a bank run, through deposit insurance, for example. More important is reverse stress testing. In contrast to “normal” stress testing, reverse stress testing takes a business failure outcome as its starting point and asks how great the stress in a scenario has to be for failure to actually occur. The crucial question here is whether the estimated distance to default is so big that a bank can relax or whether – as with all other stress testing results – corrective measures are necessary.
Being creative enough to find scenarios that can hit a bank hard and identify the relevant risk factors (e.g. changes in interest rates and exchange rates) is vital. This is closely linked to the question of who defines the scenarios – banks themselves or banking supervisors (internal vs supervisory stress testing)?
Supervisory stress testing
Internal stress testing is “tailored” to the portfolios examined and their risk drivers. As a result, it is able to identify risk hot spots. Supervisors are quite familiar with such stress testing, which gives them insight into vulnerabilities at individual banks.
Supervisory stress testing has further objectives: for one thing, allowing a comparison of banks’ resilience and, for another, analysing the impact on the banking sector as a whole, with the focus on the stability of the financial system. This objective call, at least to some extent, for a standardised approach because only then can individual results be aggregated usefully. However, this means that the parameters cannot be tailored as well to the different bank portfolios. From an internal bank perspective, the “right” scenarios may not be fully simulated with the right “severity” in some portfolios, even though an economic slump, for instance, will affect every bank. This consequently limits the ability of supervisory stress testing to identify vulnerabilities at individual banks. So, despite supervisory stress testing’s general effectiveness, its significance should not be overrated.
2014 stress test
Unlike its forerunners, the 2014 stress test was embedded in the comprehensive assessment (CA), where it built on and complemented the asset quality review (AQR). The AQR’s task was to assess the prudential reliability of asset valuations in bank balance sheets and to make adjustments where necessary. These adjustments led to a reduction in banks’ capital resources. The CA had a number of sensible objectives intended to protect the European Central Bank (ECB) against taking over supervision of banks with troubled legacy assets accumulated while these banks were being supervised by national supervisors.
The 2014 stress test also adopted the scenario approach. Besides a baseline scenario, an adverse scenario – a stress scenario based on a general economic downturn – was simulated. Under this scenario, changes in the inflation rate and unemployment rate consistent with adverse conditions were assumed. In addition, country-specific assumptions regarding the trend in real-estate prices were made.
The adverse scenario had to be transformed by banks into the impact on the CET1 ratio (measurement of a bank’s core equity capital compared with its risk-weighted assets (RWAs)). The adverse scenario has a negative effect on the quality of a bank’s loan portfolio and results in poorer internal ratings with higher probabilities of default (PDs). This, in turn, increases RWAs, since higher PDs raise the capital requirements under the internal ratings-based approach (IRBA) and thus the CET1 ratio denominator. Moreover, a poorer loan portfolio leads to further loan losses and write-downs which reduce capital and hence also the CET1 ratio numerator. With securities marked to market, e.g. bonds, market risk factors such as the interest rate level were stressed in such a way that this led to lower market values and consequently to losses that were, as a rule, transformed into increases in RWAs using internal models. So a feature of the EU-wide stress test is that supervisors based it on internal models which, after being carefully vetted, could be used to calculate capital requirements.
In addition, the methodology used included the very strict but unrealistic assumption of a static balance sheet over the three-year stress test time horizon. This assumption led to the stress scenario’s negative impact being overestimated because corrective risk management measures to reduce risks that would have been taken if such stress had occurred in reality were ignored.
To pass the test, the participating banks had to meet certain minimum CET1 ratios after applying both scenarios. These were 8% in the baseline scenario and 5.5% in the stress scenario. This means that, before applying the stress scenario, banks had to have a much higher CET1 ratio to be able to clear these hurdles. In the run-up to the launch of the comprehensive assessment, many banks therefore undertook capital increases where they assumed that their “buffer” would not be sufficient for them to reach hurdle level. This effect led to a significant improvement in the participating banks’ capital position ahead of the stress test.
The hurdles were much higher than the statutory minimum requirements. Being extremely conservative, freely estimated thresholds, clearing or failing to clear them was thus an arbitrary result. The very tough parameters outlined above, which went well beyond those for the previous stress test, not surprisingly led to a number of banks tripping over the hurdles and requiring more capital to jump them. More important than clearing the hurdles or not is an analysis of how many percentage points the CET1 ratios dropped by. This analysis illustrates banks’ different risk exposures best. The severity of the stress test was reflected in a significant reduction in CET1 ratios by 3.4 percentage points on average.
Even banks that require additional capital may have a sustainable business model and need not be about to fail. Conversely, the limited usefulness of standardised analysis and the fact that it only delivers a snapshot of the risk exposure at a given moment means that banks which pass the stress test may nevertheless run into difficulties after a while. This is not an argument against such analysis but only for bearing in mind the test’s limits.
It is along these lines that the question posed in the title of this article should be answered: stress testing is no magic bullet – yet, despite its limitations, it makes a significant contribution to ensuring the viability of individual banks and the banking system as a whole. It is far from being a medicine without any beneficial effect on the solvency of European banks. Only if this were the case, would we be reliant on the placebo effect. Instead, there is evidence that stress testing has a distinctly positive impact on European banks’ resilience.
Future of supervisory stress testing
There is always a next stress test around the corner: so even if no stress test is planned for 2015, further regular tests are expected afterwards. The methodology is also under discussion and improvements, albeit no fundamental ones, are still possible. Instead of banks’ internal models, stress testing could be based on supervisors’ own risk calculations. This is currently being debated. Such a fundamentally different approach does not necessarily produce more meaningful results, however. We believe that supervisors can usually measure risk much less accurately than banks themselves.