Bankers call for overhaul of EBA stress tests

Support for multiple scenarios, but only if fixed assumptions and variables are scaled back

Europlaza building in Paris, home to the EBA

Photo: 11h45_Paris La Defense

The European Banking Authority is failing in its effort to make its biennial stress tests more relevant for internal bank risk management, according to senior executives at two of the European Union’s largest lenders.

“The exercise is too complex, it is too unrealistic, and it is not aligned with bank risk management,” Diego Maria Serra, senior solvency manager at Spain’s BBVA, said of the EBA’s stress tests.

“[The EBA] wants banks to merge stress testing into the internal process of risk management, but to be completely honest, many banks are not doing that – it’s just a one-off exercise that you deliver and then forget about,” he added.

The exercise is too complex, it is too unrealistic, and it is not aligned with bank risk management

Diego Maria Serra, BBVA

The bankers’ main criticism is that the stress tests are based on a single scenario that might be obsolete by the time the results are available. The scenario for the 2023 EBA stress test was released in January, with the results published in July.

“In those six months, a lot of things happened, so maybe the scenarios get outdated,” said Serra.

Another complaint is that the EBA’s growing use of standardised top-down assumptions for components such as non-interest income has made the stress tests less useful for internal risk management.

“Each time we go for standardisation of the rules or for some additional layers of methodological buffers, we are losing the value-added for risk analysis,” said Marc Irubetagoyena, global head of stress testing at BNP Paribas.

Serra also questioned the validity of some of the EBA’s standardised assumptions – for instance, that banks will continue paying dividends to shareholders even if their capital is depleted in a stress scenario. “When we were in the Covid pandemic, they told us we cannot pay dividends,” he said, “but when they perform the stress test exercise, they tell us that we have to pay those dividends, so it is hugely inconsistent, it is not realistic.”

Serra and Irubetagoyena were speaking at a webinar hosted by the European Banking Federation and S&P Global Market Intelligence on September 12.

A better scenario

One remedy to the problem would be update the scenario during the stress-testing process if the initial version is overtaken by events. This year, after the collapse of Silicon Valley Bank, the EBA sought additional data on bond holdings that are not marked at fair value on banks’ balance sheets, which was one of the risk factors that contributed to the failure of SVB.

However, Serra believes the current methodology makes it difficult to introduce substantive changes to the scenario mid-test. “We would be able to provide the data, but from the EBA and ECB side, I don’t think it would be possible to pass all the evaluation, quality assurance [and] data quality checks,” Serra said. “Maybe they have to shorten the process and risk losing some data quality.”

Related

Do all roads lead to multi-scenario Fed stress tests?

After SVB downfall, EBA stress test seeks out unrealised losses

EBA eyes top-down stress test for credit risk

Another option is to run multiple scenarios to ensure other potential risk paths are covered. “It would be meaningful to look at a scenario of a long pattern of high interest rates and also a long pattern of low interest rates,” said Irubetagoyena. “For some other risk factors […] it wouldn’t be that informative, so we should select the right risk factors on which we create that differentiation.”

In addition to making the stress test more relevant for risk management, he added that it was important to build relevant scenarios because the stress test results feed into EU Pillar 2 capital requirements via the supervisory review and evaluation process (SREP).

“For feeding the SREP process, it would be an improvement to have meaningful risk factors and a broader range of possible outcomes covered,” said Irubetagoyena. “If you don’t have a representative set of scenarios, it can be a concern because you take a one-sided view on the Pillar 2 expectations.”

Again, however, both bankers expressed concern about moving to a multi-scenario stress test without first simplifying the existing methodology. Serra said he would be “scared” of trying to run multiple scenarios using the current process.

“For market risk alone, the EBA published 740 variables, so just imagine multiplying that by three or four times,” Serra said. He added that extra scenarios could also extend the current six-month timeline to publish the results. Since the stress test is run using the balance sheet data from the end of the previous year, the results become less meaningful as more time passes.

Serra suggested one way to streamline the process would be to allow banks to use the same models as for their internal capital adequacy assessment process (Icaap), which also forms part of the SREP dialogue with supervisors.

“If you use your Icaap models with the scenarios, you can also run some idiosyncratic scenarios only for your own bank, and it would be perfectly integrated – it would be relevant for the market, for your supervisor, and for your own bank,” he said.