Stress-testing: still worth the stress?

Stress tests are only as valuable as regulators’ ability to predict future shocks accurately

The surge of supervisory stress-testing that began after the financial crisis had a clear purpose: to identify banks in danger of imminent distress and reassure the market that those still standing were solvent and could survive the downturn.

More than a decade later, issuance costs for bank debt could hardly be lower, even for subordinated bonds. That suggests a level of confidence among investors that they understand the risks in the banking sector. And yet, bankers say the data demands from regulatory stress-testing are, if anything, becoming more intense. 

That has prompted the European Banking Authority to consider a review of its biennial stress-test process before it is run in 2022, with a particular focus on the “usefulness” of the entire undertaking. 

Currently, European supervisors feed the results into their Pillar 2 capital add-ons regime. But there is precious little disclosure of how far this affects capital requirements and how the add-on is broken down: for instance, the contribution of data and governance difficulties revealed during the test versus actual solvency risks uncovered by the results. 

“The effects of the test are very far from the attention of the front office,” one banker told Risk.net recently, which doesn’t sound like the Pillar 2 capital impact really moves the dial.

The Bank of England has gone one stage further, deploying the overall results across the sector as an input for the countercyclical capital buffer applied to all banks. The greater the loss, the higher the buffer. The US Federal Reserve has mooted going further still, by incorporating the test into the individual Pillar 1 capital requirements of each bank in place of the existing fixed 2.5% capital conservation buffer.

There is sympathy for the BoE’s method of deploying the results as a macro indicator of risk. But US banks have been far more critical of the Fed’s so-called stress capital buffer, arguing it duplicates existing risk-based capital requirements such as the Basel framework and the buffers for global systemically important banks.

No better than Basel?

There’s certainly a case for seeing the stress test as a cruder version of Basel risk weightings. The Basel internal model methodologies require banks to calculate capital based on the probability of a range of loss outcomes. A stress test focuses on the impact of a single – or perhaps a few – specific scenarios chosen by the supervisor. Hence the outputs are only as valuable as the regulator’s ability to predict the most appropriate stress scenario.

US lobbyists have argued for an average of impacts over multiple scenarios, and the EBA review will consider adding extra scenarios to the existing single-shock approach.

But would this make the tests any more useful for the banks themselves? The largest, most sophisticated firms already run tests internally, sometimes daily. Alas, the data requirements from supervisors for annual testing rarely match up with internal risk management.

Increasingly, regulators seem to see the stress-test exercise more as a way to identify process failures rather than balance sheet risks. And banks see them as a capital constraint that is misaligned with the Basel risk-based ratios, or even just as an extra layer of bureaucracy. The annual round of regulatory stress-test results is in danger of becoming an expensive circus that doesn’t specifically reduce risk in the system – especially if, as Risk.net uncovered recently, banks are gaming the results. 

If the concern is that banks are modelling and assessing risk incorrectly, then perhaps a more valuable use of supervisors’ limited time and resources would be to deploy their challenger models directly to the Basel framework. The Fed already does this for stress-test outputs. Regulators can then step up scrutiny of those banks with model outputs that deviate a long way from the challenger, rather than imposing a stress test for all risks on all banks. 

Some of this analysis of model risk is already built into the Basel framework. Regulators took a long time to agree on how to validate internal models under the Fundamental Review of the Trading Book. Ensuring rigorous internal model validation might do more to identify any risks the banks are missing than a few weeks of quality assurance each year during the stress test. 

Stress-testing – Special report 2019
Read more