The Federal Reserve Board has finalised two rules changing how it designs and applies its annual stress test of large banks. One creates recurring public input on hypothetical scenarios and significant model changes. The other averages results from two consecutive annual tests when setting certain banks’ stress capital buffers. The package also adopts models for the 2027 test and opens a separate consultation on a possible revision to fee-income modelling.
The test is intended to assess whether large banking groups could absorb losses in a severe hypothetical downturn and still support households and businesses. Its results feed into a firm-specific stress capital buffer, an additional layer above minimum capital requirements. The buffer is based in part on how far a bank’s capital ratio is projected to fall under stress, alongside an allowance for planned dividends.
More notice of scenarios and models
The transparency rule requires annual public input on the scenarios and any material changes to supervisory models. Scenarios describe an imagined economic and financial shock; they are not forecasts. Models estimate how a bank’s balance sheet, income and losses might respond. Making these elements more visible gives outside parties a formal opportunity to comment on the framework before it is used.
The Board has adopted the models for the 2027 stress test after a prior period of public feedback. The calendar for that cycle sets out proposed scenarios for comment by 10 January 2027, with final scenarios due by 28 February. The final rule also revises the scenario-design framework and makes procedural changes to the test timetable.
For banks with large trading books, the test will include two global market shock components each year. The Board will use whichever component produces the larger losses for each firm when calculating its result. That change is intended to retain a demanding test while making the market shock process more clearly specified.
Buffer averaging comes later
Under the second rule, a firm that takes part in stress tests in two consecutive years will have its stress-capital decline component calculated using the average of its results in those two tests. This is a symmetric two-year measure: the latest result and the preceding year’s result receive equal weight. If a firm does not participate in consecutive tests, the rule instead uses its most recent result. That distinction matters for some firms tested every other year.
The implementation dates need to be separated. Stress capital buffer requirements will move from a 1 October to a 1 January effective date, giving firms an additional three months to adjust; the first such requirement is due to take effect on 1 January 2028. Averaging itself starts one year later, with requirements effective from 1 January 2029. This staging is designed to ensure that the average uses results from tests whose models have incorporated public input.
The change addresses abrupt year-to-year swings in firm-specific requirements, not a promise that banks’ capital levels will fall. The expected effect is about 50 per cent less annual volatility in capital requirements, with no material change expected in aggregate requirements. Retrospective modelling across the 2024–26 stress-testing cycles indicates an aggregate reduction of roughly 1 per cent of required common equity tier one capital. These are model-based projections, not a guarantee of any individual bank’s future buffer or lending decisions.
A continuing debate over resilience
Stress tests matter beyond regulatory calculations because banks may respond to a higher buffer by retaining more earnings, issuing capital or changing the size and composition of their lending. A more predictable requirement may make planning easier. But reducing volatility must be balanced against keeping the test responsive to new risks and sufficiently rigorous. Governor Michael S. Barr dissented from the transparency rule, warning that public disclosure and annual consultation could make models less responsive to emerging vulnerabilities and encourage banks to manage towards the test. He supported using multiple market-shock scenarios and applying the larger loss.
A separate proposal would revise the model used to project noninterest income, including fee income, to better reflect differences in how banks generate revenue. That element is not final: it is open for comment, with submissions due 60 days after publication in the Federal Register. If adopted, the revision would replace the current model.
The rules therefore change both the process and the timing of the capital calculation, while leaving a further modelling question unresolved. Proposed scenarios for the 2027 test are due for public input by 10 January, with final scenarios by 28 February; proposed material model changes for the 2028 test are scheduled for publication by 31 August 2027. The first averaged buffer requirement is scheduled for 2029, after the new input process has had time to shape the models used in the calculation.
