The Fed’s Stress-Test Overhaul Cuts Volatility, Not Risk

The Federal Reserve has changed how its annual stress test will shape large-bank capital requirements. The final rules make the models and scenarios more open to public scrutiny, add a second market shock for large trading firms, and average two years of stress-test results when setting the stress capital buffer.

The Fed estimates the package will reduce year-to-year volatility in capital requirements by about 50% without materially changing aggregate capital. That is the reported policy outcome. The management question is harder: when a risk measure becomes more stable, how do leaders make sure it has not become less responsive?

The judgment

• Implication: Smoother regulatory capital requirements should make planning more predictable, but they do not justify smoother internal risk assumptions. Finance leaders still need a current-year view that can react quickly when the balance sheet or market changes.
• What would change the conclusion: The case for averaging would weaken if evidence shows it delays capital responses to rapidly rising concentrations, new business models, or losses that last year’s test did not capture.
• Management action: Preserve the latest standalone stress result beside the new two-year average, reconcile the difference, and require an explicit explanation whenever the blended measure is less conservative than the current-year result.

What the Fed changed

On September 30, the Federal Reserve finalized two rules and requested comment on a related model revision. The first rule requires annual public input on stress scenarios and material model changes. It updates the scenario-design framework, adopts models for the 2027 test, and adjusts the testing calendar.

For banks with large trading books, the Fed will now run two global market shock components and use whichever produces the larger loss for each firm. That change increases risk sensitivity where a single market path could miss an important vulnerability.

The second rule changes the calculation of the stress capital buffer. For firms tested in both years, the Fed will average the two most recent supervisory stress-test results. The averaging begins in 2028 so that only models that have gone through the new public-input process enter the calculation.

The Fed also proposed revising its noninterest-income model to better distinguish how different banking business models generate fee income under stress. Comments are due 60 days after publication in the Federal Register.

Lower volatility is useful—but it is not the same as lower risk

Capital requirements that swing sharply from one year to the next can make dividends, buybacks, lending capacity, and business-line investment harder to plan. Some of that movement may reflect real changes in a bank’s exposures. Some may come from scenario design or model changes. Averaging can reduce the noise.

But the same mechanism that dampens noise also dampens new information. If the current test identifies a materially larger loss than the prior year, the two-year average will initially produce a smaller capital effect than the latest result alone. That is arithmetic, not a flaw. It becomes a governance problem only if management mistakes a smoother requirement for evidence that the underlying risk is smoother.

The distinction resembles the one in our analysis of how CECL can affect lending capacity. Regulatory accounting and capital measures influence behavior, but they do not replace an economic view of expected loss, funding pressure, or risk concentration.

Three numbers belong in the capital discussion

A bank board or finance committee should not receive only the final averaged buffer. It should see at least three related measures:

• Latest standalone result. What does the current scenario and current balance sheet imply without averaging?
• Two-year regulatory average. What requirement follows from the Fed’s new method?
• Internal economic-capital view. What loss-absorbing capacity does management believe is needed after considering risks, concentrations, and scenarios that may not be fully captured by the supervisory test?

The reconciliation matters more than any single figure. If the latest result is worse than the average, management should explain whether the difference reflects a new exposure, a harsher but plausible scenario, a model change, or ordinary volatility. If the internal view is lower than the regulatory result, the burden of proof should be higher—not because the regulator is always right, but because optimism about one’s own portfolio is rarely a control.

Transparency creates a new model-risk task

Public input on equations, variables, assumptions, and scenario design can improve accountability and make capital planning less dependent on guessing how a supervisory model works. It should also make model disagreement more productive because banks, investors, and researchers can identify specific assumptions instead of arguing with a black box.

Transparency can create a second-order risk, however: organizations may become better at optimizing to the published test. A portfolio can look resilient against a known framework while remaining vulnerable to a different shock. The Fed’s decision to use the larger loss from two global market shocks is one response to that problem.

Management should apply the same principle internally. A board should ask for at least one scenario that is not simply a variation of the regulator’s case. The purpose is not to outguess the Fed. It is to identify where the bank’s own strategy, customer mix, fee model, funding structure, or trading book creates a loss path that a standardized test could miss.

Averaging should not delay escalation

The main control risk is timing. A two-year average necessarily carries information from the prior period. That is helpful when the latest movement is noise. It is less helpful when risk is building quickly.

Finance and risk teams should therefore establish triggers that override the comfort of the average: rapid concentration growth, material funding-cost changes, weaker collateral, a deterioration in fee income, a counterparty event, or a current-year stress loss substantially above the blended measure. The trigger does not have to dictate a capital action automatically. It should force an explanation and a decision.

That approach is consistent with our earlier analysis of the price of financial resilience: the right question is not whether more capital is always better, but what protection and flexibility each additional dollar buys under a credible downside.

The broader CFO lesson

Most organizations do not calculate a regulatory stress capital buffer, but many use rolling averages, normalized margins, blended forecasts, and multi-period hurdle rates. Those tools can improve decisions when they separate signal from noise. They can also hide turning points.

The control is simple: keep the smoothed measure, keep the latest unsmoothed result, and reconcile them. The free Forecast Scenario Planner can help finance teams put a current downside case beside the base view and link the difference to management action.

The Fed’s overhaul is therefore not a retreat from stress testing. It is a shift in how transparency, volatility, and risk sensitivity are balanced. The practical judgment is to take the planning benefit without allowing a more stable number to become a more comfortable story.

Sources

• Federal Reserve Board, “Federal Reserve Board finalizes changes to enhance the transparency and public accountability of its stress test and reduce volatility in its stress test-related capital requirements”, September 30, 2026.
• Reuters, “Fed finalizes bank ‘stress test’ overhaul”, September 30, 2026.