The Proper Use of a Bank Stress Test
A bank stress test is most useful when it is read neither as a prophecy nor as a theatrical worst-case story. It is a structured decision tool: a deliberately adverse scenario is applied to a bank’s starting exposures through a chosen model system, and the resulting capital path is examined for vulnerabilities. The Federal Reserve’s 2026 exercise and the IMF’s euro-area top-down illustration make that common logic visible from two different institutional settings. Together, they support a final judgment about how to use a stress-test number—and how not to use it.
The central rule is simple: describe the result with its conditions attached. “The 32 banks will fall to an 11.2 percent CET1 ratio” is a forecast claim the evidence does not support. “Under the Federal Reserve’s 2026 severely adverse scenario, models projected an aggregate CET1 minimum of 11.2 percent for the 32 participating banks, using the exercise’s assumptions and nine-quarter horizon” is a conditional statement the evidence does support. The second sentence is not weaker. It is more precise about what has actually been measured.
One mental model: conditional resilience
The Federal Reserve and IMF examples can be understood through one chain:
Starting bank conditions → hypothetical scenario → model translation → profitability and credit-risk channels → capital path and minimum → decision under stated limits.
The starting conditions are not a footnote. They include the participating banks, their financial condition, their loan and securities portfolios, their business focus, and their risk characteristics. The scenario is not a forecast. It is a set of macro-financial paths chosen to place pressure on those starting conditions. The model translation determines how unemployment, GDP, property prices, interest rates, spreads, and equity prices become losses, income, provisions, risk-weighted assets, or other capital effects.
The channels then explain how those effects reach capital. Profitability affects the capital available through earnings and retained earnings. Credit risk affects the risk-weighted assets that capital must support. Market and counterparty positions can add other losses. The capital ratio is the resulting summary after these routes are combined over a defined horizon. A minimum is the point at which the modeled pressure is greatest within that path.
This chain makes the Federal Reserve’s 2026 headline intelligible. The aggregate CET1 ratio began at 12.8 percent at 2025:Q4, reached a projected minimum of 11.2 percent during the nine-quarter horizon, and rose to 12.7 percent by 2028:Q1. The report also projected nearly $708 billion in losses, $719 billion in pre-provision net revenue, and only $1.2 billion in cumulative pre-tax net income. Those figures are not separate stories. They are different views of the same conditional system.
The IMF paper illuminates the same chain through a top-down research framework for 91 systemically important euro-area banks over 2023–25. Its baseline and adverse scenarios feed profitability models and credit-risk models. The profitability channel affects retained earnings; the credit-risk channel affects risk-weighted assets. The framework also tests alternative ways of representing risk distributions. Its results—resilience under the baseline, moderate capital pressure under the adverse scenario, and negligible differences between the two alternative distributional mappings in the reported exercise—do not compete with the Federal Reserve’s results. They show how a different model architecture can make the conditional chain more explicit.
The two examples should not be treated as directly comparable measurements. The Federal Reserve report is a supervisory exercise for 32 participating banks under a 2026 scenario and its supervisory assumptions. The IMF document is a working paper describing a top-down illustration based on an earlier euro-area consultation exercise. Their samples, dates, scenarios, models, and purposes differ. What they share is not a common answer about future bank capital. They share a way of asking a resilience question.
That distinction protects against a common category error.