On Wednesday the Federal Reserve published its 2026 stress test results. All 32 banks passed. They absorbed $708 billion in hypothetical losses under a scenario that included 10% unemployment, a 39% collapse in commercial real estate, a 30% decline in home prices, and a 58% equity drawdown. The aggregate common equity tier 1 ratio fell 1.6 percentage points. The test worked as designed, with one exception: the results do not count.
In February, the Fed voted to freeze stress capital buffers at their current levels through 2027. The buffers that determine how much capital each bank must hold will not change based on this year's exam. The test ran. The grades published. Nothing moves.
The Architecture
The stress test was born from the wreckage of 2008. Tim Geithner's Supervisory Capital Assessment Program tested the 19 largest banks and published the results with unprecedented transparency. The market confidence it restored became the model. Dodd-Frank codified the practice as an annual exercise. But as the Fed expanded the roster and refined the scenarios each year, the supervisory models grew more complex and less visible. The Fed derived from the results a stress capital buffer unique to each bank. The buffer set a floor on how much capital the bank had to hold and a ceiling on how much it could return to shareholders. The opacity became the mechanism. If you do not know what the examiner is testing, you cannot study for the exam. You have to be prepared for everything. By 2024, the stress test had become the single most consequential annual event in American banking regulation.
For fifteen years, the banking industry tried to open the book. The Bank Policy Institute called the models volatile and arbitrary. In court filings, the industry described the resulting capital requirements as "vacillating and unexplained." The Fed resisted. The models stayed closed. Capital requirements stayed binding.
Then the Supreme Court killed Chevron deference. In Loper Bright v. Raimondo, decided June 28, 2024, the Court held that agencies cannot claim deference for their own interpretations of ambiguous statutes. The ruling concerned a fishing company's challenge to an observer-funding mandate. Its blast radius reached the entire administrative state. For financial regulators who had operated under the assumption that courts would defer to their technical expertise, the ground shifted overnight.
The Filing
On December 23, 2024, the Fed announced it would open its stress test models to public comment. On Christmas Eve, the American Bankers Association, the Bank Policy Institute, the U.S. Chamber of Commerce, the Ohio Bankers League, and the Ohio Chamber of Commerce filed suit in the Southern District of Ohio. The complaint alleged the stress test models had been "adopted in secret" in violation of the Administrative Procedure Act. No public comment had ever been solicited on the models that determined billions of dollars in capital requirements. The statute of limitations on challenging the original framework was expiring in February 2025. They filed or forfeited the right to challenge forever.
The timing tells the story. The Fed moved first, one day ahead of the filing. Vice Chair for Supervision Michelle Bowman said the freeze would "give us the opportunity to correct any deficiencies in our supervisory models based on that feedback." The models the Fed had enforced for fifteen years may have deficiencies. The Fed chose not to bind capital requirements to results from those models while it determines what the deficiencies are.
The Return
JPMorgan moved within hours of the stress test release. Dividend raised to $1.65 per share from $1.50. New $50 billion share repurchase program, effective July 1. The firm's stress capital buffer stays at 2.5%. Its total CET1 requirement stays at 11.5%. The bank did not wait for the results because the capital buffers were already locked.
KBW analysts estimated that if this year's results had been binding, Morgan Stanley, Citigroup, Citizens Financial, and KeyCorp would have seen the largest reductions in their capital buffers. The test, if enforced, would have freed capital. Instead the buffers remain at levels set by prior exams, and the banks that would have benefited most from updated results are paying the cost of an obsolete score.
The conventional reading: this is a pause. The Fed updates its models. Public comment improves transparency. New requirements take effect in 2027, more defensible and durable.
The alternative: a stress test whose models banks can see, comment on, and challenge in federal court is a fundamentally different instrument. The original worked because the models were black boxes. Open the book and the exam measures test-taking ability, not resilience.
This is the same transformation unfolding across the Fed. Warsh removed forward guidance from the FOMC statement. Rate consensus now swings 75 basis points in a week. The stress test gave banks a binding constraint on capital. Loper Bright removed the legal foundation. The constraint became advisory. The banks did not need Congress. They needed a courtroom in Ohio and a Supreme Court that had already done the work.