Krishnan Ranganathan is a former Executive Director at Nomura and has worked on both Dodd Frank and EMIR implementation. An alumnus of Harvard Business School, he is currently a Guest Faculty at various B-schools bridging the gap between theory and practice via his “Eight Bridges” initiative.
Few legislators achieve the strange immortality of becoming an adjective. Sarbanes-Oxley managed it, so did Dodd-Frank. Barney Frank, who died on May 19, was by wide consent the funniest man in the House. But the word that will outlast him is the hyphenated one he shares with a former senator from Connecticut.
The legacy that outlives him
The Dodd-Frank Act of 2010 reached nearly every corner of finance: derivatives, proprietary trading, capital, resolution, and consumer protection. Parts have since been trimmed, delayed, litigated over, or quietly softened. The structure remains. Yet to remember Frank merely as co-author of a sprawling statute is to miss the most ingenious thing attached to his name: not a rule, but a device. The Annual Stress Test attacked the oldest problem in bank regulation: once a rule is fully known, a motivated bank will learn to game it.
The crisis that created it
The idea had an inauspicious birth. In 2009, twenty days into the Obama administration, with markets in freefall, the President held a prime-time press conference and announced that his Treasury Secretary would, the very next day, provide the administration’s new plan. Somewhere in Washington, Timothy Geithner watched and thought: “What? The plan was not ready”. The next morning’s speech was widely panned. Stocks fell 5% that day, then kept falling — a further 22% over the following month — before hitting bottom. The “deer in the headlights” label stuck: he was articulating a policy still being written, on a timeline set by someone else, before an audience looking for certainty.
Yet the speech passed into history for what it contained —the seed of the idea that ended the crisis: a stress test of the 19 largest institutions. By announcing it, the government placed itself behind all 19 — too big to fail, precisely what markets needed to hear. The numbers were modest. Against $180bn for AIG, $443bn in TARP and $831bn in stimulus, the total shortfall came to $75bn: large enough to seem serious, small enough not to spook. Then came something genuinely new — bank-by-bank detail, data previously guarded like nuclear secrets. In crises, opacity is gasoline. In 2009, transparency was water.
When uncertainty was the point
Frank's real contribution was to bolt that emergency improvisation permanently to the floor. The annual test now feeds into capital requirements and constrains how much the largest banks can return through dividends and buybacks. That matters because ordinary capital rules invite optimization. Once the formula is stable, banks hire enough talent to arbitrage it. Stress testing was more unsettling because it preserved supervisory discretion: the scenario changes, the models are not in the banks’ control, and the damage is revealed after the fact. The industry complained for years that the regime was opaque, capricious, even arbitrary. That was not a design bug. It was the design.
The exam that no longer counts
That was then.
The 2026 round arrived last month wearing the same clothes but carrying less weight. It put 32 banks through a hypothetical global recession — unemployment at 10%, commercial real estate prices down 39%, house prices down 30%, equities down 58% — and found they would absorb more than $708bn in losses while remaining above minimum requirements. "Today's results underscore the strength of the banking system," said Michelle Bowman, the Fed's Vice Chair for Supervision.
The most interesting number, though, was announced back in February: zero —the effect this year's results were permitted to have on capital requirements, which the Fed froze at 2025 levels until 2027 while it renovates the framework — a review prompted by industry litigation. The banks knew before sitting the exam that the grade would not count. JPMorgan marked results day with a 10% dividend rise and a $50bn buyback authorisation; Goldman and Morgan Stanley followed within hours — rather like leaving a check-up and ordering a second dessert.
The benign headline, meanwhile, owed less to stronger banks than to a kinder ruler. The 2026 scenario floored 10-year Treasury yields at 2.3%, against 1% last year; since the Fed's models project interest income forward from recent earnings, a higher rate floor manufactures income across the nine-quarter horizon. Projected pre-provision revenue duly climbed for the third year running, delivering the smallest capital drawdown since 2020 — in a scenario that was, on paper, harsher than last year's. The credit losses were real enough. The offset was an assumption.
The wrong crisis
The deeper problem is what the test cannot see. The regional bank failures of 2023 involved no exotic derivatives: long-duration assets funded by flight-prone deposits turned a rate shock into a solvency crisis inside a week. The 2026 scenario is structurally blind to that species of disaster, because the Fed's recession template assumes rates fall as capital flees to safety. Silicon Valley Bank died of rates rising. The test literally cannot fail a bank the way American banks most recently failed. Nor does it examine the banks holding the bulk of the country's commercial real estate: the 32 institutions in the exercise are the giants, while CRE concentration — and the refinancing wall — sits largely with the regional lenders outside the frame. The exam is administered to the students least likely to flunk it, on material from the previous decade's syllabus.
From stress test to ritual
There is, to be fair, a respectable case for the Fed's new openness: public comment, Ms Bowman argues, brings accountability to an exercise that sets binding capital requirements, and the courts were unlikely to keep indulging a black box. But transparency about models converts a test into a syllabus, and banks are excellent students. The political economy of the rest is depressingly familiar. The institutions rescued in one crisis spend the following expansion lobbying to make the next test gentler, narrower, or less consequential. And so the apparatus survives with its teeth filed down: capital buffers frozen, supervisory headcount cut by 30%, Basel "Endgame" reforms trimming capital further, and a growing habit of treating a pass mark as proof of health.
This is the irony Frank would have appreciated. The most effective part of the post-2008 settlement worked because it was hard to domesticate. It introduced uncertainty into a sector that prefers its constraints legible, negotiable, and engineerable. Perhaps the largest banks really are far safer than they were. But the history of finance suggests a standing rule: whenever a supervisory exercise becomes an annual ceremony whose purpose is to certify calm, look closely at what has been left outside the frame.
Nothing to see here, the ritual assures us — provided you keep your eyes inside the frame. The exit is on your left.
