CORRECTED-BREAKINGVIEWS-Goldman’s exam score raises study-guide questions

By Stephen Gandel

NEW YORK, July 2 (Reuters Breakingviews) - Goldman Sachs GS.N has gone from troublemaker to star pupil. The investment banking powerhouse led by David Solomon aced its Federal Reserve stress test a year after being one of the worst performers. Its top score frees up some $20 billion, paving the way for a 33% dividend increase and more stock buybacks. A mysteriously strong trading grade, however, casts doubt on plans to hand out cheat sheets.

Created after the 2008 crisis to help shore up financial stability, the Fed's annual reviews help determine how much capital banks need to cushion potential losses. Goldman and four others challenged the results in 2020, the first year that appeals were allowed. Solomon argued last year that regulators were too pessimistic with their projections.

The 2025 improvement in Goldman’s trading business assessment was strong, but also strange. In previous tests, the Fed estimated losses as high as $21 billion from buying and selling stocks, bonds and commodities. This year it was a mere $300 million. Other banks did a bit better on this portion of the exam, too, in part because market conditions were less harsh. Still, if Goldman’s hypothetical trading losses had tracked the 37% average of its four closes peers, they would have tallied about $13 billion.

Although moving private equity losses into a different category also helped Goldman, the biggest reason for the uplift probably relates to the Fed’s methodology. It picks a single day, usually in the fall of the previous year, to test a bank's trading positions. If the date selected, October 11 this past year, happens to be idiosyncratic for some reason, the results can be unrealistic. Indeed, the Fed said “atypical client behavior” flattered some of the 2025 outcomes.

Banks routinely grouse about the tests, especially the cloak over how the Fed grades, in large part because their own internal analyses often yield different numbers. Goldman itself calculated nearly $4 billion of trading losses. The fix isn’t to provide more helpful cheat sheets, but rather to use more durable inputs. For example, average trading exposures over time would make more sense than a single day. Collecting so much extra data might be costly, but a relative bargain to financial calamity.

Follow Stephen Gandel on LinkedIn and X.

CONTEXT NEWS

Goldman Sachs said on July 1 that it plans to increase its common stock dividend by 33%, to $4 a share from $3, following results from the U.S. Federal Reserve’s comprehensive capital analysis and review stress test.

The Fed released the results on June 27, with all 22 banks passing the annual exam.

In December, the central bank’s regulators said that, due to an “evolving legal landscape,” they were likely to propose significant changes that would improve the stress tests’ transparency and consistency.