Fed stress-test reforms hand banks a January comment window on the scenarios that set their capital

The Bank Policy Institute's account of the old regime explains where the excess capital came from. "A bank 'passing' the test simply means that it is not required to raise capital or shrink assets immediately, and in that sense, banks 'passed' the test. But banks passed because they now all hold large excess capital as an uncertainty buffer given the randomness of each year's test results," the group said ^[10]. Banks had little insight into what pushed their requirements up year after year ^[11]. As the institute tells it, they held capital against the test itself (or rather, against its randomness) on top of what their losses called for ^[10].

Averaging over two years ^[4] changes that randomness in a way an investor can model. One year's result is half of the averaged figure, and it stays in the average for two consecutive cycles ^[1]. A bad scenario moves a bank's result half as far in the year it runs, then keeps weighing on it the following year. A good year pays off at the same half speed ^[1]. The second scenario each year is a separate guard, aimed at window dressing ^[13].

Part of the reason for the comment window is legal. Banks sued, arguing that tests which change capital must go through formal notice and comment, and Bowman has described publishing the scenarios as needed to comply with the Administrative Procedure Act ^[8]. "Some of these have been difficult lessons, especially for an institution like the Federal Reserve Board," she said ^[9].

I see three ways this could play out for bank capital. Banks could treat steadier results as permission to run leaner, and the excess the institute describes shrinks ^[10]. They could keep the buffer, in which case the reform narrows the swing in requirements without moving their level much. Or the fight moves to January, and the uncertainty that used to sit in the result moves into the comment docket instead. I think the second is the most likely for the first two cycles. A two-year average needs two years of results before anyone can watch it behave ^[1]. A treasurer who built a buffer over years against random outcomes has little reason to release it on the strength of one averaged test. The view is wrong if large banks cut their capital targets after the first averaged result.

Bowman's case for the rule is about who can check the Fed's work. "The public will now have greater assurance that the risks banks take will be reflected appropriately in their stress test losses and their capital requirements," she said ^[14]. Governor Michael Barr's second objection runs the other way. "Calcified models will also allow banks to optimize their balance sheets to the test, rather than focusing on underlying risk," Barr said ^[7]. Reform proponents say that worry overstates how far a bank can change its balance sheet for a snapshot ^[12]. Alongside the published tests, the Fed will use exploratory stress tests and banks' internal tests to shape supervisory priorities ^[5]. The report does not include an estimate of how much the final rule changes any bank's capital requirement.