Averaging Is a Capital Rule

The Federal Reserve Board’s final rule on how large banks’ stress capital buffers are calculated was published in the Federal Register on Friday 2 October (91 FR 62636, Docket No. R-1866). It takes effect on 1 December. The averaging it adopts will not touch any bank’s requirement until 1 January 2029.

The gap between those dates is the substance of the rule. Under it, the Board will no longer treat one year’s stress test as the whole answer. It will average the last two. That is a change to the capital rule, not to the test: the test is still run each year and the same projected decline in capital is still measured. What changes is how much of any one year’s result a bank has to hold capital against.

What the buffer is

The stress capital buffer dates from a framework the Board set up in March 2020. For each covered firm the Board takes the difference between its starting and its lowest projected common equity tier 1 ratio under the severely adverse scenario. The notice calls that the “stress capital decline component.” It then adds four quarters of the firm’s planned common dividends, as a share of risk-weighted assets, and applies a floor of 2.5 percent. A firm can change its planned dividends or ask for reconsideration after it sees a preliminary figure. The final figure becomes part of its ongoing capital requirements.

Firms subject to Category I to III standards are tested every year. Category IV firms are generally tested every other year. That distinction decides who benefits from averaging.

How the averaging works

For a firm in two consecutive annual tests, the Board will average the stress capital decline from each of the two most recent tests, with equal weight. The notice sets the order of operations: average the declines, then add the dividend component, then apply the 2.5 percent floor. The dividend component is not averaged; it is the firm’s own plan for the year the requirement is in force. For illustration only, and not any bank’s figures: a firm whose projected decline moves from 3.0 to 5.0 percentage points would see the decline component rise to 4.0 in the first year, and reach 5.0 only if the second test repeated the result.

Two exceptions matter. A Category IV firm that is not in two consecutive tests is not averaged; its requirement follows its latest test. And when a “material change” leads the Board to recalculate a firm’s requirement, it will not average, because carrying the old balance sheet forward “may misrepresent the firm’s risk level,” in the Board’s words.

Who gains, on the Board’s numbers

The notice’s Table 1 recalculates requirements from 2020 to 2025 as if the rule had been in force. Average year-over-year movement in a firm’s requirement falls from 64 basis points to 47, which the Board calls a decline of about 27 percent. By category, this publication calculates the following from the table: Category I falls from 46 to 23 basis points (50 percent), Categories II and III from 98 to 73 (about 26 percent), and Category IV from 45 to 36 (20 percent). The notice says averaging offers “minimal reduction in volatility” to firms tested every other year, who can opt in by taking part in consecutive annual tests. On the same table, the one-quarter delay alone, with no averaging, would bring Category IV to 43.

The average level barely moves: 3.83 percent under the current framework, 3.80 under the rule. Staff estimated in a memorandum to the Board dated 31 August that the rule would have lowered average requirements by three basis points. The notice attributes a small downward effect to the interaction of averaging with the floor, for firms whose requirement moves above and below 2.5 percent. Commenters who opposed averaging argued that banks would distribute more capital to shareholders. The Board replied that the goal is to reduce volatility, “not to reduce their overall level,” and rejected averaging only increases, which it said would “mechanically reduce capital requirements.” Its own table puts that alternative’s average at 3.68 percent.

The cost the Board names is timeliness. It describes the primary concern as “the slower responsiveness of stress capital buffer requirements to changes in firm risk profiles and economic conditions.” The average gap between the date of the financial data and the date a requirement takes effect rises from 11 months to 18. The rule also moves the annual effective date from 1 October to 1 January, so a firm has two quarters to comply with a new figure rather than one.

The calendar

The averaging does not start when the rule does. On 4 February the Board voted to maintain current stress capital buffer requirements until 2027, so that new ones could be calculated with models that reflect public feedback. The notice says each firm therefore keeps its current requirement until a new one takes effect on 1 January 2028, calculated without averaging. The regulatory text enforces this through a transition paragraph: the calculation in calendar year 2027 uses the non-averaged method. The first averaged requirement uses the 2027 and 2028 declines for firms in both tests, and takes effect on 1 January 2029. The Board gives the reason as ensuring that requirements are calculated only with models that have been subject to public input, and it did not allow firms to opt in earlier.

That is later than the April 2025 proposal, under which the first averaged requirement would have used the 2024 and 2025 tests. The Board’s 29 August 2025 release had said that, if the rule were finalised as proposed, that year’s results “will be averaged with those from 2024.” The mechanism is as proposed; the start is three test cycles later. The Board’s own 30 September announcement says it will “begin averaging stress capital buffer requirements in 2028,” and the American Bankers Association’s summary says the change “will take effect in 2028.” The notice’s own dates show that 2028 is the year of the test; the requirement it produces is effective in 2029.

What is not known

The Board’s record of votes lists the final rules approved on 29 September by six votes to one, with Governor Michael S. Barr against. His statement is addressed to the companion transparency rule, which he says will make the stress tests “less responsive to emerging risks”; it does not discuss averaging. The Board’s analysis rests on 2018–2025 results, so it cannot say how averaging would behave in a real downturn. The staff memorandum estimates that the whole package, this rule together with the transparency rule and the model changes, would reduce year-to-year volatility by about 50 percent. That is a staff estimate made before the rules were final, not a result.

Averaging works in both directions. A bank that reduces its risk will see only half of the improvement in its requirement at first, and a bank that adds risk will be charged only half at first.