Capital buffer only after the stress test: BaFin overhauls its Pillar 2 Guidance for smaller banks

On 8 October 2026 Germany’s Federal Financial Supervisory Authority (BaFin) published supervisory notice 08/2026 (BA), and with it a new methodology for its Pillar 2 Guidance (P2G), known in Germany as the Eigenmittelempfehlung. The guidance is the capital cushion a supervisor expects a bank to hold on top of all binding requirements and buffers, so that it can come through a severe crisis without eating into its mandatory buffers. Under the notice, an institution receives guidance only if its capital ratios in the adverse scenario fall below its required capital plus 500 basis points. The scenario is that of the stress test for less significant institutions (LSIs). The notice runs until 30 August 2028. The first zeros from this year’s stress test take effect on 1 December.

BaFin framed the public story itself. Institutions would be “noticeably relieved”, its announcement says, and the press release on the stress test speaks of “about 40 per cent fewer institutions” receiving guidance. That is accurate, but it counts institutions rather than capital. The documents establish one thing: a threshold placed in front of an unchanged formula. According to figures from the supervisor, reported by Börsen-Zeitung on 24 September, the number of institutions with positive net guidance falls from 616 to 369. Working through who is affected, on an assumption explained below, shows that only some of the institutions above the threshold actually lose positive guidance. For most of them, what disappears is a letter whose result was zero anyway. None of the sources reviewed puts a figure on the capital released. The derivations below rest on the formula and the published numbers. They are not a forecast for any individual bank.

The formula stays, but a threshold now stands in front of it

The notice describes the mechanics in a sentence that preserves more than it changes: the formula is “based on the previous methodology”. Capital depletion in the adverse scenario, meaning the drop from the starting capital ratio to its low point under stress, is multiplied by a calibration factor that BaFin has set at 65 per cent for 2026. As before, the result is capped at 1,000 basis points, and it continues to be offset against the 2.5 per cent capital conservation buffer (CCB) under Section 10c(1) of the German Banking Act (KWG). That leaves at most 750 basis points net. According to an article by the Bavarian cooperative association Genossenschaftsverband Bayern, published in December 2024, the factor was already 0.65 in the 2024 stress test and 0.75 in 2022. What is new is the threshold, which BaFin had announced on 30 March, and the way institutions learn of their result.

This threshold is more demanding than the rulebook requires. The SREP guidelines of the European Banking Authority (EBA), to which the notice refers, already permit a supervisor to set no guidance where an institution does not breach its TSCR in the adverse scenario. BaFin adds a further margin of 500 basis points. Take a typical institution without systemic importance. Its buffer stack is the 2.5 per cent conservation buffer, the 0.75 per cent countercyclical buffer and the sectoral systemic risk buffer for residential real estate, which adds at most another percentage point. The threshold sits above that entire stack. An institution that no longer receives guidance therefore still holds all its buffers after the stress, plus a reserve of 75 to 175 basis points of Common Equity Tier 1. That is my own calculation from the rates in force, not a statement by BaFin, but it explains why the supervisor calls the threshold a “risk-based early-warning threshold”.

The chart shows a second threshold that barely features in the debate. Because the conservation buffer absorbs the first 250 basis points of any guidance, net guidance remains only once capital depletion exceeds 2.5 divided by 0.65, roughly 3.85 percentage points. According to the press release published on 24 September, aggregate depletion across all participants was 3.8 percentage points, just below that mark. The aggregate says nothing about how individual institutions are distributed. It does explain why so many institutions already came out at zero net under the old methodology: banks with moderate losses were covered by the buffer.

In most cases, the template replaces the letter

The first noticeable change concerns how an institution learns of its guidance. Under the notice, the stress test template indicates whether guidance is likely (tab A2, cell J81). An institution that sees a zero there will “as a rule” receive no letter from BaFin. The zero counts as communicated through the template and takes effect on 1 December of the stress test year. Nikolas Speer, BaFin’s Chief Executive Director of Banking Supervision, put it this way when he presented the results on 24 September: “If no own funds are recommended, the process ends right there.” Until a new letter arrives, however, the previous guidance continues to apply. Where guidance does follow after all, the institution hears from BaFin by 30 November at the latest.

The notice itself acknowledges a limitation of the indicator. It tests only whether capital ratios fall below the threshold and leaves out the offset against the conservation buffer, because the calibration factor is not yet known when the stress test starts. An institution may therefore see guidance indicated in the template and still end up at zero net, in which case it does receive a letter. According to the supervisor this applies to 129 institutions. The obvious objection, that institutions now work out their own guidance, has been anticipated. The Bundesbank’s completion instructions make the indicator subject to “supervisory data quality assurance and discretionary case-by-case decisions”, and the notice announces corrections “on the basis of material data errors in the LSI stress test”. For mergers, BaFin determines the guidance as before by weighting it by total risk exposure and takes the threshold into account, and where an institution within a group receives no guidance, none is recommended at group level either.

The forty per cent are measured against 616 institutions, not 1,113

The figure that stuck from the press conference has a denominator that the press release leaves out. Nikolas Speer said in his speech that “40 per cent of the institutions that would mathematically receive guidance, but are very comfortably capitalised”, would no longer receive any. The accompanying slide shows roughly 50 per cent of institutions with guidance under the old methodology and roughly 30 per cent under the new one, and states that “2 in 5 institutions with previous P2G benefit from the threshold”. Börsen-Zeitung supplied the absolute numbers, citing the supervisor: 498 of the 1,113 participants fall below the threshold, 129 of these show zero net after the buffer offset, which leaves 369. Under the old methodology the figure would have been 616. As 247 is 40.1 per cent of 616, the figures are consistent.

These numbers make it possible to work out who is actually relieved, on one condition: institutions below the threshold receive the same net guidance under the old and the new methodology, because formula, factor and cap have not changed. On that basis, all 247 institutions that lose positive guidance sit above the threshold. The remaining 368 of the 615 institutions above the threshold were already at zero net under the old methodology. For them the main change is that no letter arrives. For 247 institutions a recommendation disappears, and the supervisor does not publish its size. For the other 498, nothing changes in the calculation. The split is my own calculation from the supervisor’s figures. BaFin itself does not give it.

Nikolas Speer added that the approach “leaves no institution worse off, but many better off”, and as far as the design goes, the calculation bears him out. A threshold placed in front of an unchanged formula can only relieve. The result for an individual bank can still shift. Nikolas Speer himself spoke of a “deliberately tightened calibration of the scenario”, and BaFin resets the factor after each stress test. A footnote in the notice states that the 500 basis point margin “may be adjusted in future stress tests if necessary”. Anyone who treats the threshold as fixed in capital planning is therefore planning with a parameter that BaFin explicitly keeps open.

The ECB sorts significant banks into overlapping bands

For the significant institutions it supervises directly, the European Central Bank (ECB) approaches the same problem differently. Since 2021 it has used a two-step approach. First, it places each bank in one of four buckets according to capital depletion in the EU-wide stress test, which the chart on its website divides into steps of three percentage points. Second, it sets the guidance within the bucket’s range, “or in exceptional cases outside it”, taking into account the bank’s risk profile and the year in which its capital ratio reached its low point. In the ECB’s own example, a bank whose ratio falls by eight percentage points lands in the third bucket and is “likely to have a Pillar 2 guidance of no more than 2.75%”. There is no cap, and the buckets have not changed since 2021. According to the ECB’s press release published on 18 November 2025, average guidance across significant banks falls to 1.1 per cent in 2026, from 1.3.

Two differences matter in practice. The ECB measures only depletion, a difference between two ratios, and has published no threshold on the level of the stressed capital ratio. BaFin places exactly such a level threshold in front of a continuous formula. And the ECB says of its buckets that each range overlaps with its neighbours “to avoid cliff effects”. The EBA guidelines spell out the same aim only for buckets, which BaFin does not use, so there is no breach of the rules. BaFin’s design has a different advantage: it can be calculated, and each institution sees its result in its own survey form. The Dutch central bank also uses its LSI stress test as the starting point for guidance, according to an announcement published on 7 September, and it has not published a threshold either.

The threshold creates a cliff for banks with heavy depletion

Anyone who has traded barrier options will recognise the structure. Above the threshold the guidance is zero, below it the formula applies, a knock-in, in derivatives parlance. With barrier options, the area around the barrier is the hardest to hedge, because a few basis points decide the entire payout. Something similar applies here, with a restriction that defuses the cliff considerably. A jump arises only for institutions whose depletion exceeds the roughly 3.85 percentage points from which any net guidance remains. A bank with moderate losses that falls just below the threshold still ends up at zero. For a bank with six percentage points of depletion, the worked example by Genossenschaftsverband Bayern would leave guidance of 1.4 per cent, triggered by missing the threshold by a few basis points.

In the statements and reports reviewed, nobody has so far criticised this cliff, the German Banking Industry Committee (Deutsche Kreditwirtschaft) included. The committee, speaking through the National Association of German Cooperative Banks (BVR) on 24 September, did not comment on the new methodology. It asked the supervisor to proceed “with care” when setting guidance, “so as not to impair lending capacity without need”, and called for the countercyclical buffer and the residential real estate buffer to be withdrawn. Daniel Quinten, a member of the BVR board, called the smaller data request “an important step towards relief” and added that the “collection and implementation burden for institutions” remained “high”. The cliff is therefore this article’s own reading, and the sources reviewed contain no industry position on it. It still matters for capital planning, because a bank whose stressed ratios sit just above the threshold has a reason to hold that margin that has nothing to do with the size of any guidance it might otherwise receive.

A second objection concerns timing. The guidance is not binding, and Section 6d(4) KWG makes clear that an institution that does not fully cover it faces no distribution restrictions as long as it meets its requirements and buffers. Börsen-Zeitung rightly noted that the supervisor nonetheless expects institutions “to comply with the recommendations”. A bank just above the threshold in 2026 and just below it in 2028 gets the full guidance back, calculated with a factor that BaFin will have reset by then. The notice runs until 30 August 2028, after which BaFin expects to turn it into a circular if the practice becomes established. The EBA’s new SREP guidelines, which apply from 1 January 2027, change little about the logic of the guidance.

The relief reaches 247 institutions, its size remains unquantified

In an interview on the stress test that BaFin itself published, the interviewer put the obvious point to Nikolas Speer: “That doesn’t sound like the great big cut in red tape.” Nikolas Speer answered with the reporting burden, saying the greatest relief was a sharp cut in data points, by up to a third for all institutions. That is an honest way to put it. The new methodology saves letters and, for 247 institutions, a recommendation that previously sat in their capital planning. In the sources reviewed, nobody, whether supervisor, association or trade press, has put a euro figure on it, and the size of the guidance that falls away is not published.

At the same event, the supervisors also urged caution. Michael Theurer, the Bundesbank board member responsible for banking supervision, pointed out that institutions plan for return on assets to rise from 0.43 per cent to 0.64 per cent by 2028, while one in five expects its Common Equity Tier 1 ratio to fall. According to the press release, several dozen institutions could no longer meet their capital requirements in the stress scenario, fewer than in 2024 but enough not to read the threshold as a free pass. The new methodology rewards institutions that remain comfortably capitalised after stress, and that is a sensible priority for a supervisor. For 368 institutions it mainly changes the post. For the 498 below the threshold it changes neither the calculation nor the letter.

Recommendations

Glossary