On 11 June 2026 the European Banking Authority (EBA) published the draft methodology for the 2027 EU-wide stress test, together with templates and template guidance. The consultation ran until 10 July 2026 at 23:59 and has now been closed for almost two months. The final methodology has yet to appear. Between those two points lies the window in which a participating bank has to prepare without knowing the final text.
The figure most often quoted from the package promises relief: 55% fewer data points. It comes from the authority’s own impact assessment, it is well documented, and it is still the reason some plans will go wrong this year. Read it as a promise for 2027 and you have confused the volume of data requested with the route by which it travels. The shape of the forms changes at once; the data path changes only in 2029. One full exercise sits in between, and a bank will still have to feed it from its own ad hoc collections.
What: draft methodology, templates and template guidance for the 2027 EU-wide stress test, published on 11 June 2026
Sample: 63 banks from the EU and Norway, 47 of them in the euro area
Consultation: closed on 10 July 2026 at 23:59, with late responses expressly disregarded
Cut-off of the exercise: 31 December 2026, with projections running from 2027 to 2029
Still open: a date for the final methodology, a start date for the exercise and a date for the results. The authority gives none of them
The 55% measures the volume of data, not the route it travels
The figure comes from the impact assessment accompanying the consultation, where it is carefully bounded. The measures could “lead to an overall reduction in stress test data requirements of around 55% compared to the 2025 exercise”. The reference point is the 2025 exercise, not an average and not the reporting framework as a whole.
That distinction matters more than it may appear. The same round of reforms produces a second figure that is often quoted alongside it: for EU harmonised reporting as a whole, the authority has announced a roughly 50% cut in data points of around 50%. Different measure, different scope. Anyone putting both in a slide deck should say which is which.
No absolute figures appear anywhere, for 2025 or for 2027. If you come across one, ask where it comes from.
The relief is real, but it arrives later
The real lever behind the simplification is the link to routine supervisory reporting. Selected stress test data requirements, above all the credit risk starting points, are to move into regular reporting, specifically FINREP, COREP and ESG reporting, so that future exercises rely less on ad hoc collections.
The next sentence in that same source qualifies it:
In the original: “This proposal would apply for the 2029 stress test and onwards (as FINREP/COREP amendments are planned to have as first reference as of September 2027). Still for the 2027 stress test, the stress test templates will mimic the FINREP/COREP/ESG reporting proposals included in the amended ITS on supervisory reporting.”
In practice, banks will still deliver through separate ad hoc templates in 2027. Those templates are modelled on the future reporting forms, which does ease the work of completing them and prepares the later migration. The data path, however, stays separate. Only from the 2029 exercise onwards is the supervisor meant to draw starting points out of routine reporting rather than requesting them.
For resource planning, this is the distinction that matters most. Treat the 55% as grounds for a smaller stress test team in 2027 and you have banked a saving that this round only partly delivers. Do nothing at all and you overlook the fact that the new reporting templates are coming anyway: the FINREP and COREP amendments are due to apply with a first reference date of September 2027. Under the amended implementing technical standards they include the new F 49.01 and F 49.02 templates in FINREP and C 09.05 in COREP for exposure amounts under the internal ratings-based (IRB) approach.
The date that is easy to miss
Two dates sit close together in the methodology, and they do different jobs.
The reference to end 2025 relates only to the sample. The exercise runs on a sample of banks covering “broadly 75% of the banking sector in the euro area, each non-euro area EU Member State and Norway, as expressed in terms of total consolidated assets as of end 2025”. That wording matters, because it is more precise than the widespread shorthand of 75% of the EU banking sector. It means roughly 75% per jurisdiction, not aggregated across the Union.
The exercise's own cut-off date is 31 December 2026. It fixes the starting balance sheet, the applicable accounting regime and the recognition of internal models. The methodology leaves no discretion here: all balance sheet and profit and loss projections across 2027 to 2029 are to be computed under the accounting regime in force at that date, and new or amended internal models must be used where they have been approved and effectively implemented by then.
The first date that binds a bank is therefore not the launch of the exercise but a date in the current year. Anyone hoping to carry a model approval into the exercise has until the end of December to get it done.
The climate module adds work rather than replacing it
Climate risk enters the exercise for the first time, and the authority itself tempers expectations: the risks are “assessed through a dedicated module and will not affect the core stress test results”. The module leaves the capital outcome untouched. It still creates work.
It consists of two separate scenarios. Transition risk runs across the full 2027 to 2029 horizon and assumes an abrupt, stringent shift in climate policy in an unprepared, fossil-fuel-dependent world. Each bank chooses between two sets of shocks: climate variables such as carbon price, emission and energy price paths, or shocks to gross value added by country and economic sector.
Physical risk bites in the first projection year only. It models simultaneous riverine flooding across the member states of the European Economic Area at a a 1% annual probability, drawn from the Joint Research Centre flood hazard maps at 90-metre resolution.
Two constraints shape the workload more than the scenario does. The static balance sheet assumption applies here as well, and mitigation is expressly ruled out: neither retrofits nor the repair of damaged property may be recognised within the horizon, and no management action aimed at reducing climate-related losses may be assumed. Second, the module covers only on-balance-sheet credit risk exposures held at amortised cost and looks solely at the profit and loss effect, leaving risk exposure amounts aside. Market risk is out of scope.
Reporting runs through two new templates for transition and physical risk, built along the lines of the new FINREP template F 49. What banks tend to underestimate is granularity, not volume: sector classification by economic activity, energy performance ratings on residential property and flood severity per exposure are attributes many institutions do not hold in full.
The timetable is missing, the cut-off is not
The authority deliberately consulted early to make preparation easier. A timetable was not part of it. The documents give no date for the final methodology, none for the launch of the exercise and none for publication of results. The widespread expectation of an early 2027 start comes from consultancies rather than from the authority. The industry workshops, too, exist so far only as an announcement, with no schedule attached.
That is a fact of planning, not a complaint. It means the preparation window hangs on a date that is fixed, the cut-off on 31 December 2026, rather than on one that has yet to be set.
London cuts the number of exercises, Brussels the volume of data
Look beyond the cut-off and the same intention turns up in London, executed differently. Both of Europe’s large supervisors are working to reduce the burden that stress testing places on firms, and they pull different levers to do it. The Bank of England published its approach on 29 November 2024, effective from 2025: the annual cyclical scenario, run every year since 2015, becomes a “Bank Capital Stress Test”, which the Bank says it expects to run “every other year”.
It justifies this in the same document as a “material efficiency gain” and as a way of keeping the burden on participating firms proportionate. In the intervening years the supervisor does the arithmetic itself, in desk-based tests built on its own estimates.
What London does not touch is the route. Participating firms still report through a dedicated infrastructure, the Stress Test Data Framework. Submissions go through the BEEDS platform as Excel templates, with its own manual and data dictionary.
Brussels does the opposite: the European Banking Authority keeps the cyclical exercise and cuts the data burden instead, by drawing the starting points out of routine reporting in the medium term.
Two supervisors, one burden, different levers: London cuts the number of exercises and still collects on its own infrastructure; Brussels cuts the volume of data per exercise and intends, eventually, to stop collecting separately at all. Where they end up is worth noting: the EU exercises also run on a two-year cycle, in 2025, 2027 and 2029. For a group authorised on both sides of the Channel, the two easings do not add up. They arrive twice, from different directions, and each lands on a different team.
How long the Brussels approach can take is visible in an earlier version of the same promise. AnaCredit was adopted in 2016 to collect granular credit data a single time and reuse it; the first reporting reference date was 30 September 2018.
Eight years after the regulation, the European Central Bank’s 2024 overview of its Integrated Reporting Framework still records “possible duplications and overlaps in reporting”, and holds out the prospect that ad hoc requests will become less frequent only under the new framework.
That framework is set to replace the AnaCredit regulation, with the first regular reporting due in 2031. This is not a verdict on AnaCredit, and the ECB does not offer one. It is the observation that the same promise has been made before, and that making good on it will have taken fifteen years. For planning purposes that does not mean ignoring 2029. It means treating it as a commitment rather than as a date.
Recommendations
Now: new or amended internal models enter the exercise on a mandatory basis where they are approved and effectively implemented by the cut-off. That is a hard gate, not a formality: whatever is not in place by then no longer shapes the outcome. Line up model validation, supervisory approval and stress test planning this autumn, not in January.
In the budget round: the relief works through the shape of the templates in 2027, not through the data path. Delivery remains separate. The saving that would justify a smaller plan arrives with the 2029 exercise. Scaling down the 2027 plan means planning against what the methodology actually says.
This autumn: F 49.01, F 49.02 and C 09.05 arrive independently of the stress test, with a first reference date of September 2027. Because the 2027 stress test templates are modelled on precisely those templates, running the two initiatives separately means doing the work twice. One data model for credit risk starting points serves both.
Before year end: the supervisor supplies the scenario. What a bank has to supply is sector classification by economic activity, energy performance ratings on residential mortgage collateral and the geographic detail behind the flood shock. A gap analysis on those three attributes is the cheapest step with the largest effect, and it holds whatever the final methodology turns out to say.
Glossary
Data point: the unit in which supervisory reporting is counted, roughly one cell in one template for one reporting date. It is what the 55% refers to, which is why the figure says nothing about how much work each cell costs to produce.
FINREP and COREP: the two sets of reporting templates used across EU banking supervision, one for accounting data and one for own funds and exposures. They are submitted regardless, which is precisely why the stress test starting points are meant to be drawn from them rather than collected separately.
Implementing technical standards (ITS): the binding rules that fix what supervisory reporting looks like, down to the individual template. When the ITS change, the templates change, and everything modelled on them has to follow.
Static balance sheet assumption: the requirement to hold the balance sheet constant across the projection horizon. It bars a bank from reacting to the shock inside the model, and it is the reason the climate module may not credit any retrofits.
BEEDS: the Bank of England’s submission platform for regulatory data, and the route by which UK firms file their stress test returns. It is the counterpart to the separate collection the EBA now wants to retire.