Reader question: If a bank is allowed to use internal models for some risk-weighted assets, what stops an unusually low model estimate from driving regulatory capital far below the amount produced by standardized Basel approaches?
The Basel output floor is a constraint on the total risk-weighted assets that result when banks use internal-model approaches. It compares the bank’s pre-floor RWA with a percentage of the RWA that would result from the relevant standardized approaches.
A useful simplified representation is:
Floor RWA = φ × Standardised RWA,
Final RWA = max(Pre-floor RWA, Floor RWA),
where φ is the applicable output-floor percentage.
The final Basel target is 72.5%. Under the Basel reference transition schedule, the percentage phases in from 50% in 2023 to 72.5% in 2028. The Basel reference percentage for 2026 is 65%. National implementation dates and transitional schedules can differ, so an implementation must use the rule applicable to its jurisdiction rather than assuming the global Basel reference date is automatically the local legal date.
What this page owns — and what it does not
This page owns:
pre-floor/model-eligible RWA + standardized RWA + applicable transition percentage → floor-adjusted RWA.
It does not replace the Basel IRB credit-risk algorithm, which builds credit RWA from PD, LGD and other inputs; FRTB market-risk algorithms; or the Basel leverage ratio, which is a non-risk-based exposure backstop.
This is regulatory-capital mathematics and educational explanation. It does not assess the capital adequacy of any specific bank and is not financial advice.
Why the output floor exists
Internal models can reflect portfolio-specific risk more precisely than one standardized rule, but model choices can also create large variation in RWA across banks.
The output floor limits how far aggregate internal-model RWA can fall below the standardized benchmark. It therefore preserves some benefit from approved internal models while setting a lower bound linked to common standardized methods.
It is best understood as:
model flexibility subject to a standardized minimum output.
Step 1: calculate pre-floor RWA
The bank first calculates the RWA produced under the Basel approaches it is permitted to use before the output-floor constraint.
This can include combinations of:
- standardized credit-risk RWA;
- approved internal-ratings-based credit-risk RWA;
- standardized or approved market-risk approaches;
- operational-risk RWA under the applicable framework;
- counterparty-credit and CVA components;
- other required RWA components.
Call the resulting aggregate:
RWApre.
Step 2: calculate the standardized RWA benchmark
The bank must also construct the RWA that is relevant to the output-floor calculation using the Basel-specified standardized approaches.
Call this:
RWAstd.
This is not necessarily a published accounting number already sitting in one system. For a model-using bank, it can require a parallel standardized calculation across portfolios that are otherwise managed under approved internal approaches.
This creates a major data requirement:
the bank needs enough standardized-approach data even where internal models are used for primary RWA.
Step 3: select the applicable floor percentage
Basel’s reference transitional schedule is:
| Date | Basel reference floor percentage |
|---|---|
| 1 Jan 2023 | 50% |
| 1 Jan 2024 | 55% |
| 1 Jan 2025 | 60% |
| 1 Jan 2026 | 65% |
| 1 Jan 2027 | 70% |
| 1 Jan 2028 | 72.5% |
Therefore, under the Basel reference timetable:
φ2026 = 0.65.
But local law can implement Basel reforms on a different timetable. A production engine must store jurisdiction and effective-date metadata rather than hard-code “current year → Basel percentage” globally.
Step 4: calculate the floor RWA
Suppose standardized RWA is 500.
Under the Basel reference 2026 percentage:
Floor RWA = 0.65 × 500 = 325.
At the final 72.5% target:
Floor RWA = 0.725 × 500 = 362.5.
The same standardized benchmark therefore produces a higher floor as the transition percentage increases.
Step 5: test whether the floor binds
Suppose:
Pre-floor RWA = 300.
Standardised RWA = 500.
At the 2026 Basel reference percentage:
Floor RWA = 325.
Since 325 > 300:
Final RWA = 325.
The output floor binds.
If pre-floor RWA were 350 instead, then:
Final RWA = max(350, 325) = 350.
The floor would not bind.
A binding indicator
A useful machine diagnostic is:
Binding = 1 if φ × RWAstd > RWApre; otherwise 0.
The floor uplift can be reported as:
Floor Uplift = max(φRWAstd − RWApre, 0).
And the relative uplift as:
Floor Uplift % = Floor Uplift / RWApre.
These quantities are useful for explaining why capital ratios move even if model parameters do not change.
Step 6: propagate final RWA into capital ratios
Suppose CET1 capital is 40.
Before the floor:
CET1 ratio = 40 / 300 = 13.33%.
With the 2026 Basel reference floor binding at RWA 325:
CET1 ratio = 40 / 325 ≈ 12.31%.
At the final 72.5% floor, if standardized RWA remains 500 and pre-floor RWA remains 300:
CET1 ratio = 40 / 362.5 ≈ 11.03%.
No capital disappeared. The denominator increased because the regulatory floor became more binding.
The floor does not overwrite internal models
If the floor binds, the internal-model calculations do not become meaningless. They can still determine portfolio-level risk measures, risk management and parts of the pre-floor RWA calculation.
The output floor acts at the required aggregate regulatory level by imposing a lower bound on total RWA.
This distinction matters because a bank may remain model-sensitive even when its reported capital ratio is currently floor-bound.
The floor is not a 72.5% risk weight
A common misunderstanding is to say that the bank’s assets receive a 72.5% risk weight.
They do not.
The 72.5% is applied to the standardized RWA benchmark, not directly to asset exposure amounts.
If an exposure of 100 receives a standardized risk weight of 50%, its contribution to standardized RWA is 50. At the final floor, its indirect benchmark contribution would be part of:
72.5% × standardized aggregate RWA,
subject to the actual Basel aggregation rules.
Why standardized-data quality becomes model-capital quality
A bank can have excellent internal models but poor standardized-approach data mapping.
If the output floor binds, an error in standardized classification can directly change final RWA and capital ratios.
Therefore standardized calculation infrastructure is no longer merely a fallback or disclosure calculation for model banks. It can become the binding denominator owner.
Counterexample: improving an internal model may not improve the reported capital ratio
Suppose pre-floor RWA falls from 300 to 280 after a valid model improvement, while floor RWA remains 325.
Final RWA remains:
max(280, 325) = 325.
The economic/model measure improved, but the reported denominator did not move because the floor was already binding.
This is not a model failure. It is the intended behavior of a binding constraint.
Counterexample: standardized RWA can raise final RWA even when model RWA is unchanged
Suppose model RWA stays at 300 but a standardized exposure mapping correction raises standardized RWA from 500 to 540.
At φ = 65%:
Old floor = 325.
New floor = 351.
Final RWA rises from 325 to 351 despite no change in the internal model.
This demonstrates why standardized data and model data require separate controls.
Counterexample: the output floor and leverage ratio can bind at different times
A bank with low-risk, high-volume assets can be constrained by the non-risk-based leverage ratio even when the output floor does not bind.
A bank with large differences between model-based and standardized RWA can be constrained by the output floor while its leverage ratio remains comfortable.
The two backstops are therefore complementary rather than substitutes.
National implementation can differ from the Basel reference schedule
Basel standards establish an international reference framework, but they become binding through domestic implementation.
Jurisdictions can adopt different start dates, transitional arrangements and legal details. For example, UK and EU implementation timetables have not simply mirrored every Basel reference date.
A current system should therefore resolve:
legal entity → jurisdiction → applicable rule version → applicable φ.
Using 65% merely because the calendar says 2026 would be wrong for a jurisdiction whose local transition has a different effective timetable.
Transitional national-discretion cap
The Basel transitional arrangements permit supervisors, during the phase-in period, to exercise national discretion to cap the incremental increase in RWA caused by the output floor at 25% of a bank’s RWA before application of the floor.
This is not a universal automatic cap. It is a national-discretion transitional option.
An implementation must therefore distinguish:
Basel permission ≠ local activation.
A stylized 25% transitional-cap example
Suppose pre-floor RWA = 300 and an unconstrained floor calculation would raise final RWA to 390.
The increase is 90, or 30% of pre-floor RWA.
If the competent authority has activated the Basel transitional discretion exactly as permitted, the allowed RWA increase could be limited to:
25% × 300 = 75.
A capped transitional RWA would then be 375 under that simplified illustration.
The engine must not apply this cap unless the relevant local rule actually does so.
Inputs and outputs
A robust output-floor engine can require:
- regulatory consolidation perimeter;
- pre-floor RWA by risk type;
- standardized-approach RWA by risk type;
- jurisdiction and legal entity;
- applicable floor percentage;
- transitional effective dates;
- national-discretion cap status where relevant;
- CET1, Tier 1 and total capital;
- rule/version identifiers;
- source-data and mapping lineage.
Outputs can include pre-floor RWA, standardized RWA, floor RWA, final RWA, binding flag, floor uplift, capital ratios before and after the floor, phase-in impact and jurisdictional-rule diagnostics.
Evidence polarity: what supports confidence?
Evidence for a reliable floor calculation includes pre-floor RWA reconciling to approved capital reports, standardized RWA reproducible from line-level exposures, floor percentages tied to the correct jurisdiction/effective date, binding status matching independent calculation, capital ratios reconciling after denominator substitution and transitional discretion documented where used.
Evidence against confidence includes 72.5% applied directly to exposure rather than standardized RWA, a global Basel percentage hard-coded for every jurisdiction, standardized RWA missing modelled portfolios, floor uplift changing after a mapping update with no traceable exposure change, or a national-discretion cap applied automatically without legal basis.
Weak links in implementation
Perimeter mismatch. pre-floor and standardized RWA cover different legal entities.
Standardized shadow-calculation gaps. internal-model portfolios lack required standardized attributes.
Wrong phase-in date. Basel reference schedule is confused with local law.
Percentage/unit error. 65 is multiplied instead of 0.65.
Double-flooring. a floor-adjusted RWA is accidentally used again as the pre-floor input.
Capital-ratio denominator drift. one report uses floor RWA while another uses pre-floor RWA.
National-discretion assumption. the optional 25% transitional cap is treated as mandatory.
Risk-type omission. a component required in the standardized benchmark is missing.
Diagnostics: how to test the algorithm
- non-binding test: set pre-floor RWA above φ × standardized RWA and confirm final RWA remains pre-floor RWA.
- binding test: set pre-floor RWA below the floor and confirm final RWA equals the floor amount.
- equality test: make pre-floor RWA exactly equal to floor RWA and verify no artificial uplift.
- phase-in test: run the same standardized RWA through 50%, 55%, 60%, 65%, 70% and 72.5% Basel reference factors.
- jurisdiction test: assign identical banks to two jurisdictions with different local implementation dates and verify different applicable percentages where legally required.
- capital-ratio test: confirm all risk-based capital ratios use final RWA consistently once the floor binds.
- mapping test: change one standardized exposure class and trace the effect from standardized RWA to floor RWA to capital ratio.
- transitional-cap test: activate the optional cap only under an explicit local-rule flag.
- reconciliation test: pre-floor and standardized calculations reconcile to the same regulatory perimeter.
- parallel implementation test: reproduce the aggregate calculation independently in a spreadsheet.
What would falsify confidence?
Confidence should be withdrawn if final RWA is ever below both pre-floor RWA and the applicable floor amount; if standardized RWA cannot be reconstructed; if the floor percentage lacks a jurisdiction/effective-date source; if capital ratios use inconsistent denominators; or if the same inputs produce different results because processing order changes.
Alternatives and limits
The output floor is not a risk model. It is a regulatory constraint placed around risk models and standardized approaches.
It does not tell the bank which model estimate is economically correct, nor does it eliminate model risk. A model can be poor even when the floor does not bind, and a model can be sound while the floor binds because standardized RWA is materially higher.
Internal model validation, standardized approach controls, leverage ratio, stress testing and supervisory review remain separate layers.
How this connects to the surrounding knowledge estate
Credit-model RWA connects to Basel IRB. Trading-book RWA connects to FRTB. The separate gross-exposure backstop is the Basel leverage-ratio algorithm. The output floor sits above those RWA engines as a final denominator constraint rather than replacing them.
Verification and update triggers
Preserve the Basel and local rule versions, jurisdiction mapping, phase-in schedule, standardized RWA calculation, pre-floor RWA source, transitional-discretion flags and capital figures. Revalidate after Basel amendments, domestic implementation changes, acquisitions, model approvals/withdrawals, standardized-approach remapping, or any unexplained change in floor uplift.
Primary and high-quality references
- Basel Framework, RBC90 — Transitional arrangements, including the Basel output-floor phase-in schedule from 50% to 72.5% and the national-discretion transitional cap.
- Basel Committee on Banking Supervision, Basel III: Finalising post-crisis reforms, December 2017, introducing the 72.5% output floor in the final reform package.
- Bank of England Prudential Regulation Authority, Implementation of the Basel 3.1 standards: near-final policy statement — part 2, 2024, illustrating jurisdiction-specific implementation of the output floor.
- Bank of England Prudential Regulation Authority, Implementation of the Basel 3.1 standards — Output floor, explaining standardized versus modelled RWA construction.
Educational boundary: This article explains the Basel output-floor calculation. It does not assess any bank’s capital adequacy or provide personalized financial advice.