Credit Edge

The CRR3 Output Floor: A Strange Backstop for Credit Risk Measurement

Basel 3.1 has a so-called output floor that places strict limits on the amount of benefits a bank can gain from using internal models to meet supervisory capital requirements. However, this counterintuitive backstop itself suffers from inconsistencies.

Friday, March 22, 2024

By Marco Folpmers


The latest update to the capital requirements regulation under Basel 3.1 reinforces the notion that EU supervisors strongly favor a standardized approach (SA) to credit risk measurement over an advanced internal ratings-based (A-IRB) approach. This is particularly true with respect to one of the most controversial elements of Basel 3.1: the output floor, which is scheduled to go into effect on January 1, 2025.

marco-folpmersMarco Folpmers

The output floor incentivizes the adoption of the SA by banks, partly by restricting the advantages they can gain from applying the A-IRB approach. Consistency in the calculation of risk-weighted assets is the line of reason most cited by prudential authorities (especially the European Banking Authority) when explaining the output floor changes that have been implemented and the reason SA is favored over A-IRB. Today, supervisors argue, there is too much variability in RWA computations.

Banks, supervisors further contend, have in the past used diverse internal models to comply with a wide range of regulatory capital requirement measures for the same hypothetical portfolio. More problematic, however, may be the inconsistencies in Basel’s near-final rules for capital requirements regulation (CRR3), specifically with respect to the formulas for calculating the output floor during a five-year transition period. Let’s now discuss the mechanics of this backstop and why it is such a rough and counterintuitive measure.

Three Regimes – Analyzing the Output Floor

The output floor is the “lock on the door” – or the backstop for the A-IRB approach. It puts a limit on the amount of capital benefit a bank can obtain from its use of internal models relative to the SA. 

The BCBS describes this restriction as follows: “Banks’ calculations of risk-weighted assets generated by internal models cannot, in aggregate, fall below 72.5% of the RWAs computed by the standardized approaches. This limits the benefit a bank can gain from using internal models to 27.5%.” It needs to be applied, moreover, not merely at the group level but at all levels of consolidation.

Consequently, when determining the final total risk exposure amount (image001-2), the standardized total risk exposure amount (image002-1, or the output of applying the SA) is an important factor. The output produced by the internal ratings-based (IRB) approach is the unfloored total risk exposure amount (image003-2); to determine a firm’s final image008-3, this IRB output must be compared to image002-1 , via the following formula: image004-3.

Essentially, this means that the image008-3 cannot go below 72.5% of the standardized image008-3. However, this is not a straightforward rule change. The output floor will be a phased implementation, and the 72.5% figure will only be reached after a transition period that runs from January 2025 through December 2029.

Specifically, the output floor will gradually increase from 50% in January 2025 to 70% in January 2029; it will not reach its final value of 72.5% until January 1, 2030.

During the transition period, the formula for calculating the output floor will be modified as follows: image005.

Under this formula, the output floor will be limited to 25% of the unfloored image008-2. In the graph below, we show how the transition will work in 2029, when image006-1.


In the figure, we have added the 45° line in light green, as well as its multiple (1.25) in gray. Moving along the x-axis (from the right to the left), we can see that there are three different regimesor possible areas for the unfloored 𝑇𝑅𝐸𝐴 as a percentage of the capital requirement as computed under the standardized approach:

  • The first regime shows the outcome when the unfloored image008-2 is equal to or above 70% of standardized image008-3. In this case, the bank is free to apply the IRB approach, and the ultimate image008-3 is equal to the output of the IRB approach. This regime is “IRB risk sensitive” in that more risk means higher capital requirements (as measured by internal models) – and vice versa.
  • The second regime depicts a situation where the unfloored image008-3 is below 70%, but equal to or above 56% of the standardized image008-3. In this case, the IRB approach is not effective. The final capital requirement is therefore equal to the outcome of the SA; PDs and LGDs do not matter in this example, and the regime is not “IRB risk sensitive.”
  • The third regime shows what could happen when the unfloored image008-3 is below 56% of the standardized image008-3. In this case, the final image008-3 is equal to the gray line – i.e., it’s 1.25 times the unfloored image008-3. This regime is again “IRB risk sensitive.” In terms of absolute distance in percentage points, the gap between image008-3 and image003-2 (the vertical distance between the gray line and the light green line) becomes smaller as image003-2 decreases.

The outcomes of all three scenarios are baffling. Firstly, both the 72.5% output floor and the transition path toward it (captured in variable image010-1) seem quite arbitrary. Secondly, once we accept that the advantage of the IRB approach is limited to capital requirements above image011-2, it is strange that trust is suddenly restored for low, unfloored image008-2.

Indeed, for image006-1, trust is restored for unfloored image008-2 below 56%. I don’t think that there is any supervisor or financial risk manager who can explain why, for low, unfloored image008-2 (from 56% and below), the output floor is again IRB sensitive – even with a diminishing gap between floored and unfloored image008-2.

Parting Thoughts

The near-final text of CRR3 contains baffling specifications of the output floor. This leads us to ask the following unresolved questions: (1) How was the 72.5% value of the floor calibrated, if it was indeed statistically calibrated rather than politically motivated? (2) Why should banks stop fully relying on the IRB capital requirement below this exact percentage? and (3) Why is confidence in the IRB capital requirement restored as soon as the gap between floored and unfloored image008-2 is larger than 25%?

Supervisors and legislators claim to want to reduce variability in banks’ RWA calculation – but have yet to clarify the logic behind the output floor. Banks that are forced to comply with the Basel 3.1 rules for capital requirements are at least owed a detailed explanation about why the output floor modifications, and the preference toward standardized approach over internal models, make sense.

Dr. Marco Folpmers (FRM) is a partner for Financial Risk Management at Deloitte the Netherlands.


BylawsCode of ConductPrivacy NoticeTerms of Use © 2024 Global Association of Risk Professionals