The Future of IRB CCF Models: Understanding the Impact of the 2025 Draft Guidelines. Part 2

In July 2025, the European Banking Authority (EBA) published the draft Guidelines on Credit Conversion Factor (CCF) estimation and application methodologies (hereafter, draft CCF GLs) for public consultation, which is scheduled to conclude on 29 October 2025. These draft CCF GLs represent a significant milestone in the history of CCF models, as they mark the first comprehensive formalization by the EBA of regulatory expectations and interpretations under the CRR framework. While the document is still in the draft form, it provides valuable insight into the regulatory vision of the future of the CCF models, enabling industry experts to assess potential impacts and prepare for the coming changes.

This blogpost is to analyze the implications of selected changes introduced in the draft CCF GLs. Specifically, it will assess the direction of the expected impact of these changes on existing CCF models and respective RWA (risk-weighted assets) impacts, and evaluate the operational and methodological adjustments likely required from banks’ modelling teams to align their current IRB (Internal Rating Based) models to the new regulatory environment.

The analysis provided focuses specifically on aspects most relevant to the model development process for retail portfolios.

The CCF is one of the three parameters making up the A-IRB models. Its role is to calculate a bank's exposure to off-balance-sheet items. In essence, a CCF is a percentage used to estimate the portion of an undrawn credit commitment that a borrower is likely to draw down between the estimation date and the time of default (ToD).

As per Regulation (EU) 2024/1623 (known as CRR3), the CCF should be defined as:

where  is the amount drawn at ToD,  - amount drawn at the time of the estimation (reference date   and  - undrawn limit at the time of the estimation.

The introduction of incomplete recoveries as part of the CCF in-default concept is expected to increase RWA compared to the current approach, where banks without in-default CCF models assigned a value of zero to defaulted exposures, assuming that no further changes in exposure are foreseen. However, this increase is expected to remain limited, as the proportion of defaulted exposures in the application portfolio is relatively small compared to banks’ overall portfolios. Furthermore, additional drawings after default are not expected to be common in the retail segment, as banks are generally expected to restrict such drawings through their internal lending standards.

Additional drawings after default (DaD) are the funds that a borrower withdraws from an available credit line or facility after a formal default event has occurred. For example, in case of a credit card exposure, the borrower might have defaulted on a consumer loan but still has the capacity to withdraw money on their credit card and, hence, has additional post-default drawings.

Article 182(1)(c) of the CRR3 stipulates that CCF estimates must include additional DaD. However, Article 182(3) introduces an amendment for retail exposures, allowing institutions to allocate post-default drawings either to the CCF or to the LGD estimates.

In the presence of post-default drawings, the CCF formula can be rewritten as:

where DaD – are the drawings after default. Traditionally, this DaD is measured by adding all additional drawings to the EAD after discounting:

In a “going concern” obligor does large post default drawings, such formula can lead to very high additional drawn amounts and thus high observed CCFs and low LGDs. This effect will be compounded in case of very granular data on drawings and recoveries, i.e. when a bank observes all the post default drawings and reimbursements and not only the monthly difference in the balance amount.

At the level of the measured loss, these two effects will cancel out. However, in case a model does not perform perfectly (which is expected), the processing of additional drawings can bias the RWA results. Such potential biases are amplified by the floors for the parameters’ quantification: for some retail products we have difference between the floor for LGDs and for CCFs, while the integration of the post default drawings can be incorporated in either parameter at the discretion of the modeler.

This difference in floor value can lead to an “arbitrage” between CCF and LGD definitions, resulting in a difference in RWA. To illustrate this point, let’s consider the following simple illustrative example of a small business owner classified as retail and having a small business line of credit. After the loan has defaulted, the line of credit remains classified as in default, although the bank does not prevent the obligor from continuing to draw additional funds.

Let’s suppose that 12 months before the default the drawn amount is €1K and the undrawn amount €4K. The EAD is €3K, the additional drawn amount (calculated as the sum of the discounted post default drawings) is €4K and recoveries (including the fees capitalized after default) amount to €8K. We get the following parameters for the two possible methodologies:

By applying the 15% LGD floor for other retail exposures secured by other physical collateral, we get the different levels of expected losses when including the post default drawings into the LGD or into the CCF. Such floor is in practice applied at the LGD estimate level (i.e., calibration level), but here in our simplifying example there is only one exposure so we apply it at the exposure level. Continuing with our illustrative example, we can see that changing the post-default drawing amount will change the difference in measured loss between including it into the CCF and into the LGD. This loss difference relation is non-linear and becomes 0 as the LGD without including the post default drawings exceed the 15% threshold:

To avoid very high realized CCFs and very low realized LGDs, the draft CCF Guidelines propose to calcule the additional drawings as the difference between the exposure at default and the highest discounted drawn amount observed during the default period.

There are two major differences in this new formula. The first one is that the full drawn amount is discounted at the considered date instead of discounting each drawing at the date of its occurrence. The second primary difference is that the formula implicitly accounts for recoveries as the total drawn amount is affected by both drawings and repayments.

This new formula introduces a discrepancy compared to the existing methodology for incorporating additional drawings (for retail exposures) into the LGD, which is currently done as a sum of discounted cash flows. This inconsistency may lead to 'cherry-picking', i.e., selecting the calculation method that results in the lowest recognized losses.

This new definition of the additional drawn amount has the potential to drastically change the measured CCFs when the post-default drawings are an important component of the CCF. To illustrate this point, we will take a simple example.

Let’s suppose again that 12 months before the default the drawn amount is €1K and the undrawn amount €4K. The amount drawn at default is €3K, the discount rate 5% and the post default drawings (positive amounts) and recoveries (negative amounts) are the following stream of cash flows:

With the former formula (discounted sum of post default drawings), the additional drawn amount is €3,945.5. With the new formula, it becomes €2,389.3. We get the following parameters for the two formulae:

The new additional drawn amount formula, by reducing the value of the amount added to the drawn amount, decreases the recorded loss. For retail exposures, this creates an opportunity to select the lowest recorded loss by choosing between incorporating the post default drawings into the CCF or into the LGD:

The draft CCF GLs require that, consistent with the LGD framework for defaulted exposures, institutions implement a modelling approach where reference dates (potentially multiple) are determined based on observed post-default drawing patterns. For CCF in-default, the use of a fixed 12-month horizon is not permissible.

Institutions must also define a maximum drawing period, representing the time in default after which no further drawings are assumed possible. This period is expected to be aligned with the MRP (maximum recovery period) applied for LGD.

Consider an illustrative example with three facilities:

• Facility A: Credit Limit = 10.000, EAD = 8.000, max(t) = 18
• Facility B: Credit Limit = 5.000, EAD = 4.500, max(t) = 30
• Facility C: Credit Limit = 20.000, EAD = 18.000, max(t) = 24

All three facilities exhibit additional DaD, as shown in Figure 2.

Assume that based on the institution’s internal analysis of contractual practices defining DaD in the retail segment, it was observed that the drawing behavior changes materially after 6 months in default. Consequently, reference dates at 1 month and 6 months are defined for the purpose of CCF-in-default estimation, in line with the regulatory requirement to select reference dates based on observed post-default drawing patterns. Additionally, we assume that at portfolio level, no drawings or recoveries occur beyond 24 months in default. The institution therefore sets the maximum drawing/recovery period (MRP) at 24 months, aligned with both the EBA GLs and the maximum recovery period as defined in the LGD model.

The LRA CCF in-default will be calculated for each of the selected reference dates separately (LRA CC₁,  LRA CCF₆ and LRA CCF₂₄) and then applied to the facilities that are in default at least as long as the selected reference date (Table 3):

Table 3: Example of application of the LRA CCFs by reference date

For each of the reference dates (R_k), the respective LRA CCF in-default is calculated as simple default-weighted average as follows:

where  is the number of facilities that are in default at time  is realized CCF of facility  at time  that is computed in accordance with the defined draft CCF GLs formula for CCF DaD as below:

The table below summarizes the key focus areas of the draft CCF GLs outlined above and their expected impact on RWA (graphically in the respective column below). The requirements with “high” (red arrows) RWA impact are expected to cause the most noticeable increase in the RWA values. The implementation of the “medium” impact points (orange arrows) is expected to result in limited RWA increase. The remaining requirements are not associated with a predetermined increase or decrease in RWA, as their impact may vary depending on the characteristics of each individual portfolio.

In addition, the summary table provides an indicative assessment of the implementation effort required to adapt the existing models in the European banks (based on common market practices) to the draft requirements, should they come into force in the current form. In this context, low effort refers to adjustments such as sample filtering or model recalibration; medium effort corresponds to changes in risk differentiation or the reconstruction of risk driver functions; and high effort involves fundamental changes to model design and a full redevelopment.