EBA Publishes Consultation Paper on CCF Estimation
On 2 July 2025 the European Banking Authority (“EBA”) published draft guidelines on the methodology for Credit Conversion Factor (“CCF”) estimation (hereafter “Draft GL”) in a public consultation as part of the continuation of the IRB repair program.
The Draft GL builds on previous work done in the form of the PD and LGD estimation guidelines (EBA/GL/2017/16), CDR on IRB assessment methodology (EU nr 2022/439), LGD downturn estimation guidelines (EBA/GL/2019/03), the supervisory handbook on validation (EBA/REP/2023/29), Regulation (EU) No 2024/1623 (“CRR3”), and the Basel 3 framework (BCBS D424) and expands on these in a CFF context with the aim of achieving consistency in definitions and methods for PD, LGD, and CCF estimation with respect to revolving credit facilities for which banks apply IRB-CCF estimates. As such, much of the guidance in the Draft GL formalizes existing regulatory expectations already in place for PD and LGD estimation but introduces a number of modifications and simplifications with respect to CCF estimation reflecting the relatively lower materiality and narrower scope of CFF compared to PD and LGD. Among these simplifications is the introduction of minimum “fixed CCF values” of at least 100% where lack of historical data significantly hinders the reliable estimation of own estimates of CCF and the possibility for banks to apply the same non-defaulted CCF estimates to defaulted exposures thus removing the need for banks to develop and calibrate separate CCF in-default models and estimates.
The Draft GL follows a similar overall structure as EBA’s guidelines on PD and LGD estimation (EBA/GL/2017/16) which serves as the starting point for much of the guidance on CCF estimation. More detailed guidance is provided with respect to the various parts of the estimation phase of CCF models and estimates on the one hand and subsequent application of CCF models and estimates on the other. The guidance on the estimation phase covers model development (“risk differentiation”), calibration of estimates and downturn adjustments (“risk quantification”), and the application of appropriate adjustments and MoC to the CCF estimates. The guidance on the application phase covers among other things how conservatism and human judgement shall be applied in the application of CCF models and how the review of estimates process shall be carried out.
Calculation of Realized CCFs
A key element of the Draft GL concerns the definition and calculation of realized CCFs and the level at which CCF estimation shall be carried out where one credit facility may comprise multiple revolving credit contracts. Here the Draft GL builds on the Basel 3 framework in BCBS D424. Depending on the utilization rate at the reference date, realized CCF shall be calculated using either (i) the so-called undrawn limit factor (“ULF”) approach or (ii) where the utilization rate at the reference date equals or exceeds 100%, the so-called limit factor (“LF”) approach, with a fixed 12-month reference date prior to default providing also a treatment for “fast defaults” where the fixed 12-month reference date prior default is not possible to be applied. To address the extreme values in realized CCF that may result from using the ULF-approach when facilities are close to being fully drawn at the reference date banks are suggested to define a threshold in terms of the utilization rate for the so-called region of instability (“ROI”) and apply the LF-approach also for facilities with utilization rates below but close to 100% in order to quarantine their CCF estimates from the negative effects of the ROI.
Moreover, the Draft GL reiterates the requirement in point (a) of article 182(1b) of CRR3 that extreme values of realized CCF should not be handled by capping observed CCF values but instead banks should identify relevant risk drivers in the model development phase to effectively differentiate these observations and assign them to separate grades or pools. Additional guidance in the Draft GL with respect to the calculation of realized CCF concerns the treatment of multiple defaults on a single facility, incomplete recovery and drawdown processes, interest and fees, and additional drawings after default which seek to achieve alignment with the calculation of realized LGD set out in EBA/GL/2017/16.
Risk Quantification and Calibration Approach
In terms of risk quantification, the Draft GL introduces a more structured approach to the calibration phase of CFF estimation which aligns closely with the steps to be followed for LGD estimation set out in section 6.3 of EBA/GL/2017/16. Among other things, the GL proposes that banks are required to include all relevant default observations in the historical observation period for calibration purposes up until the time of estimation without any unjustified exclusions, and that the possibility of additional drawings on incomplete drawdown observations shall be taken into account in a conservative way in the estimation of LRA-CCF where additional drawings are included by the banks in their CCF estimates. Worth noting here also is that EBA proposes in the Draft GL that the LRA-CCF is calculated as the arithmetic average of realized CCF on all default observations in the historical observation period. This differs from the approach stipulated by the ECB that CCF should be calculated as the arithmetic average of yearly averages of realized CCFs as laid down in paragraph 204(c) of section 7.4 on “Credit risk” in EGIM (February 2024) which the Draft GL explicitly prohibits in paragraph 110. Moreover, no unequal weighting of most recent years’ observations in the data is allowed in the calculation of LRA-CCF.
The downturn framework in the Draft GL closely follows EBA’s guidelines on LGD downturn estimation (EBA/GL/2019/03) with some modifications and simplifications added. Most notably, whereas the impact of a downturn on LGD may be estimated using either a “haircut” and/or “extrapolation” approach, only the latter approach is permitted with respect to downturn CCF estimation without the option of extrapolating at risk driver level, unlike for LGD. Also, banks may apply the same downturn component to their CCF estimates for defaulted exposures as to the CCF estimates for non-defaulted exposures thus removing the need for separate quantification of a downturn component for these estimates.
Model Performance Testing and Representativeness
Further guidance is also laid down proposing that testing of model performance, including out-of-sample and out-of-time testing on a representative testing sample, shall be carried out to ensure that the CCF model’s discriminatory power is satisfactory and that the model does not suffer from overfitting as well as ensuring that exposures assigned to the same grades or pools are sufficiently homogenous within these grades or pools and sufficiently heterogenous across grades or pools. The Draft GL also sets out that additional conservatism in the application of CCF estimates shall be applied for exposures in the application portfolio along the same lines as those applied to PD and LGD estimates at least for exposures with (i) missing information on risk drivers as well as for exposures with (ii) outdated or (iii) missing CCF rating assignments. Further, the Draft GL introduces several simplifications for CCF representativeness testing, where for example the representativeness requirements for the development sample are considered subordinate to the actual model performance and thus relaxed, as well as proposing simplifications for the representativeness dimensions.
Consultation Period and Finalization
The consultation round on the Draft GL will run until 15 October 2025 during which respondents will be able to submit feedback on the Draft GL for the EBA to take into consideration before finalization of the guidelines.
Contact us or visit our Risk & Finance page to learn more.
