CRR Model

Approval for the Internal Ratings-Based Approach (IRBA) requires a formal application to the competent authority pursuant to Art. 143 CRR. Institutions must demonstrate that their internal rating systems fully meet the requirements of Art. 143 to 191 CRR. This includes robust PD (Probability of Default) estimates for the Foundation IRB approach, plus own LGD and EAD estimates for the Advanced IRB approach.

Key requirements include at least three years of system usage before application, complete documentation of model methodology, independent validation processes, and adequate data quality. The supervisor assesses whether all minimum requirements for data history, discriminatory power, calibration, and governance are met. Material model changes or scope extensions also require separate approval.

ADVISORI supports credit institutions through the entire approval process — from model development and documentation to supervisory dialogue.

Under the Foundation IRB approach (F-IRBA), institutions estimate only the Probability of Default (PD) using their own models. For Loss Given Default (LGD), Exposure at Default (EAD), and the Credit Conversion Factor (CCF), supervisory standard values apply.

Under the Advanced IRB approach (A-IRBA), institutions additionally estimate LGD, EAD, and CCF with their own internal models. This requires more extensive data histories, more rigorous validation processes, and stronger model governance. Under CRR III, the scope of A-IRBA is restricted: for exposures to financial institutions and large corporates with consolidated annual revenue exceeding EUR 500 million, only F-IRBA is permitted. Additionally, input floors apply for PD (minimum 0.05%), LGD, and CCF, preventing model estimates from producing excessively low values.

The choice of approach directly impacts the level of risk-weighted assets (RWA) and consequently capital requirements.

The output floor under Art. 92a CRR III limits the capital benefit that banks can achieve through internal models compared to the Standardised Approach (SA). From 2025, risk-weighted assets (RWA) calculated using internal models may not fall below a specified percentage of SA-calculated RWA.

Implementation is phased: 50% from January 2025, increasing annually by 5 percentage points to 72.5% by 2030. This means IRBA banks must maintain a full parallel SA calculation alongside their internal models. Institutions whose IRB models previously achieved significant capital advantages over the standardised approach must expect higher own funds requirements.

Banks should conduct portfolio-specific impact analyses early to assess which exposure classes remain economically viable under IRBA and where switching to the SA may be more advantageous.

With the implementation of the Fundamental Review of the Trading Book (FRTB) through CRR III, requirements for internal market risk models (Internal Model Approach, IMA) are fundamentally revised. The previous Value-at-Risk approach is replaced by Expected Shortfall as the central risk measure.

Institutions wishing to use IMA must apply for approval at individual trading desk level. Each desk must pass a Profit-and-Loss Attribution Test (PLA) and a backtesting test. Desks failing these tests automatically fall back to the Standardised Approach (SA-TB). Capital requirements under FRTB-IMA comprise Expected Shortfall, a Stressed Expected Shortfall component, and a Default Risk Charge.

The approval threshold has risen significantly, leading many institutions to evaluate whether switching to the standardised approach is more economical for certain desks. ADVISORI supports desk-level analysis and strategic FRTB implementation.

Validation of PD, LGD, and EAD models under CRR III follows a structured framework based on Art. 174 to 191 CRR and EBA guidelines on model validation. Independent validation must occur at least annually and cover all model components.

For PD models, validation includes discriminatory power analysis (Gini coefficient, Accuracy Ratio), calibration tests, and stability analysis across different time periods. LGD models are tested by comparing realised loss rates against estimated values, validating both downturn LGD and long-term averages. EAD models and CCF estimates are verified by comparing actual drawdown at default with model predictions.

Under CRR III, additional input floors apply (PD minimum 0.05%, LGD minimum 25% for unsecured corporate exposures), ensuring model estimates do not fall below regulatory minimums. Validation must document correct application of these floors.

CRR III fundamentally changes the cost-benefit calculation for internal models. The output floor (72.5% of SA-RWA) caps the maximum capital advantage of IRBA over the standardised approach at 27.5%. Meanwhile, costs for model development, validation, governance, and IT infrastructure remain.

Institutions must assess on a portfolio-by-portfolio basis whether the remaining capital benefit justifies model maintenance costs. Analyses show that particularly for mid-sized banks with total assets between EUR 5 and 30 billion, selective IRBA application to specific exposure classes may be more economical than blanket usage.

Assessment factors include: current difference between IRB and SA RWA per portfolio, costs of parallel SA calculation (output floor requirement), regulatory change frequency, and the strategic value of granular risk data for internal management. ADVISORI conducts detailed economic analyses and develops optimal model strategies under CRR III.

ADVISORI supports credit institutions across all aspects of CRR model implementation and optimisation. Our services encompass strategic model planning, quantitative model development, preparation for supervisory approval, and ongoing model maintenance.

For IRB approval, we support development and calibration of PD, LGD, and EAD models, preparation of audit-proof documentation, and dialogue with supervisory authorities. For the output floor, we conduct portfolio-specific impact analyses and develop optimal strategies combining IRBA and SA.

For FRTB implementation, we analyse IMA economics per trading desk and support implementation of PLA test and backtesting infrastructure. Our model validation teams perform independent reviews per EBA guidelines and identify optimisation potential. Through regular regulatory updates, we keep our clients’ models aligned with current requirements.