An IRB model, or Internal Ratings-Based model, is a credit risk framework that lets banks use their own historical data and statistical models to estimate the probability that borrowers will default, replacing the standardized risk weights set by regulators with institution-specific calculations. Basel.NXT significantly tightens the rules around IRB models by restricting which asset classes can use the approach, setting minimum input floors, and introducing an output floor that limits how much capital relief banks can claim compared to the standardized approach. The sections below walk through how IRB models work, what has changed, and what banks need to do to stay compliant.

How does an IRB model calculate credit risk capital?

An IRB model calculates credit risk capital by estimating three core risk parameters for each exposure: the Probability of Default (PD), the Loss Given Default (LGD), and the Exposure at Default (EAD). These inputs feed into a regulatory formula that produces a Risk-Weighted Asset (RWA) figure, and the bank must hold a minimum percentage of capital against that figure. The logic is straightforward. If a bank can demonstrate through rigorous internal data that a particular segment of borrowers has a low historical default rate, the IRB approach rewards that with a lower capital requirement than a blunt, one-size-fits-all standardized weight would produce. The bank applies its PD estimate to a supervisory formula, combines it with LGD and EAD, and arrives at an expected loss and an unexpected loss component. Regulatory capital is primarily held against unexpected losses, since expected losses are meant to be covered by provisions. The appeal of IRB has always been risk sensitivity. Banks with well-managed, low-risk portfolios can hold less capital than their peers with riskier books, which creates a direct link between good credit risk management and capital efficiency.

What is the difference between Foundation IRB and Advanced IRB?

The difference between Foundation IRB (F-IRB) and Advanced IRB (A-IRB) comes down to how many risk parameters the bank estimates itself. Under F-IRB, banks supply their own PD estimates but use supervisory values set by regulators for LGD and EAD. Under A-IRB, banks estimate all three parameters internally, giving them greater flexibility but also requiring more robust data and model governance. In practice, A-IRB delivers more capital sensitivity because a bank can reflect the specific recovery characteristics of its own collateral and lending structures in its LGD model. A mortgage portfolio backed by high-quality real estate, for example, might attract a lower LGD under A-IRB than the supervisory haircut applied under F-IRB. The trade-off is complexity and regulatory scrutiny. A-IRB models require years of internal default and loss data, independent validation, and ongoing supervisory approval. Regulators have grown increasingly skeptical about the variability of A-IRB outputs across institutions, which is one of the main reasons Basel IV has moved to constrain the approach.

What changes does Basel IV make to the IRB approach?

Basel IV makes three major changes to the IRB approach: it removes A-IRB for certain asset classes entirely, it introduces minimum input floors for PD and LGD, and it tightens the definition of default and loss estimation standards. Together, these changes reduce the degree to which banks can use internal models to minimize capital requirements. Under Basel IV, A-IRB is no longer permitted for exposures to large corporates, banks, and other financial institutions. These asset classes must now use F-IRB or the Standardized Approach. The rationale is that default data for these counterparties is too sparse for banks to build statistically reliable LGD models, so supervisory values are considered more appropriate. Minimum input floors set a lower bound on the parameters banks can use in their models. For example, PD cannot fall below 0.05% for most corporate exposures, and LGD floors apply to both secured and unsecured lending. These floors prevent banks from feeding optimistic assumptions into their models to produce artificially low RWAs. The changes also tighten how banks must define default, how they must treat cured exposures, and how they must account for economic downturn conditions in LGD estimates. The overall direction is clear: more consistency, less model discretion.

What is the output floor and how does it affect IRB banks?

The output floor is a Basel IV rule that prevents a bank’s total RWAs, calculated using internal models, from falling below 72.5% of the RWAs that would result from applying the Standardized Approach to the same portfolio. It acts as a capital backstop, ensuring that internal models cannot produce capital requirements dramatically lower than the standardized baseline. For IRB banks with highly optimized internal models, the output floor can be a significant constraint. A bank whose A-IRB models have historically produced RWAs well below the standardized equivalent may find that the floor effectively forces it to hold more capital than its models alone would require. The floor applies at the consolidated level, which means a bank cannot offset a floored portfolio against an unconstrained one without limit. The practical impact varies by institution and portfolio mix. Banks with large retail mortgage books or well-collateralized SME portfolios may feel the floor more acutely, because those are often the asset classes where internal models have historically delivered the greatest capital savings relative to standardized weights. Banks that are already close to the standardized output will experience less disruption.

Why are regulators restricting IRB model use under Basel IV?

Regulators are restricting IRB model use under Basel IV because studies conducted after the financial crisis revealed excessive variability in RWA outputs across banks using similar portfolios. Two banks holding identical exposures could produce materially different capital requirements simply because of different modeling assumptions, making it difficult for supervisors and investors to compare capital adequacy across institutions.

This variability undermined the credibility of risk-weighted capital ratios as a measure of financial resilience. If internal models can be tuned to produce lower RWAs without a corresponding reduction in actual risk, the capital framework loses its purpose. Regulators found that some of the variability was legitimate, reflecting genuine differences in portfolio quality, but a significant portion appeared to stem from modeling choices rather than underlying risk. The Basel Committee’s response was to reduce model flexibility in areas where data is thin or where variability was hardest to justify. Removing A-IRB for bank and large corporate exposures, setting input floors, and introducing the output floor are all mechanisms designed to narrow the range of permissible outcomes without abandoning the risk-sensitive framework altogether.

How should banks prepare their IRB models for Basel IV compliance?

Banks preparing for Basel IV IRB compliance need to audit their existing models against the new parameter floors, assess which portfolios are affected by the A-IRB restrictions, and run parallel calculations under both internal models and the Standardized Approach to understand where the output floor will bind. The earlier this analysis starts, the more time a bank has to adjust capital planning and data collection strategies. A practical preparation checklist includes:

• Reviewing all PD and LGD model outputs against the new minimum input floors and identifying exposures where current estimates fall below the regulatory thresholds
• Mapping corporate, bank, and financial institution exposures to determine which must migrate from A-IRB to F-IRB or the Standardized Approach
• Running output floor calculations at the consolidated level to quantify the capital impact and identify the portfolios driving the shortfall
• Updating default definitions and loss estimation methodologies to align with the revised Basel IV standards, including downturn LGD requirements
• Stress testing the revised capital position under different economic scenarios to understand how the new framework behaves in adverse conditions

Data quality sits at the heart of all of this. Basel IV’s tighter standards mean that gaps in historical default and loss data will have a direct impact on model eligibility and parameter estimates. Banks that have invested in centralizing and cleaning their credit risk data will be better positioned to demonstrate compliance and defend their model outputs to supervisors. Technology also plays an important role. Running parallel calculations across IRB and standardized approaches, applying output floor logic, and stress testing capital positions across scenarios requires a platform that can handle these calculations at scale without the delays that come with overnight batch processing. At ElysianNxt, our Basel IV credit risk solution covers the full Revised IRB framework, including output floor calculations, input floor application, and integration with ECL models, so you can move from data to decision in real time rather than waiting until the next morning to see your numbers.