IRB versus the standardised approach

Two ways to weight the same exposure, and a rule that limits how much the difference can be worth.

The difference in one line

The standardised approach asks the supervisor for a weight. The IRB approach asks the bank's model for a probability of default and a loss given default, then puts those through a formula. Everything else follows from that.

Side by side

FeatureStandardisedIRB
Risk weight sourcePrescribed by class and ratingFormula using internal PD and LGD
Data requirementExposure and ratingDefault history, recovery data, validation
MaturityNot an inputInput for corporates
CollateralAffects the weight through prescribed rulesAffects LGD
Supervisory approvalNot requiredModel permission required
Sensitivity to riskCoarseHigh

Same exposure, two answers

A corporate borrower with no external rating, exposure 10,000,000, PD 1%, LGD 40%, maturity 2.5 years.

  • Standardised, CRR: risk weight 100%, RWA 10,000,000, capital 800,000.
  • IRB, CRR3: risk weight 82.06%, RWA 8,205,940, capital 656,475.

The IRB approach saves about 18% of the capital for this borrower. For a low-risk corporate with PD 0.1%, the gap is far larger. For a weak borrower with PD 10%, IRB can be more expensive than the standardised weight, which is the point: the model is supposed to discriminate.

Where IRB pays and where it does not

  • Pays: large, diversified, low-default portfolios such as prime mortgages, high-quality corporates and card portfolios with long histories.
  • Does not pay: small portfolios where data is thin, high-default segments, and exposures that the standardised approach already weights cheaply.
  • Costs: model development, validation, supervisory approval and ongoing monitoring. A bank needs scale before the capital saving covers the running cost.

The output floor

Since 2025 the benefit of IRB is capped. Total risk-weighted assets cannot fall below 50% of the standardised result, rising to 72.5% by 2030. A bank whose models produce a very low number still has to hold capital against the floor. The comparison above is a single exposure; the floor works on the whole institution.

Which one should you use

For a single-exposure estimate, use whichever approach the bank actually has permission for. Mixing them in one number produces a figure that no supervisor would recognise. If the question is academic, calculate both and read the difference as the value of the model.

Questions

Can a bank use both approaches?

Yes, with supervisory permission. The permanent partial use rules decide which portfolios stay standardised.

Is IRB always cheaper?

No. For high-default exposures the modelled weight can exceed the standardised one.

Does the standardised approach use maturity?

No. Maturity is an IRB input for corporate exposures.

Related