Part A defined what counts as a GIRR risk factor. This part supplies the actual numbers: the bucket structure, the risk weight assigned to each tenor, and the full correlation matrix that Article 8A’s aggregation formulas — WSk = RWk × sk, and the within-bucket and across-bucket formulas — use to turn individual sensitivities into a GIRR capital requirement.
Before the specifics, one general principle applies to everything in this article and every risk-class article that follows: the prescribed risk weights and correlations from here through MAR21.89 have all been calibrated to the liquidity-adjusted time horizon relevant to each risk class — the same liquidity horizon concept we defined back in Article 4A’s glossary.
Sourced from MAR21.40.
GIRR Buckets: One Per Currency
The bucket structure for GIRR is the simplest of any risk class in this series: each currency is its own separate delta GIRR bucket. Every risk factor in the risk-free yield curves for a given currency — regardless of which specific curve within that currency it comes from — is grouped into that one currency bucket.
Sourced from MAR21.41.
Risk Weights by Tenor
For calculating weighted sensitivities, each of the ten prescribed GIRR tenors carries its own risk weight.
| Tenor | Risk Weight |
| 0.25 year | 1.7% |
| 0.5 year | 1.7% |
| 1 year | 1.6% |
| 2 year | 1.3% |
| 3 year | 1.2% |
| 5 year | 1.1% |
| 10 year | 1.1% |
| 15 year | 1.1% |
| 20 year | 1.1% |
| 30 year | 1.1% |
Two further flat risk weights apply outside the tenor structure: the inflation risk factor and the cross-currency basis risk factor are each assigned a risk weight of 1.6%.
Sourced from MAR21.42–MAR21.43 (Table 1).
The Square-Root-of-2 Discount for Major Currencies
For a specific list of currencies named by the Basel Committee — EUR, USD, GBP, AUD, JPY, SEK, CAD, plus a bank’s own domestic reporting currency, whatever that happens to be — a bank may, at its own discretion, divide the risk weights above by the square root of 2. This effectively reduces the risk weight by roughly 29%, reflecting the deeper liquidity and more robust hedging markets available in these major currencies compared to the general case.
Sourced from MAR21.44, including footnote [12].
Correlation Within a Bucket: Same Tenor, Different Curve
Where two weighted sensitivities sit in the same bucket (same currency), at the same assigned tenor, but come from different curves — for example, the five-year point on an OIS curve versus the five-year point on a three-month Euribor curve — the correlation parameter ρkl between them is set at 99.90%.
One practical simplification is allowed specifically for cross-currency basis risk: because onshore and offshore curves for the same currency must be treated as two different curves (as established in Article 9A), a bank may choose to aggregate all cross-currency basis risk for a given currency pair — “Curr/USD” or “Curr/EUR” — across both onshore and offshore curves using a simple sum of weighted sensitivities, rather than running the full correlation-based aggregation.
Sourced from MAR21.45.
Correlation Within a Bucket: Different Tenor, Same Curve
This is the densest single table in GIRR: the correlation between weighted sensitivities at different tenors, on the same curve, within the same currency bucket.
| 0.25y | 0.5y | 1y | 2y | 3y | 5y | 10y | 15y | 20y | 30y | |
| 0.25y | 100.0% | 97.0% | 91.4% | 81.1% | 71.9% | 56.6% | 40.0% | 40.0% | 40.0% | 40.0% |
| 0.5y | 97.0% | 100.0% | 97.0% | 91.4% | 86.1% | 76.3% | 56.6% | 41.9% | 40.0% | 40.0% |
| 1y | 91.4% | 97.0% | 100.0% | 97.0% | 94.2% | 88.7% | 76.3% | 65.7% | 56.6% | 41.9% |
| 2y | 81.1% | 91.4% | 97.0% | 100.0% | 98.5% | 95.6% | 88.7% | 82.3% | 76.3% | 65.7% |
| 3y | 71.9% | 86.1% | 94.2% | 98.5% | 100.0% | 98.0% | 93.2% | 88.7% | 84.4% | 76.3% |
| 5y | 56.6% | 76.3% | 88.7% | 95.6% | 98.0% | 100.0% | 97.0% | 94.2% | 91.4% | 86.1% |
| 10y | 40.0% | 56.6% | 76.3% | 88.7% | 93.2% | 97.0% | 100.0% | 98.5% | 97.0% | 94.2% |
| 15y | 40.0% | 41.9% | 65.7% | 82.3% | 88.7% | 94.2% | 98.5% | 100.0% | 99.0% | 97.0% |
| 20y | 40.0% | 40.0% | 56.6% | 76.3% | 84.4% | 91.4% | 97.0% | 99.0% | 100.0% | 98.5% |
| 30y | 40.0% | 40.0% | 41.9% | 65.7% | 76.3% | 86.1% | 94.2% | 97.0% | 98.5% | 100.0% |
This entire table is generated from a single formula, not chosen point by point — worth knowing if you ever need to sanity-check or extend it:
| ρ_kl = max( exp( −θ × |T_k − T_l| ÷ min(T_k, T_l) ), 40% ) where T_k and T_l are the tenors relating to WS_k and WS_l respectively, and θ (theta) is set at 3% |
| The Source Document’s Own Worked Example The correlation between a sensitivity to the one-year tenor and a sensitivity to the five-year tenor of the Eonia swap curve, in the same currency, is calculated as: max(e^(−3% × |1−5| ÷ min(1,5)), 40%) = max(e^(−3% × 4 ÷ 1), 40%) = max(e^(−0.12), 40%) = max(88.69%, 40%) = 88.69% — matching the value shown at the intersection of the 1-year row and 5-year column in the table above. The intuition behind the formula: the further apart two tenors are relative to the shorter of the two, the faster the correlation decays — but it can never fall below the 40% floor, no matter how distant the tenors. |
Sourced from MAR21.46 (Table 2) and footnote [13], including the full worked example.
https://www.bis.org/basel_framework/chapter/MAR/21.htm?inforce=20230101&published=20260323
Correlation Within a Bucket: Different Tenor AND Different Curve
Where two weighted sensitivities sit in the same bucket but differ in both tenor and curve, the correlation is simply the Table 2 value from above, multiplied by 99.90%.
| ρ_kl (different tenor, different curve) = ρ_kl (from Table 2) × 99.90% |
| The Source Document’s Own Worked Example The correlation between a sensitivity to the one-year tenor of the Eonia swap curve and a sensitivity to the five-year tenor of the three-month Euribor swap curve, in the same currency, is (88.69%) × (0.999) = 88.60% — taking the 88.69% figure calculated just above and applying the same-currency, different-curve discount. |
Sourced from MAR21.47, including footnote [14].
Correlation Involving Inflation and Cross-Currency Basis
| Correlation Pair | Value |
| Inflation curve sensitivity ↔ a given tenor of the relevant yield curve | 40% |
| Cross-currency basis curve sensitivity ↔ a given tenor of the relevant yield curve | 0% |
| Cross-currency basis curve sensitivity ↔ the inflation curve | 0% |
| Cross-currency basis curve sensitivity ↔ another cross-currency basis curve (if relevant) | 0% |
Sourced from MAR21.48–MAR21.49.
Correlation Across Buckets (Different Currencies)
For aggregating GIRR risk positions across different buckets — meaning across different currencies entirely — the cross-bucket correlation parameter γbc is set at a flat 50%, regardless of which two currencies are involved.
Sourced from MAR21.50.
https://www.bis.org/basel_framework/chapter/MAR/21.htm?inforce=20230101&published=20260323
Putting It All Together
With Articles 9A and 9B combined, GIRR is now fully specified end to end: what counts as a risk factor, how sensitivities are measured, and every number needed to risk-weight and aggregate them, from the tenor-level risk weights through to the cross-currency correlation. Plugging these numbers into the formulas from Article 8A gives a complete GIRR capital requirement, calculated under each of the three correlation scenarios from Article 8B.
| Building Block | Where It’s Defined | Covered In |
| Risk factor definitions | MAR21.8 | Article 9A |
| Sensitivity formula (PV01) | MAR21.19 | Articles 8C and 9A |
| Bucket structure | MAR21.41 | This article |
| Risk weights by tenor | MAR21.42–21.44 | This article |
| Within-bucket correlations | MAR21.45–21.49 | This article |
| Across-bucket correlation | MAR21.50 | This article |
| Aggregation formulas | MAR21.4–21.7 | Articles 8A and 8B |
Looking Ahead
GIRR is complete. The next article moves to the second risk class: credit spread risk for non-securitisations — the 18-sector bucket structure that captures spread risk on corporate and sovereign bonds outside the securitisation framework.
Frequently Asked Questions
How many GIRR buckets are there?
One per currency. Every risk factor in a currency’s risk-free yield curves — regardless of which specific curve it comes from — is grouped into that single currency bucket.
Which currencies qualify for the square-root-of-2 risk weight discount?
EUR, USD, GBP, AUD, JPY, SEK, CAD, and a bank’s own domestic reporting currency. The discount is optional, applied at the bank’s discretion.
What is the correlation between GIRR sensitivities in different currencies?
A flat 50%, regardless of which two currencies are involved — this is the across-bucket correlation parameter γbc used in the cross-bucket aggregation formula from Article 8A.
Link to previous article