How does ISDA SIMM calculate initial margin, and what are the risk sensitivity buckets?

ISDA SIMM (Standard Initial Margin Model) computes initial margin for non-centrally cleared derivatives using a standardized sensitivity-based approach. It organizes risk factors into classes, risk types, and buckets, then aggregates using prescribed correlations to produce a portfolio-level IM.\n\nSIMM Hierarchy:\n\n`mermaid\ngraph TD\n A[\"SIMM Total IM\"] --> B[\"Interest Rate\"]\n A --> C[\"Credit Qualifying\"]\n A --> D[\"Credit Non-Qualifying\"]\n A --> E[\"Equity\"]\n A --> F[\"Commodity\"]\n A --> G[\"FX\"]\n B --> H[\"Delta\"] \n B --> I[\"Vega\"]\n B --> J[\"Curvature\"]\n H --> K[\"Bucket 1: USD
Bucket 2: EUR
Bucket 3: GBP
...\"]\n K --> L[\"Tenor sensitivities:
2W, 1M, 3M, 6M,
1Y, 2Y, 3Y, 5Y,
10Y, 15Y, 20Y, 30Y\"]\n`\n\nCalculation Steps:\n\nStep 1 -- Compute sensitivities. For each trade, calculate delta (PV change per 1 bp rate shift), vega (PV change per 1% vol shift), and curvature (second-order PV change for large rate moves).\n\nStep 2 -- Assign to buckets. Group sensitivities by currency (IR), sector/rating (credit), industry (equity), or commodity type.\n\nStep 3 -- Weight sensitivities. Multiply each raw sensitivity by its prescribed risk weight (RW). For example, USD 10Y IR delta has a risk weight of approximately 61 bps.\n\nWeighted sensitivity: WS = RW x s\n\nStep 4 -- Aggregate within bucket. Using intra-bucket correlations (rho):\n\nK_b = sqrt(Sum_i Sum_j rho_{ij} x WS_i x WS_j)\n\nStep 5 -- Aggregate across buckets. Using inter-bucket correlations (gamma):\n\nTotal = sqrt(Sum_b Sum_c gamma_{bc} x S_b x S_c)\n\nwhere S_b is the net bucket-level contribution.\n\nWorked Example:\nVanguard Derivatives has a bilateral portfolio with Clearwater Bank. IR delta sensitivities for the USD bucket:\n\n| Tenor | Raw Sensitivity (per bp) | Risk Weight | Weighted Sensitivity |\n|---|---|---|---|\n| 2Y | +$42,000 | 56 bps | +$2,352,000 |\n| 5Y | -$28,000 | 52 bps | -$1,456,000 |\n| 10Y | +$61,000 | 61 bps | +$3,721,000 |\n\nIntra-bucket correlation between 2Y and 10Y: 0.71\nIntra-bucket correlation between 2Y and 5Y: 0.85\nIntra-bucket correlation between 5Y and 10Y: 0.83\n\nK_USD = sqrt(2,352^2 + (-1,456)^2 + 3,721^2 + 2 x 0.85 x 2,352 x (-1,456) + 2 x 0.71 x 2,352 x 3,721 + 2 x 0.83 x (-1,456) x 3,721) (in thousands)\n\nAfter computation: K_USD = approximately $4,580,000\n\nThis USD bucket result then aggregates with EUR, GBP, and other currency buckets using inter-bucket correlations.\n\nKey Design Features:\n- Risk weights are calibrated to stressed market conditions (not average vol)\n- Correlations are fixed by ISDA governance, updated annually\n- Concentration thresholds increase margin for large positions\n- The model is transparent and replicable (unlike internal models)\n\nStudy SIMM methodology in our FRM Part II course.