Elderlake Partners

04 · A-SA — sensitivities-based method

Art 325c–325ay CRR

Purpose & history

Capital from the desk's own greeks

The old standardised approach priced positions by notional buckets — blind to hedges, blind to curvature. The A-SA's insight is to charge capital on the same objects the front office already manages: delta sensitivities to prescribed risk factors, vega where there is optionality, and a curvature add-on that stress-shifts each factor up and down and keeps the worse residual beyond delta. Netting happens where risk actually offsets — within a bucket by supervisory correlation ρ, across buckets by γ — and because hedges can fail in both calm and panic, the whole aggregation runs three times (correlations ×1.00 / ×1.25 / ×0.75) taking the worst. DRC then charges default separately (sensitivities see spreads, not jumps-to-default), and RRAO sweeps up payoffs no sensitivity can represent. It is a model in standardised clothing — which is precisely why it can serve as everyone's floor and every desk's fallback.

Requirements

Sensitivity pipelineDeltas/vegas per prescribed factor and tenor from validated pricing libraries, reconciled to front-office risk; curvature scenarios computable position-by-position.
Bucketing & mappingEvery position mapped to class → bucket → factor; index look-through where required; basis risks (onshore/offshore, cross-currency) captured.
DRC dataObligor hierarchies, seniority per instrument, maturity scaling; net JTD computed within obligor.
RRAO registerExotic underlyings and residual-risk instruments tagged at trade capture — 1.0% vs 0.1% is a booking-quality question.
GovernanceMonthly production with independent control; the A-SA is also the IMA fallback and the benchmark supervisors compare desks against — errors here surface twice.

Calculations

Delta with a hedge — one bucket Long equity A: WS₁ = +€3.0m · short equity B (same bucket): WS₂ = −€2.0m · ρ = 25%
K_b = √(3² + 2² + 2×0.25×3×(−2)) = √(13 − 3) = €3.16m
vs €5m gross — imperfect correlation buys 37% netting, no more. Under the ×0.75 scenario ρ = 18.75% → K_b rises: the worst scenario governs.
Curvature — why delta is not enough Sold €10m of options, delta-hedged: delta ≈ 0, but the ±RW shift loses €0.6m on both sides → CVR⁺ = CVR⁻ = 0.6 → curvature charge €0.6m on a "flat" book.
DRC — netting inside the obligor Long €10m senior bond, short €4m same issuer: net JTD = 6m × LGD 75% = €4.5m × RW(BBB) 6% = €0.27m. Long-short across different obligors nets only via the hedge-benefit ratio — issuer identity is everything.
RRAO €20m notional longevity-linked note: 20 × 1.0% = €0.2m, flat, regardless of hedging — the charge for being unmodellable by construction.

Changes & sources

Old SAA-SA
Notional blocks, crude offsetsSensitivities, correlated aggregation, three scenarios
No curvature conceptExplicit curvature add-on
Specific risk chargesDRC on net JTD + RRAO
  • Sources: Art 325c–325ay; RTS on curvature treatment, FX/materiality; EBA reporting ITS since 2021. From 2027 the A-SA is also the floor input and benchmarking object — its numbers acquire three audiences at once.

Independent analysis by Elderlake Partners, as at August 2026. Not legal, accounting or supervisory advice.