Reader question: A derivatives counterparty can remain solvent while its credit spread widens and the bank’s mark-to-market falls. How does Basel capitalise that credit-valuation-adjustment risk separately from ordinary counterparty default risk?
The Basel CVA framework treats CVA as a market-value risk. It asks how the regulatory CVA of covered transactions changes when counterparty credit spreads and relevant market risk factors move. Banks calculate capital using either the Basic Approach to CVA (BA-CVA) or, with supervisory approval and sufficient infrastructure, the Standardised Approach to CVA (SA-CVA). BA-CVA uses supervisory risk weights, EAD, maturity and hedge-recognition formulas. SA-CVA is sensitivity-based: the bank calculates regulatory CVA and its sensitivities to prescribed credit-spread and market risk factors, applies risk weights and correlations, then aggregates them into capital.
The key distinction is conceptual: CVA capital is not the same as the expected accounting CVA reserve and not the same as capital for the counterparty’s outright default.
What this page owns — and what it does not
This page owns:
covered counterparty portfolio + regulatory CVA risk factors + eligible hedges → BA-CVA or SA-CVA capital → CVA RWA.
It does not replace XVA valuation, SA-CCR, or large-exposure concentration limits. Those are different risk questions.
This is public prudential mathematics, not legal or regulatory advice for a real bank.
Why CVA exists at all
Suppose a bank is owed positive future value by Counterparty A. If A’s credit spread widens, the market value of the bank’s claim falls even before A defaults.
Accounting CVA captures this reduction in fair value. Basel CVA capital then asks how much capital is needed for adverse changes in CVA over the risk horizon.
This is different from counterparty default capital, which asks what happens if A actually defaults.
Covered transactions form the CVA portfolio
Basel applies the CVA framework to a defined portfolio of covered transactions and eligible CVA hedges. OTC derivatives are central to the scope, while the framework also addresses securities-financing transactions that are fair-valued for accounting purposes when their CVA risk is material.
Transactions excluded from CVA capital should not be mixed into the calculation simply because they have counterparty exposure under another prudential rule.
Two approaches: BA-CVA and SA-CVA
The current Basel framework provides two approaches:
- BA-CVA: a supervisory formula using counterparty risk weights, EAD, maturity and limited hedge recognition;
- SA-CVA: a sensitivity-based method requiring regulatory-CVA modelling and supervisory approval.
SA-CVA is more risk-sensitive but requires stronger systems, data, model governance and a dedicated CVA risk-management function.
BA-CVA: the reduced version
Every bank using BA-CVA must be able to calculate the reduced version.
The reduced BA-CVA does not recognise CVA hedges. At counterparty level, the formula uses a stand-alone CVA charge built from:
- counterparty credit-spread risk weight;
- effective maturity of the relevant netting set;
- EAD measured under the same counterparty-credit-risk methodology used for minimum capital;
- Basel supervisory parameters.
The portfolio formula then combines systematic and idiosyncratic components rather than simply summing every counterparty charge linearly.
Why EAD comes from the counterparty-credit-risk engine
BA-CVA does not invent an entirely separate exposure model.
For each netting set, EAD is calculated as it is for counterparty-credit-risk capital. If the bank uses SA-CCR for that counterparty, the CVA engine consumes the approved SA-CCR EAD.
This creates an important integration test: the same netting set should not have one unexplained EAD in SA-CCR capital and a different EAD in BA-CVA.
Counterparty spread risk weights
In BA-CVA, Basel maps counterparties to supervisory CVA risk weights reflecting expected credit-spread volatility. The mapping depends on prescribed credit quality and sector categories.
The weight is not the counterparty’s probability of default.
It is a supervisory proxy for how volatile the counterparty’s spread—and therefore CVA—can be.
The discount scalar
The Basel BA-CVA framework applies a discount scalar:
DSBA-CVA = 0.65.
The scalar modifies the raw BA-CVA aggregation into the final capital requirement.
A calculation that reproduces the unscaled supervisory formula but omits the discount scalar is not the Basel capital requirement.
Full BA-CVA: recognise eligible credit-spread hedges
The full BA-CVA is designed for banks that hedge CVA risk.
It starts from the reduced BA-CVA structure but adds recognition for eligible single-name and index credit hedges under detailed conditions.
The full formula contains a supervisory parameter:
β = 0.25,
which creates a floor limiting how much hedging recognition can reduce BA-CVA capital.
Why a hedge cannot reduce capital to zero automatically
A CDS hedge can protect against counterparty spread widening, but real-world basis risk remains:
- the hedge reference entity may differ from the counterparty;
- maturity can differ;
- the hedge can respond differently to market liquidity;
- the CVA exposure profile itself changes over time.
The BA-CVA hedge floor recognises that imperfect hedge effectiveness should remain capitalised.
Eligible hedge rules matter
Basel permits specified single-name credit instruments and index hedges subject to eligibility rules.
For example, a single-name credit instrument can reference the counterparty directly or, under prescribed conditions, a legally related or sufficiently similar entity.
A hedge that looks economically plausible but falls outside the rulebook should not be recognised mechanically in regulatory CVA capital.
SA-CVA: calculate regulatory CVA first
SA-CVA starts from the bank’s regulatory CVA for each counterparty.
Basel requires regulatory CVA to represent expected future loss from counterparty default under the assumption that the bank itself is default-free.
At minimum, regulatory CVA uses:
- a term structure of market-implied counterparty PD;
- market-consensus expected LGD;
- simulated paths of discounted future exposure.
This is a market-consistent regulatory quantity, not necessarily identical to the bank’s accounting CVA implementation.
SA-CVA then differentiates CVA
Once regulatory CVA is available, the bank calculates sensitivities to prescribed risk factors.
Basel identifies two broad kinds of drivers:
- counterparty credit-spread risk factors;
- market risk factors that change the value and exposure of covered transactions, such as interest rates and FX.
The bank effectively asks:
How much does regulatory CVA change for a prescribed small shock to each risk factor?
A simple sensitivity picture
If regulatory CVA is 10 million and a one-basis-point widening of a counterparty spread increases regulatory CVA by 80,000, then the spread sensitivity is approximately:
80,000 per bp.
The SA-CVA framework does not stop there. It risk-weights and aggregates sensitivities across counterparties, tenors, risk classes and prescribed correlations.
Why market factors enter CVA capital
CVA depends on both default probability and expected exposure.
If interest rates move, a swap’s future positive exposure can change. If FX moves, the replacement value of a cross-currency position can change.
So even if counterparty credit spreads are unchanged, CVA can move because the exposure profile changes.
This is why SA-CVA includes market-risk-factor sensitivities in addition to counterparty-spread sensitivity.
SA-CVA borrows the sensitivity-aggregation architecture
Conceptually, SA-CVA resembles the sensitivities-based market-risk framework:
- calculate sensitivities;
- apply supervisory risk weights;
- aggregate within buckets using prescribed correlations;
- aggregate across buckets and risk classes;
- include recognised hedge sensitivities with the prescribed sign and treatment.
But the risk factors and calibrations are CVA-specific.
SA-CVA needs supervisory approval
Basel requires a bank using SA-CVA to meet minimum operational and governance conditions.
Among them, the bank must be able to model exposure and calculate CVA and CVA sensitivities at least monthly, and it must have a CVA desk or similar dedicated function responsible for CVA risk management and hedging.
This is an important model-governance boundary: a bank cannot adopt SA-CVA merely because it can code the formula.
BA-CVA versus SA-CVA
| BA-CVA | SA-CVA | |
|---|---|---|
| Core input | EAD, maturity, supervisory spread weights | Regulatory CVA sensitivities |
| Hedge recognition | none in reduced version; limited in full version | risk-factor sensitivity based |
| Model requirement | lower | higher |
| Supervisory approval | baseline route | required |
| Risk sensitivity | coarser | richer |
CVA capital is separate from accounting CVA reserve
An accounting CVA reserve reduces the fair value of a derivative asset.
CVA capital covers the risk that CVA itself changes adversely.
A bank can therefore have:
- a current accounting CVA reserve;
- counterparty default capital;
- additional CVA risk capital.
These should not be added or netted casually because they answer different questions.
CVA capital is also different from DVA
DVA reflects changes in value from the bank’s own credit risk.
Basel regulatory CVA is constructed under the assumption that the bank itself is free from default risk for the purpose of the CVA risk capital calculation.
So accounting bilateral valuation adjustments and regulatory CVA capital do not have identical boundaries.
RWA conversion
As with other Basel capital requirements, CVA capital can be expressed as RWA through:
CVA RWA = 12.5 × CVA capital requirement.
This lets CVA risk join credit, market and operational risk in the bank’s total RWA denominator.
Inputs and outputs
A CVA-capital engine can require:
- covered transaction population;
- counterparty/netting-set identifiers;
- EAD and effective maturity;
- counterparty sector and credit-quality mapping;
- eligible CVA hedges;
- regulatory-CVA model;
- market-implied PD curves;
- market-consensus ELGD;
- simulated discounted future exposures;
- CVA sensitivities;
- Basel/local rule version and supervisory approval status.
Outputs can include:
- reduced BA-CVA charge;
- full BA-CVA hedge-recognised charge;
- SA-CVA charge by risk class;
- CVA RWA;
- hedge effectiveness and basis-risk diagnostics;
- counterparty and bucket contribution reports.
Evidence polarity: what supports confidence?
Evidence for a reliable CVA-capital result includes covered transactions that reconcile to the derivatives inventory, EAD matching the approved CCR engine, maturity inputs matching netting sets, counterparty spread mappings documented, eligible hedges validated independently, regulatory CVA reproducing from market inputs, and sensitivities stable under small shock sizes.
Evidence against confidence includes missing counterparties, CVA hedges that reduce capital but are not actually eligible, inconsistent EAD across SA-CCR and CVA, sensitivities that change wildly with numerical bump size, stale credit curves or a SA-CVA implementation without required supervisory approval.
Counterexample: counterparty default risk can be unchanged while CVA capital rises
If market volatility increases or exposure becomes more sensitive to rates and FX, SA-CVA can rise even if the counterparty’s current PD is unchanged.
CVA risk is about mark-to-market sensitivity of expected counterparty loss, not only today’s default probability.
Counterexample: a perfect notional hedge can be a poor CVA hedge
A CDS with the same notional as a derivative portfolio can still mismatch CVA because exposure is time-varying and conditional on market states.
Notional equality is not the same as sensitivity equality.
Counterexample: accounting CVA can fall while regulatory capital rises
If current CVA reserve falls because spreads tighten but volatility or hedge basis deteriorates, the risk of future CVA movement can rise.
Level and risk are different quantities.
Counterexample: SA-CCR capital and CVA capital can move in opposite directions
Collateral changes or netting can reduce counterparty EAD while credit-spread volatility rises, increasing CVA sensitivity.
The two capital frameworks are connected but not monotonic copies of each other.
Weak links in implementation
scope mismatch. Covered transactions differ between systems without explanation.
EAD inconsistency. BA-CVA uses stale or independently rebuilt exposure rather than the approved CCR value.
maturity mismatch. Netting-set maturity rules are simplified incorrectly.
hedge eligibility error. Economic hedges are treated as regulatory hedges automatically.
credit-curve staleness. Regulatory CVA uses outdated counterparty spreads.
numerical sensitivity noise. bump sizes are too small or too large.
wrong-sign hedge treatment. hedge sensitivities increase rather than offset CVA risk.
model boundary drift. accounting CVA and regulatory CVA assumptions are mixed silently.
Diagnostics: how to test the engine
- scope test: reconcile covered transactions and eligible hedges to source inventories.
- EAD test: BA-CVA netting-set EAD must match the approved CCR method.
- reduced/full test: full BA-CVA cannot avoid calculating the reduced version.
- hedge-floor test: extreme hedge size should not drive full BA-CVA below the framework’s hedge-recognition floor.
- shock-sign test: widening counterparty spreads should generally increase positive CVA loss.
- sensitivity-bump test: halve and double numerical shocks and verify stable finite-difference sensitivities.
- market-factor test: rate and FX shocks should change CVA where exposure is sensitive to those variables.
- approval test: SA-CVA use must be tied to valid supervisory approval and governance evidence.
- RWA test: CVA RWA should equal 12.5 times the capital requirement under the relevant disclosure convention.
- independent-pricer test: selected counterparties should reproduce regulatory CVA and sensitivities independently.
What would falsify confidence?
Confidence should be withdrawn if CVA capital cannot be reproduced from approved inputs; if BA-CVA EAD differs from CCR without explanation; if ineligible hedges reduce capital; if sensitivity signs are inconsistent with economic behaviour; if SA-CVA approval is absent; or if the model cannot separate accounting XVA assumptions from regulatory CVA requirements.
Alternatives and limits
BA-CVA offers simpler, more supervisory capitalisation. SA-CVA offers richer risk sensitivity but adds modelling and governance cost. Internal economic CVA risk can be measured with even more detailed simulation, full revaluation, stress testing and P&L explain.
No Basel capital formula removes model risk. CVA depends on future exposure, counterparty spreads, LGD, collateral, wrong-way risk and market dynamics. Capital measurement compresses those uncertainties into a prudential framework; it does not eliminate them.
How this connects to the surrounding knowledge estate
XVA algorithms explain the valuation adjustments. SA-CCR supplies regulatory exposure for many BA-CVA implementations. SIMM is another sensitivity-aggregation framework but serves bilateral initial margin rather than CVA capital. The resulting CVA RWA flows into balance-sheet optimisation.
Verification and update triggers
Preserve Basel/local rule version, approach assignment by counterparty, supervisory approval status, EAD source, maturity calculations, hedge eligibility, regulatory-CVA model, market data and sensitivity engine version. Revalidate after Basel amendments, CCR methodology changes, CVA model redevelopment, hedge-strategy changes or material unexplained CVA P&L.
Primary and high-quality references
- Basel Committee on Banking Supervision, MAR50 — Credit valuation adjustment framework.
- Basel Committee, Basel Framework, including SA-CVA regulatory-CVA and sensitivity requirements.
- Basel Committee, DIS51 — CVA risk disclosures.
- Basel Committee, Counterparty credit risk overview.
Educational boundary: This article explains public Basel CVA-capital mechanics. It does not determine any real bank’s regulatory capital or provide personalized financial advice.