XVA Curves: What They Are and How They Are Observed

XVA is not a single number. It is the aggregation of several components, each driven by a different “curve” or economic input. In a CSA negotiation, understanding how these curves are constructed, observed, and challenged is critical, because it determines how much value is released when moving to a collateralised structure.

CVA Curve (Credit Valuation Adjustment)

CVA represents the expected loss arising from counterparty default. It is driven by three elements: the exposure profile of the trade or portfolio, the probability of default of the counterparty, and the loss given default.

The CVA curve is effectively the term structure of default risk. It shows how the market prices the probability of default at different maturities.

This is the most observable component of XVA. It is typically derived from the CDS spreads of the counterparty. These spreads are bootstrapped into a hazard rate curve, which in turn produces survival probabilities over time. Where a direct CDS curve does not exist, dealers use proxy curves based on similar credits, which introduces some discretion.

Because it is anchored in traded market instruments, CVA is relatively transparent and defensible. However, differences still arise through the choice of CDS reference, liquidity adjustments, and proxy selection. These differences are often material and negotiable, but they can always be benchmarked back to observable market data.

A further area of discretion arises in the treatment of recovery rates. Dealers will sometimes argue that the recovery rate applicable to derivatives differs from that implied by the CDS market, and may introduce an adjustment to the CVA curve on that basis. The practical effect of this is typically to reduce the CVA benefit available to the client (i.e. reduce CVA release) in the negotiation.

In practice, this is often a negotiating lever. While there are cases where structural features - such as seniority of claims or close-out mechanics - could justify a differentiated recovery assumption, such arguments should be treated with caution. Recovery is fundamentally a counterparty credit parameter, and CDS markets already embed the market consensus on recovery expectations. Deviating from this requires a clear and defensible rationale.

These assumptions can and should be challenged directly, benchmarked against observable CDS-implied recoveries and peer dealer treatment, and tested against the actual legal framework governing the exposure. Unsupported adjustments in this area are a common source of value leakage and are typically straightforward to identify and eliminate.

FVA Curve (Funding Valuation Adjustment)

FVA reflects the cost or benefit of funding uncollateralised exposure. It arises because a dealer must fund any negative cash position associated with the trade over its life.

The FVA curve is typically presented as the term structure of the dealer’s funding cost over risk-free rates, built as a spread over the OIS discount curve.

Unlike CVA, this curve is only partially observable. Dealers infer it from a combination of their own bond issuance levels, CDS spreads, and internal treasury transfer pricing frameworks. There is no single agreed market curve. This creates significant scope for variation. Different dealers use different funding assumptions, and even within the same dealer the methodology may vary over time. Key areas of negotiation include whether the curve reflects unsecured or secured funding, and whether funding is treated symmetrically.

For the avoidance of doubt, netting sets, collateral terms, and collateral reuse affect the exposure profile (i.e. the EPE to which FVA is applied), not the funding curve itself. Blurring this distinction is a common source of confusion and, in some cases, a mechanism for embedding additional margin within the pricing.

A critical point is that FVA is inherently symmetric. The same framework that generates a funding cost when the dealer is out-of-the-money (negative MTM) must also generate a funding benefit when the dealer is in-the-money (positive MTM), reflected through the discounting of receivables. Any asymmetry - where costs are included but benefits are reduced, deferred, or excluded - represents a pricing choice rather than a structural necessity.

This symmetry provides a practical mechanism to enforce discipline in negotiation. By requiring a consistent two-way treatment, it becomes significantly easier to identify and challenge “convenient” assumptions, including selective curve construction, asymmetric application of funding spreads, or inconsistent treatment across trades or counterparties.

More fundamentally, there is an unresolved tension in how FVA should be defined. In principle, the value of a derivative should not depend on the funding cost of a specific dealer. If it did, identical trades would have different economic values depending on the intermediary, and would tend to collapse in value as the dealer’s own credit quality deteriorates.

A more consistent interpretation is that funding should clear at a market-implied rate for the risk being transferred, rather than the internal funding cost of a particular institution. In practice, however, markets do not enforce this discipline. Dealers price off their own funding curves, and clients are exposed to that heterogeneity.

This makes FVA one of the primary drivers of pricing dispersion across dealers, and one of the most important areas to challenge in a negotiation. While it is difficult to impose a fully market-clearing framework, it is entirely possible to interrogate assumptions, enforce symmetry, and ensure that funding effects are applied consistently and transparently.

Therefore, here the key negotiating argument:  FVA is symmetric. If you charge us for funding when you are out-of-the-money, you must give us the benefit when you are in-the-money - on the same curve and methodology. If that symmetry doesn’t hold, we will normalise the pricing across dealers and remove the distortion.

KVA Curve (Capital Valuation Adjustment)

KVA represents the cost of holding regulatory capital against a derivatives position. It reflects the fact that trades consume balance sheet and must generate a return on that capital over their life.

The KVA curve is not a market curve in the conventional sense. It is constructed internally by the dealer, combining the projected capital requirement over time with an assumed return on equity.

The capital requirement itself is driven by regulatory frameworks such as SA-CCR. These frameworks do not use the expected exposure profile (EPE) directly. Instead, they are calibrated to stressed exposure measures - effectively a conservative estimate of potential future exposure under adverse market conditions.

As a result, the exposure profile underlying KVA is structurally higher than EPE. It reflects tail exposure rather than expected exposure, incorporating regulatory add-ons, supervisory factors, and limited recognition of netting and diversification. This creates a capital profile that is both larger and more persistent over time than the profile used for CVA.

There is also significant flexibility in how this capital profile is constructed. Dealers retain discretion in:

• The implementation of regulatory models
• The treatment of netting sets and diversification
• The allocation of capital across portfolios and counterparties
• The mapping from regulatory capital to internal economic capital

The cost of capital applied to this profile is entirely internal. Dealers typically assume a return on equity in the range of 10–15% or higher, but this is a strategic choice rather than a market-observed parameter.

A further layer of subjectivity arises in how this return is translated into a present value charge. KVA is typically computed as the discounted cost of future capital, but both the discount rate and the interaction between the cost of capital and funding assumptions are not standardised. In particular:

• The return on capital (RoE) is an internal target, not a funding cost
• The discount rate applied to future capital costs may or may not be consistent with the funding curve used elsewhere (e.g. FVA)
• This creates scope to amplify or dampen the present value of KVA through modelling choices

As a result, KVA is the least transparent and most subjective component of XVA. It is not directly observable in the market and cannot be independently verified in the same way as CVA or even FVA.

This creates two important consequences:

• The underlying exposure profile is structurally conservative (stressed rather than expected), increasing the magnitude of the adjustment
• The pricing of that profile depends on internal assumptions around return on equity and discounting, which are not externally benchmarked

Taken together, this makes KVA the area with the greatest scope for overstatement and the largest informational asymmetry between dealer and client - and therefore a primary focus in any pricing or CSA negotiation.

How These Curves Drive XVA Release Under a CSA

When a counterparty moves from an uncollateralised relationship to a two-way CSA, the economic benefit can be understood through the reduction in CVA, FVA, and KVA.

• CVA falls sharply because variation margin, and in some cases initial margin, substantially removes the dealer’s exposure to counterparty default.
• FVA also falls, because the dealer no longer needs to fund a large uncollateralised exposure. In practice, this is often the most visible part of the adjustment, because FVA is commonly reflected in the dealer’s declared mark-to-market. That makes it harder to conceal, and easier for a client to observe and challenge through pricing comparisons.
• KVA declines as well, because lower exposure generally translates into lower regulatory capital consumption and therefore a lower capital charge.

A critical distinction is that CVA and KVA are typically not included transparently in the declared mark-to-market in the same way. They often sit within internal reserve frameworks, transfer pricing, or broader profitability assessments, making them much less visible to the client. That gives the dealer more scope to retain value or to understate the economic benefit of moving to a two-way CSA.

The total XVA release is therefore the combined effect of reducing credit risk, funding requirements, and capital consumption. But from a negotiation perspective, the components are not equally observable. FVA is usually the hardest for the dealer to hide because it is more directly embedded in pricing and MTM. CVA and especially KVA are more opaque, more model-dependent, and therefore more susceptible to discretion in how much of the benefit is acknowledged or shared.

Commercial Reality in Negotiation

• CVA is anchored to observable market data and can be benchmarked and challenged directly. 
• FVA is partly market-based but involves significant dealer discretion, making it a key area for negotiation - but the dealer has typically already declared their hand here.
• KVA is primarily an internal construct, driven by regulatory interpretation and return-on-capital assumptions, and is therefore the least transparent and most open to challenge.

In practice, effective negotiation requires breaking XVA into these components and forcing each one onto its underlying economic drivers. The more each curve can be tied back to observable or defensible inputs, the less scope there is for pricing opacity and the more of the XVA benefit can be captured.