XVA · CVA · Counterparty risk

CVA, DVA, FVA and MVA explained

· 8 min read · Everix team

Short answer: the risk-free value of a derivative ignores who you trade with and how you fund it. CVA charges for the counterparty's default risk, DVA credits your own, FVA prices the funding of uncollateralised exposure and MVA prices the funding of initial margin you post. All four are computed from simulated exposure profiles, so they come from the same engine.

Four adjustments at a glance

CVA
Question it answers
What could I lose if my counterparty defaults while it owes me money?
Driven by
Expected positive exposure (EPE), counterparty default probability, LGD
Typical sign for the firm
Cost
DVA
Question it answers
What do I gain, in fair value terms, from the possibility of my own default while I owe money?
Driven by
Expected negative exposure (ENE), own default probability, own LGD
Typical sign for the firm
Benefit
FVA
Question it answers
What does it cost to fund the cash I am owed but not collateralised, or what do I save on cash I hold?
Driven by
EPE and ENE, funding spread over the discounting rate
Typical sign for the firm
Cost (FCA) and benefit (FBA)
MVA
Question it answers
What does it cost to fund the initial margin I must post over the life of the trade?
Driven by
Expected future initial margin, funding rate net of the margin remuneration rate
Typical sign for the firm
Cost

The common building block: exposure profiles

Every adjustment starts with the same question: what will this netting set be worth on each future date? A Monte Carlo engine simulates market risk factors forward (for example a two-factor Gaussian rates model such as G2++, or a stochastic-volatility model such as Heston for FX or equity), reprices every trade on every path and date, applies netting and the collateral agreement, and averages:

  • EPE(t), the expected positive exposure: the expected positive close-out exposure after recognised netting and collateral. Expected default loss additionally applies loss given default and the default model.
  • ENE(t), the expected negative exposure: the analogous expected negative close-out exposure, reported here as a signed negative value. Realised default loss also depends on recovery and close-out.
  • PFE(t), a high quantile of the positive exposure, used for credit limits rather than pricing.

Collateral matters here. Variation margin with a low threshold cuts EPE to roughly the move over the margin period of risk. Initial margin cuts it further, but creates the funding cost that MVA measures.

CVA

The usual discretised, unilateral form is:

CVA ≈ (1 − R) × Σᵢ DF(tᵢ) × EPE(tᵢ) × PD(tᵢ₋₁, tᵢ)

R is the recovery rate, DF the discount factor and PD the counterparty's marginal default probability between two dates, usually implied from CDS or bond spreads. Three practical points:

  • Wrong-way risk. The formula assumes exposure and default are independent. When they are not (for example a cross-currency swap with a counterparty whose credit worsens as its currency weakens), CVA needs a joint simulation.
  • Proxy spreads. Most counterparties of a mid-size bank have no liquid CDS. Their spreads are mapped from sector, region and rating curves, and that mapping is a governance question.
  • Capital is a different number. Accounting CVA is a valuation; CVA risk capital under CRR3 measures the risk of CVA changing. CRR Article 381 defines CVA for capital purposes as excluding the institution's own credit risk. See BA-CVA vs SA-CVA under CRR3.

DVA

DVA ≈ (1 − R_own) × Σᵢ DF(tᵢ) × |ENE(tᵢ)| × PD_own(tᵢ₋₁, tᵢ)

DVA is the mirror image of CVA. Fair value under accounting standards reflects non-performance risk, which includes your own credit risk on derivative liabilities, so bilateral CVA is CVA minus DVA. Regulators are less generous. CRR Article 33(1)(c) removes from own funds the fair value gains and losses on derivative liabilities that come from the institution's own credit risk, and Article 33(2) forbids offsetting them against counterparty credit risk. A bank whose credit worsens books a DVA gain that does not count as capital.

FVA

If a trade is uncollateralised and in your favour, you have paid out, or will pay out, cash that the counterparty does not return as collateral. You fund that cash at your own funding rate, not at the risk-free rate used for discounting. FVA prices that gap:

FCA ≈ Σᵢ DF(tᵢ) × EPE(tᵢ) × s_borrow(tᵢ) × Δtᵢ
FBA ≈ Σᵢ DF(tᵢ) × |ENE(tᵢ)| × s_lend(tᵢ) × Δtᵢ
FVA = FCA − FBA

s is the funding spread over the discounting curve, often weighted by survival probabilities. FBA overlaps conceptually with DVA, because both reward you for owing money while your credit is weak. Desks therefore choose a consistent set: for example CVA, DVA and FCA only, or CVA, FCA and FBA without DVA. Whatever you choose, document it and do not count the same benefit twice.

MVA

Under the uncleared margin rules, both sides post initial margin into segregated accounts, and CCPs require it for cleared trades. You cannot reuse that margin, so its funding cost depends on the difference between the total funding rate and the total remuneration rate on posted collateral, annualised in the same currency:

MVA ≈ Σᵢ DF(tᵢ) × E[IM(tᵢ)] × (r_funding(tᵢ) − r_IM(tᵢ)) × Δtᵢ

The hard part is E[IM(t)]: future initial margin along each simulated path. For bilateral trades that means projecting ISDA SIMM, a sensitivity-based model, into the future. Common shortcuts are regressing SIMM on simulated risk factors or scaling today's sensitivities by a decay profile. Each SIMM recalibration also changes today's starting point, which is one reason to track SIMM versions.

What changes with collateral

Qualitative pattern. Actual sizes depend on the portfolio, the CSA terms and the market.

Uncollateralised
CVA
Full exposure
DVA
Full exposure
FVA
Largest
MVA
None
Daily VM, zero threshold
CVA
Small (margin period of risk only)
DVA
Small
FVA
Small
MVA
None
Daily VM plus IM (bilateral, under UMR)
CVA
Very small
DVA
Very small
FVA
Small
MVA
Present
Centrally cleared
CVA
Exposure to the CCP, usually treated as small
DVA
Small
FVA
Small
MVA
Present (CCP IM)

Collateral turns credit risk into funding and liquidity cost. A trade that looks free of CVA under a strong CSA can still carry meaningful MVA.

KVA, briefly

Many desks add a capital valuation adjustment (KVA), the cost of holding regulatory capital (counterparty credit risk, CVA risk, market risk) over the life of the trade. It needs projected capital, which in turn needs projected SA-CCR exposures and CVA capital.

Getting the numbers right in practice

  1. One trade model and one set of market data for all adjustments, so CVA and FVA use the same exposure.
  2. Enough paths and dates for stable tails: check convergence when portfolios change, not just once.
  3. Allocation to trades. Desks and pricing need incremental or marginal XVA per trade, not just a netting-set total.
  4. Explain the change: new trades, market moves, credit spreads, model or parameter changes, day to day.
  5. Reconciliation with accounting, risk and capital, which often use different subsets of the adjustments.

Where Everix fits

Everix's Monte Carlo XVA engine (CVA, DVA, FVA and LVA, with G2++ and Heston models and per-trade allocation) has been in production at a European financial institution since 2018. Read the anonymised case study, or see the XVA module. The same trade model feeds SA-CCR and ISDA SIMM, the inputs behind KVA and MVA. Terms are defined in the glossary.

Formulas are the standard discretised forms used in practice; conventions (bilateral vs unilateral, survival weighting, funding curves) vary by firm. Regulatory references: Regulation (EU) No 575/2013, Articles 33, 381 and 382. General information, not accounting or regulatory advice.

The Everix note

Get the next article by email

IFR, CRR3, SIMM changes and AI-ready risk data. Roughly monthly, double opt-in.

By subscribing, you ask for the Everix note. Confirm by email; unsubscribe any time. Privacy policy

Questions about your own portfolio?

Talk to the quants and engineers behind Everix — 30 minutes to check the fit.