CVA, DVA and FVA can look like an intimidating pile of acronyms. They aren't. Every single one answers the same underlying question: what does it really cost to trade with an entity that might default, and to fund that trade in the real world? Hold onto that one sentence and you can reason your way through any XVA variant you haven't seen before — which is exactly the skill a Product Controller is expected to bring to the table every day.
What this session covers
By the end of this session, you will be able to:
- Explain, in plain English, why a derivative's fair value is not simply its risk-free discounted cash flow.
- Define CVA, DVA and FVA and connect each to a specific real-world risk (counterparty default, own default, funding cost).
- Walk through a worked numerical example for each adjustment.
- Describe how a Product Controller actually encounters XVA day-to-day — P&L attribution, independent price verification (IPV), and reconciliation with Front Office and Risk.
- Recognise the common control breaks around XVA and how experienced Product Controllers investigate them.
Agenda
| Topic |
|---|
| Why derivatives aren't "free" — setting the scene |
| CVA — Credit Valuation Adjustment |
| DVA — Debit Valuation Adjustment |
| FVA — Funding Valuation Adjustment |
| The wider XVA family (MVA, KVA, ColVA) |
| The Product Controller's world: P&L, IPV, controls, reconciliation |
| End-to-end case study |
| Q&A and interview-style questions |
Why Derivatives Aren't "Free" to Trade
Textbook derivative pricing (Black-Scholes, standard swap valuation) assumes two things that don't hold in the real world:
- Both counterparties will always pay what they owe, in full, on time (no default risk).
- A bank can fund the trade at the risk-free rate (no funding cost).
Neither assumption survived the 2008 financial crisis intact. Banks discovered — expensively — that counterparties do default (Lehman Brothers, Bear Stearns), that a bank's own credit spread affects the value of its liabilities, and that funding an uncollateralised derivative is not free. XVA is the collective name for the adjustments that correct textbook value for these real-world frictions.
Fair Value = Risk-Free Value − CVA + DVA − FCA + FBA − (other adjustments). Each XVA takes something the textbook price ignores and puts a number on it.
The Building Block: What Is Exposure?
Every XVA calculation starts with the same question: if my counterparty defaulted today, what would this derivative be worth to me? That number — the replacement cost — is called Exposure. It is not the trade's notional; it is the trade's mark-to-market value at the point of default, and because markets move, exposure itself is a distribution of possible future values, not a single number.
You enter a 5-year interest rate swap where you receive fixed and pay floating, with zero value at inception.
If rates fall over the next year, your swap becomes an asset to you (you're receiving above-market fixed) — you now have positive exposure to your counterparty, because if they defaulted, you'd lose that in-the-money value.
If rates rise instead, the swap becomes a liability to you — your exposure to the counterparty is zero (you'd simply owe them, default risk is now theirs, not yours).
This is why exposure is modelled as a full distribution across future dates, not a single figure — it depends on which way the market moves.
Two exposure metrics matter for XVA:
- Expected Positive Exposure (EPE): the probability-weighted average of positive future exposure — this drives CVA.
- Expected Negative Exposure (ENE): the mirror image, used for DVA — the probability-weighted average of exposure where you owe money.
CVA — Credit Valuation Adjustment
The Plain-English Definition
CVA is the market price of your counterparty's credit risk on a derivative. It is the amount by which you must reduce the value of a derivative asset to compensate for the possibility that your counterparty defaults before paying you what they owe.
Sign convention: CVA is always a cost to the bank holding the asset. It is subtracted from the risk-free value.
The (Simplified) Formula
A commonly taught simplified form, assuming independence between exposure and default (i.e., ignoring "wrong-way risk" for now):
Where: EPE(t) = Expected Positive Exposure at time t; PD(t−1,t) = marginal probability of counterparty default in that period, typically derived from CDS spreads; Recovery Rate = assumed recovery on default (commonly 40% for senior unsecured).
In words: for every future time bucket, multiply (i) how much you'd lose if they defaulted right then, by (ii) the chance they actually default in that bucket, by (iii) the loss-given-default percentage, then discount back to today and sum across all buckets.
Worked Example
A bank has a 3-year uncollateralised interest rate swap with Counterparty X. Expected Positive Exposure profile (simplified, annual buckets): Year 1 = $50 mn, Year 2 = $70 mn, Year 3 = $40 mn.
Counterparty X's 1-year CDS-implied marginal default probability is roughly 1.5% per annum in each bucket. Recovery rate assumption: 40% (so Loss Given Default = 60%).
Ignoring discounting for simplicity: CVA ≈ 60% × [(50 × 1.5%) + (70 × 1.5%) + (40 × 1.5%)] mn = 60% × [0.75 + 1.05 + 0.60] mn = 60% × 2.40 mn ≈ $1.44 mn.
This $1.44 mn is a debit to P&L on day one — the swap's fair value booked is $1.44 mn lower than its risk-free mark, purely because Counterparty X is not risk-free.
Ignoring discounting for simplicity — in practice, each year's cash flow would be discounted by the relevant OIS discount factor for that tenor.
What Drives CVA Up or Down
| Driver | Effect on CVA |
|---|---|
| Counterparty credit spread widens (CDS up) | CVA increases — higher implied default probability |
| Trade moves further in-the-money for the bank | CVA increases — higher exposure at risk |
| Collateral (CSA) posted by counterparty | CVA decreases — exposure is collateralised, not naked |
| Netting agreement (ISDA Master + CSA) in place | CVA decreases — exposure calculated at netting-set level, not trade level |
| Longer maturity | CVA generally increases — more time for default and larger exposure horizon |
| Wrong-way risk present | CVA increases beyond the simplified formula — exposure and default probability are correlated |
A note on wrong-way risk: this is when a counterparty's likelihood of default is correlated with your exposure to them — for example, an oil producer as counterparty on a swap where you're paying them based on oil prices, and their creditworthiness also weakens when oil prices fall. This is a classic exam and interview topic; the simplified formula above assumes no such correlation.
Right-Way Risk: a gold producer hedging its output. If gold prices fall, the producer's creditworthiness might suffer, but the bank's exposure to them also falls because the swap is now a liability to the producer, not an asset to the bank.
DVA — Debit Valuation Adjustment
The Plain-English Definition
DVA is the mirror image of CVA: it recognises that the bank itself might default before paying what it owes on a derivative liability. If your own credit deteriorates, the market value of your liability to your counterparty falls — because they, too, might not receive full payment from you. Counterintuitively, this shows up as a gain in the bank's own P&L.
- If Bank A's own credit spread widens (the market thinks Bank A is more likely to default), the fair value of Bank A's liabilities — including derivative liabilities — falls.
- A falling liability value is, by definition, a gain to Bank A.
- So a bank's own credit deterioration can show up as a P&L gain through DVA — this is exactly the mechanism that attracted heavy criticism after the crisis (banks reporting profits as their own credit worsened) and is a favourite conceptual question in interviews.
Formula (Mirror of CVA)
Where ENE(t) is Expected Negative Exposure — the exposure profile where the bank itself owes money — and Own PD is the bank's own default probability, typically from its own CDS or bond spreads.
Worked Example
Using the same swap book, suppose the negative exposure profile (where the bank owes the counterparty) is: Year 1 = $30 mn, Year 2 = $20 mn, Year 3 = $15 mn.
The bank's own 1-year CDS-implied default probability is 1.0% per annum, recovery 40%.
DVA ≈ 60% × [(30 × 1.0%) + (20 × 1.0%) + (15 × 1.0%)] mn = 60% × [0.30 + 0.20 + 0.15] mn = 60% × 0.65 mn ≈ $0.39 mn.
This $0.39 mn is a P&L gain on day one, added back to the risk-free value.
Ignoring discounting for simplicity — in practice, each year's cash flow would be discounted by the relevant OIS discount factor for that tenor.
Product Controller Angle: Why DVA Is Contentious
- Under IFRS 13 / ASC 820 (fair value measurement), DVA is required for financial reporting because "own credit risk" is deemed part of the fair value of a liability.
- DVA gains are often described as unrealised until the liability is extinguished. However, they are realised economically upon the termination or novation of the derivative liability. The controversy arises from the counterintuitive nature of booking a gain due to one's own credit deterioration, which critics argue can obscure the underlying economic health of the bank.
- As a Product Controller, this means CVA/DVA reconciliation often needs two views side by side: the accounting P&L view (includes DVA) and the regulatory capital view (CVA capital charge under Basel III/IV FRTB-CVA framework, which largely ignores DVA).
FVA — Funding Valuation Adjustment
The Plain-English Definition
FVA captures the cost (or benefit) of funding an uncollateralised derivative position at the bank's own funding rate, rather than at the theoretical risk-free rate (typically OIS/SOFR).
Here is the intuition: if a bank has an uncollateralised derivative that is an asset (in-the-money), it doesn't receive cash today — it will only receive cash when the trade eventually settles or is closed out. In the meantime, the bank has effectively lent money to the counterparty and must fund that position from its own treasury desk, which borrows at the bank's own unsecured funding spread over OIS — not at the risk-free OIS rate itself. That spread, applied to the expected exposure over the life of the trade, is FVA.
- FCA (Funding Cost Adjustment): the cost of funding a positive exposure (asset position) — the bank has to borrow to finance being "owed" money on an uncollateralised in-the-money trade.
- FBA (Funding Benefit Adjustment): the benefit from funding a negative exposure (liability position) — the bank is effectively receiving cheap funding by owing money on the trade.
- FVA = FCA − FBA (in various conventions the sign is presented differently — always check whether a desk is quoting FVA net or its two components separately).
Formula (Simplified)
Own Funding Spread = the bank's unsecured borrowing spread over the risk-free/OIS curve (often proxied by the bank's own bond or CDS spread, adjusted for tenor).
Worked Example
Reuse the EPE profile (50, 70, 40 mn across 3 years) and ENE profile (30, 20, 15 mn). Assume the bank's own funding spread over OIS is a flat 0.80% per annum.
FCA ≈ (50 × 0.80%) + (70 × 0.80%) + (40 × 0.80%) = 0.40 + 0.56 + 0.32 = $1.28 mn (cost).
FBA ≈ (30 × 0.80%) + (20 × 0.80%) + (15 × 0.80%) = 0.24 + 0.16 + 0.12 = $0.52 mn (benefit).
Net FVA = FCA − FBA = 1.28 − 0.52 = $0.76 mn — a net cost, reducing the trade's booked value.
Ignoring discounting for simplicity — in practice, each year's cash flow would be discounted by the relevant OIS discount factor for that tenor.
Product Controller Angle: Why FVA Debates Never Fully Die Down
- FVA overlaps conceptually with DVA — both relate to the bank's own credit/funding profile — which is why some banks historically used FVA and DVA in a way that risked double-counting the same economics. Reconciling this overlap (or confirming the desk's methodology has genuinely eliminated it) is a recurring IPV and model-validation question.
- The choice of funding curve (unsecured bond curve vs. CDS-implied vs. internal treasury transfer-pricing curve) is a major source of valuation dispersion between banks and a common audit/IPV challenge point — Product Control needs a clear, documented, consistently-applied funding curve, refreshed and independently checked.
- FVA is generally NOT required under IFRS for external financial reporting in the same explicit way CVA/DVA are, but banks widely apply it for internal pricing, deal profitability and trader P&L — so a Product Controller must know which P&L view (statutory vs. management/trading) includes it, and reconcile the difference explicitly.
The Wider XVA Family
CVA, DVA and FVA are the core three you must know cold. Beyond them, the alphabet continues:
| XVA | What It Adjusts For | One-Line Intuition |
|---|---|---|
| MVA (Margin Valuation Adjustment) |
Cost of funding Initial Margin posted on cleared/collateralised trades | Posting IM to a CCP or under uncleared margin rules ties up cash — MVA is the funding cost of that posted margin. |
| KVA (Capital Valuation Adjustment) |
Cost of holding regulatory capital against the trade over its life | Capital isn't free — shareholders expect a return on it, so KVA prices in the cost of capital consumed by the trade. |
| ColVA (Collateral Valuation Adjustment) |
Mismatch between the collateral rate received/paid and the discounting rate used | ColVA captures the mismatch between discounting at the risk-free rate and the actual collateral rate — increasingly material as rates have diverged. |
| AVA (Additional Valuation Adjustment) |
Prudential Valuation buffers under regulatory rules (distinct from accounting fair value) | A regulatory-driven haircut to fair value for prudent valuation purposes — relevant to Pillar 1 capital, not accounting P&L. |
The practical takeaway is this: every XVA follows the same template — identify a real-world friction that textbook pricing ignores, quantify the expected cost/benefit using an exposure profile, and discount it back. Once you understand CVA deeply, DVA, FVA, MVA and KVA are variations on the same theme.
XVA in the Product Controller's Day-to-Day World
This is where theory meets the desk.
6.1 Where XVA Sits in P&L
- XVA typically sits within a centralised XVA desk (sometimes under Treasury or a dedicated CVA/FVA trading desk) rather than being calculated by each trading desk individually — this centralisation itself is a control point Product Control must understand, since XVA is allocated back to originating desks.
- Daily P&L for a derivatives desk is commonly split into: Clean/Risk-Free P&L (the textbook mark-to-market) + XVA P&L (CVA/DVA/FVA movement) + any residual/other adjustments. A Product Controller's P&L explain must isolate how much of a day's P&L swing came from market moves on the underlying trade versus XVA-driven moves (e.g., a counterparty's CDS spread widening).
- A classic P&L attribution exercise: if a desk's swap book P&L moves by $5 mn overnight, the Product Controller decomposes it into rates-driven MTM, FX translation, and XVA — and within XVA, further into CVA (did counterparty spreads move?), DVA (did the bank's own spread move?), and FVA (did funding spreads move?).
XVA Allocation
- Pro-Rata Allocation: allocating the centralised XVA desk's total P&L to desks based on their gross or net exposure. Simple, but can be seen as unfair.
- Marginal / Incremental Allocation: allocating XVA based on the additional XVA cost a new trade adds to the existing portfolio. More complex, but more accurate and considered more "fair" by trading desks.
- The Conflict: a desk adding a trade that reduces net risk (and thus XVA) might expect a credit, which is often the source of P&L disputes between the XVA desk and trading desks. This is a prime interview question for a Product Controller.
6.2 Independent Price Verification (IPV) of XVA
XVA is one of the hardest things to independently price-test because, unlike a vanilla bond, there's often no single observable market quote for "CVA on this specific netting set." Product Control's IPV approach typically includes:
- Verifying the counterparty credit curve used (single-name CDS where liquid; proxy/sector curves where the counterparty has no traded CDS) against independent market data providers (Markit, Bloomberg).
- Challenging the recovery rate assumption against market convention and rating-agency data.
- Re-performing or benchmarking the exposure simulation (EPE/ENE profiles) — often via a simplified independent model — against the front-office XVA engine's output, investigating any material divergence.
- Checking the own funding curve used for FVA against Treasury's official funds transfer pricing (FTP) curve, to ensure the desk isn't using a stale or off-market funding assumption.
- Confirming netting set and CSA terms (threshold, minimum transfer amount, eligible collateral) are correctly reflected in the exposure model — a very common source of breaks, since legal CSA terms change and the XVA engine's static data can lag.
6.3 Common Control Breaks and Root Causes
| Symptom | Likely Root Cause | Product Controller Action |
|---|---|---|
| CVA moves materially with no matching CDS spread move | Stale or incorrect counterparty curve mapping; netting set static data error | Reconcile counterparty curve feed; check netting set/CSA static data against legal agreement |
| FVA differs significantly from Treasury's own view | Desk using an outdated or internally-built funding curve instead of official FTP curve | Escalate to Treasury/Model Risk; request curve reconciliation |
| DVA gain flagged in audit as unusually large | Bank's own CDS spread widened sharply (idiosyncratic or market-wide stress) | Corroborate against bank's own traded CDS/bond spread; confirm consistent with disclosed methodology |
| XVA allocation to desks doesn't sum to centralised XVA desk total | Allocation methodology (pro-rata by exposure vs. marginal contribution) inconsistently applied | Trace allocation keys; confirm methodology matches Finance policy documentation |
| Day-1 XVA on a new large trade looks off vs. expectation | Wrong-way risk not captured in simplified model; incorrect notional/tenor in exposure engine | Have Front Office/Quant re-run with full simulation; sanity-check against back-of-envelope estimate |
| "Day-1 P&L" breaks: a trade is showing a large, unexplained profit on its first day | The trading desk has not deferred the day-1 P&L and is booking the unobservable model-implied price as profit | Immediately flag the trade, quantify the difference between the model price and the traded price, and ensure it's placed into a Day-1 P&L reserve to be amortized |
6.4 Reconciliation and Governance
- Monthly/quarterly reconciliation between the XVA desk's booked P&L and Product Control's independent recompute, with tolerance thresholds and formal sign-off — this is a real deliverable, not a theoretical exercise.
- Reconciliation between the accounting fair value adjustment (which includes a point-in-time CVA/DVA per IFRS 13) and the regulatory CVA capital charge. The latter, under the Basel FRTB-CVA framework, is a capital requirement designed to cover potential future losses from CVA volatility — a fundamentally different concept from the accounting adjustment itself.
- Model governance: XVA models fall under SR 11-7-style model risk management frameworks — Product Control liaises with Model Risk on model tiering, validation status, and any model limitations (e.g., known gaps in wrong-way risk capture) that need P&L reserve treatment.
- XVA reserves: where a model has known limitations or market data is illiquid, Product Control is typically responsible for quantifying and holding an appropriate valuation reserve/adjustment, reviewed periodically.
6.5 Additional Points for Day-1 P&L
- The "Day-1 P&L" is the difference between the trade's model price (including XVA) and the "real" market price observed at inception. Since XVA models use unobservable inputs (e.g., credit curves, funding spreads), the output price is not "observable."
- The Control: if the model price is different from the price agreed with the counterparty, that difference cannot be booked as profit on Day 1. It must be deferred and amortized into P&L over the life of the trade, pending validation of the model. This is a cornerstone of the IFRS 13 fair value hierarchy.
- Product Controller's Role: Product Control must identify, quantify, and track any Day-1 P&L, ensuring it's amortized correctly and not just taken upfront by the trading desk. This is a huge control point.
End-to-End Case Study
Bring it all together with one running example, from trade inception to a P&L break investigation.
- Global Bank enters a 3-year uncollateralised (no CSA) interest rate swap with Regional Corp, notional $500 mn, Global Bank receives fixed.
- At inception: risk-free value = $0 (at-market swap). Counterparty credit spread (Regional Corp 3yr CDS-implied) = 150bps. Global Bank's own 3yr CDS-implied spread = 90bps. Own funding spread over OIS = 80bps. Recovery rate assumption (both sides) = 40%.
Six months later, interest rates have fallen. The swap is now in-the-money to Global Bank (they're receiving above-market fixed), with a risk-free mark-to-market of +$8 mn. At the same time, Regional Corp's credit spread has widened to 220bps on sector-wide stress, while Global Bank's own spread is unchanged at 90bps.
Step 1: Decompose the P&L Move
- Risk-free MTM moved from $0 to +$8 mn — this +$8 mn is pure rates P&L, unrelated to credit.
- Because the trade is now more in-the-money (higher positive exposure) AND Regional Corp's spread widened, CVA increases — both effects compound. Suppose the recomputed CVA is now −$0.9 mn (up from an inception CVA of roughly −$0.5 mn given the smaller exposure and tighter spread at that time).
- DVA and FVA also move, but more modestly, since Global Bank's own spread and funding spread are unchanged — any DVA/FVA movement here is driven purely by the change in exposure profile (ENE falls as the trade becomes more of an asset, less of a liability to Global Bank), not by a spread move.
Step 2: The Product Controller's Question
When the desk's daily P&L report shows a total swing of roughly +$7.6 mn, the Product Controller does not simply accept the number — they reconcile it to its components: +$8.0 mn rates MTM, −$0.4 mn incremental CVA (the widening of the CVA charge), plus small DVA/FVA movements netting close to zero. This decomposition is exactly what gets challenged in month-end P&L attribution and what an interviewer will probe: "Walk me through why CVA got worse even though nothing happened to our own credit."
Step 3: The IPV Challenge
Independent Price Verification for that month would specifically check: is the 220bps Regional Corp spread genuinely observed (or an appropriate proxy, if Regional Corp has no liquid CDS), and does the exposure profile used reflect the swap's actual current in-the-money position rather than a stale simulation run from a prior valuation date? A Product Controller catching a stale exposure profile here — one of the single most common real-world XVA control breaks — is exactly the kind of judgement this session is meant to build.
Quick-Reference Cheat Sheet & Interview Questions
One-Line Definitions
| Term | One-Line Definition |
|---|---|
| CVA | Cost of counterparty defaulting on money owed to you — a debit to value. |
| DVA | "Benefit" of you defaulting on money you owe — a credit to value, tied to your own credit spread. |
| FVA | Cost/benefit of funding an uncollateralised position at your own funding rate, not risk-free. |
| EPE / ENE | Expected Positive / Negative Exposure — the exposure profiles that feed CVA and DVA respectively. |
| Wrong-Way Risk | When exposure and counterparty default probability are adversely correlated, worsening CVA beyond the simple formula. |
| MVA | Funding cost of posting Initial Margin. |
| KVA | Cost of the regulatory capital the trade consumes over its life. |
Questions Commonly Asked in Product Control / BA Interviews
1. The DVA paradox
Why can a bank's own credit downgrade result in a P&L gain? (Tests DVA understanding and its accounting controversy.)
2. CVA vs FVA overlap
Walk me through the difference between CVA and FVA — aren't they double-counting the same thing? (Tests understanding of the historical FVA/DVA overlap debate.)
3. Collateral and CVA
How does a CSA (Credit Support Annex) reduce CVA? (Tests understanding of collateral and netting.)
4. No traded CDS
If a counterparty has no traded CDS, how would you estimate their default probability for CVA? (Tests knowledge of proxy/sector curve methodology — relevant to real-world IPV.)
5. Wrong-way risk in practice
What is wrong-way risk, and can you give a real-world example? (Tests conceptual depth beyond the formula.)
6. Investigating a CVA jump
As a Product Controller, how would you investigate a sudden unexplained jump in CVA on a desk's book? (Tests practical control mindset — expects an answer covering curve checks, static data/netting set checks, and exposure profile checks.)
7. Accounting scope
Is FVA required under IFRS 13? Is DVA? (Tests awareness of the accounting vs. internal-pricing distinction.)
Every XVA answers one question: what does it really cost to trade with an entity that might default, and to fund that trade in the real world?
If you can hold onto that one sentence, you can reason your way through any XVA variant you haven't seen before — which is precisely the skill a Product Controller is expected to bring to the table every single day.