---
title: "The Valuation-Adjustment Desk"
book: "Rates, Credit, XVA and Risk"
subject: quant
language: en
chapter: 20
exercises: 8
source: https://one-course.com/books/quant/6/en/chapter/20-the-valuation-adjustment-desk
---

# Chapter 20 — The Valuation-Adjustment Desk

A salesperson prices a twenty-year swap for an unrated industrial company that wants to fix the rate on a long loan. The rates desk quotes the par rate in seconds. Then the valuation-adjustment desk adds its charge: the company’s credit, the bank’s funding of the uncollateralised position, the capital the trade will tie up for two decades. In this chapter’s example the charge comes to nearly forty basis points a year, more than the bid–offer spread of the swap itself, and the client may walk to a bank with a cheaper view of any of the three. Chapters 17 to 19 built the adjustments; this chapter is about the desk that owns them: why it is central, how it prices a new trade against everything the bank already has with the client, what it can hedge and what it cannot, and how its charges reach the trading desks.

## 20.1 Why one central desk

**Definition 20.1 (XVA desk).**

An *XVA desk* is the trading desk that owns a bank’s valuation adjustments: it charges each new trade its adjustments, holds the resulting counterparty, funding and capital risks on its book, hedges what can be hedged, and reports the adjustments’ changes as its P&L.

Adjustments are properties of [netting sets](https://one-course.com/books/quant/6/en/chapter/17-counterparty-exposure#def-rc-counterparty-exposure-netting), not of trades: a swap sold by the rates desk and an FX forward sold by the FX desk to the same client net against each other, share the client’s collateral and its default. Only a desk that sees the whole [netting set](https://one-course.com/books/quant/6/en/chapter/17-counterparty-exposure#def-rc-counterparty-exposure-netting) can price a trade’s effect on it. Centralising also concentrates the expertise and the hedging, which needs default swaps, index protection and market hedges across asset classes ([Figure 20.1](#fig-rc-the-valuation-adjustment-desk-org)).

![The XVA desk between the trading desks and the markets: desks pass it the counterparty, funding and capital risks of their trades and pay a charge; it hedges credit and market risk in the market and settles funding and capital with the treasury. Schematic.](https://one-course.com/images/onecourse/chapters/quant-6/rc-the-valuation-adjustment-desk/fig-ac2986071833.svg)

***Figure 20.1.** The [XVA desk](#def-rc-the-valuation-adjustment-desk-desk) between the trading desks and the markets: desks pass it the counterparty, funding and capital risks of their trades and pay a charge; it hedges credit and market risk in the market and settles funding and capital with the treasury. Schematic.*

## 20.2 Incremental pricing and allocation

**Definition 20.2 (Incremental XVA).**

The *incremental XVA* of a new trade is the change in the [netting set](https://one-course.com/books/quant/6/en/chapter/17-counterparty-exposure#def-rc-counterparty-exposure-netting)’s (or the portfolio’s) adjustments when the trade is added: $\mathrm{XVA}(\text{set}+\text{trade})-\mathrm{XVA}(\text{set})$. It is what the trade costs the bank, and it can be negative when the trade offsets existing exposure.

**Example 20.3 (Two new swaps in one netting set).**

Chapter 17’s [netting set](https://one-course.com/books/quant/6/en/chapter/17-counterparty-exposure#def-rc-counterparty-exposure-netting) with chapter 18’s counterparty has a CVA of USD 891 342. A new five-year USD 50 million swap on which the bank pays fixed has a standalone CVA of 37 061, but adding it lowers the set’s CVA to 869 582: its incremental CVA is $-\text{USD}~21\,761$, a rebate. The same swap with the bank receiving fixed has a standalone CVA of 30 742 and an incremental CVA of $+22\,784$.

```python
def euler_cva(values: Sequence[np.ndarray], D: np.ndarray, times: np.ndarray, cpty: HazardCurve,
              recovery: float = 0.4) -> list[float]:
    """Euler contributions of each trade to the netting set's CVA (they add up to the total)."""
    V = np.sum(values, axis=0)
    ind = (V > 0).astype(float)
    q = np.array([cpty.survival(float(t)) for t in times])
    dq = q[:-1] - q[1:]
    out = []
    for v in values:
        c = (D * v * ind).mean(axis=0)
        out.append(float((1.0 - recovery) * np.sum(0.5 * (c[1:] + c[:-1]) * dq)))
    return out
```

***Listing 20.1.** Euler allocation of a netting set’s CVA to its trades. code/firm/xvaquote/firm_xvaquote.py*

Incremental charges depend on the order of arrival: the first trade with a client pays the full standalone charge, a later offsetting one earns a rebate. To report each trade’s share of an existing adjustment, the desk needs an allocation that adds up.

**Definition 20.4 (Euler allocation).**

*Euler allocation* assigns to trade $i$ the contribution $a_i = \frac{d}{d\epsilon}\mathrm{XVA}(\ldots,(1+\epsilon)V_i,\ldots)|_{\epsilon=0}$. When the adjustment is homogeneous of degree one in the trades’ sizes, as CVA is, Euler’s theorem makes the contributions add up to the total; for CVA, $a_i = (1-R)\sum_k\E[D(0,t_k)V_i(t_k)\mathbf 1_{V(t_k)>0}]\,\Delta\mathrm{PD}_k$.

**Example 20.5 (Allocating the netting set).**

Of the set’s USD 891 342 of CVA, [Euler allocation](#def-rc-the-valuation-adjustment-desk-euler) assigns 160 167 to the swap and 731 175 to the cross-currency swap, against standalone CVAs of 288 250 and 768 334 ([Figure 20.2](#fig-rc-the-valuation-adjustment-desk-euler)). The swap, which offsets part of the cross-currency exposure, receives most of the netting benefit.

![Standalone CVA of each trade against its Euler contribution to the netting set’s CVA. The contributions add up to the set’s CVA; the difference from the standalone figures is the netting benefit, shared in proportion to each trade’s role in the exposure. Data: the chapter’s tutorial.](https://one-course.com/images/onecourse/chapters/quant-6/rc-the-valuation-adjustment-desk/fig-0a05e88aed90.svg)

***Figure 20.2.** Standalone CVA of each trade against its Euler contribution to the [netting set](https://one-course.com/books/quant/6/en/chapter/17-counterparty-exposure#def-rc-counterparty-exposure-netting)’s CVA. The contributions add up to the set’s CVA; the difference from the standalone figures is the netting benefit, shared in proportion to each trade’s role in the exposure. Data: the chapter’s tutorial.*

## 20.3 What it hedges and what it cannot

CVA has two kinds of risk: the market factors that move the exposure, hedged with swaps, options and FX like any derivative book, and the counterparty’s credit, hedged with credit instruments. Many counterparties have no traded default swaps: the desk needs a spread to price them and a proxy to hedge them.

**Definition 20.6 (Proxy credit spread).**

A *proxy credit spread* is a spread assigned to a counterparty without liquid credit quotes, estimated from liquid peers that resemble it in credit quality, sector and region, or mapped to a single related name. In the European Union a proxy used for CVA capital must consider all three attributes, with at least three industry categories (public, financial, other) and four regions (Europe, North America, Asia, rest of the world).

**Example 20.7 (A proxy for the unrated client).**

Regress the logarithm of eighteen illustrative ten-year peer spreads on rating (A, BBB, BB), sector (industrials, utilities, consumer) and region (US, EU) indicators. The fit’s error is 1.3% of the spread, and for the client, mapped internally to BB, industrials, US, it predicts 337.5 basis points at ten years ([Figure 20.3](#fig-rc-the-valuation-adjustment-desk-proxy)). Applying chapter 18’s term-structure shape gives a curve from 154 basis points at one year to 338 at ten.

![Observed against fitted ten-year spreads of the illustrative peers in the proxy regression on rating, sector and region. The points sit close to the diagonal; the client’s proxy is the model’s prediction for its cell (BB, industrials, US). Data: the chapter’s tutorial.](https://one-course.com/images/onecourse/chapters/quant-6/rc-the-valuation-adjustment-desk/fig-b855acc1f2f8.svg)

***Figure 20.3.** Observed against fitted ten-year spreads of the illustrative peers in the proxy regression on rating, sector and region. The points sit close to the diagonal; the client’s proxy is the model’s prediction for its cell (BB, industrials, US). Data: the chapter’s tutorial.*

**Definition 20.8 (Contingent credit default swap).**

A *contingent credit default swap* pays, on the default of a reference entity, the [loss given default](https://one-course.com/books/quant/6/en/chapter/13-reduced-form-credit#def-rc-reduced-form-credit-pd) on the then value of a specified derivative (or [netting set](https://one-course.com/books/quant/6/en/chapter/17-counterparty-exposure#def-rc-counterparty-exposure-netting)) instead of on a fixed notional: exact protection for the CVA of that derivative, and priced by its CVA.

A proxy hedge leaves basis risk (the client’s credit does not move with its peers), [jump-to-default risk](https://one-course.com/books/quant/6/en/chapter/13-reduced-form-credit#def-rc-reduced-form-credit-jtd) (index protection does not pay when the client defaults alone) and [cross-gamma](https://one-course.com/books/quant/6/en/chapter/3-rates-risk#def-rc-rates-risk-crossgamma) (exposure and credit move together). Funding and capital charges have no market hedge at all: the desk manages them by collateral, compression, novation and the choice of counterparties.

**As of September 2026 — Proxy spreads and eligible hedges.**

The Basel Committee’s CVA framework (July 2020) requires proxy spreads for illiquid counterparties to be estimated from liquid peers by an algorithm that discriminates at least on credit quality, industry and region, allows mapping to one related name (a municipality to its country) only with supervisory justification, and recognises as CVA hedges only single-name default swaps, single-name contingent default swaps and index default swaps.

## 20.4 Charging, transfer pricing and the desk’s P&L

**Definition 20.9 (XVA charge).**

The *XVA charge* is the amount the [XVA desk](#def-rc-the-valuation-adjustment-desk-desk) takes from the originating desk when a trade is booked: the trade’s incremental adjustments, paid upfront or converted into a running spread on the trade, $\text{charge} = \mathrm{XVA}/(N\cdot A)$ with $A$ the trade’s annuity.

**Example 20.10 (The twenty-year quote).**

The client receives floating and pays the par rate of 4.11% on USD 100 million for twenty years; no CSA. On 4 000 paths the bank’s [expected exposure](https://one-course.com/books/quant/6/en/chapter/17-counterparty-exposure#def-rc-counterparty-exposure-profiles) peaks at USD 5.65 million after five years ([Figure 20.4](#fig-rc-the-valuation-adjustment-desk-ee)). With the proxy curve, the CVA is USD 1 649 494; funding at 80 basis points costs 343 872; with capital charged at a 10% hurdle (SA-CCR at a 100% risk weight, and CVA capital at the 7% risk weight for unrated industrials), the KVA is 3 141 826. Over the swap’s annuity of 13.67, that is 12.1, 2.5 and 23.0 basis points a year: 37.6 in all.

![Expected exposure of the twenty-year swap on which the bank receives fixed from the unrated client: it peaks after about five years and runs off over the remaining fifteen. Data: the chapter’s tutorial.](https://one-course.com/images/onecourse/chapters/quant-6/rc-the-valuation-adjustment-desk/fig-9972e7c74138.svg)

***Figure 20.4.** [Expected exposure](https://one-course.com/books/quant/6/en/chapter/17-counterparty-exposure#def-rc-counterparty-exposure-profiles) of the twenty-year swap on which the bank receives fixed from the unrated client: it peaks after about five years and runs off over the remaining fifteen. Data: the chapter’s tutorial.*

The desk’s P&L is the change in the adjustments it holds, net of its hedges and of the charges it receives. Charges set at inception and adjustments marked daily diverge: spreads move, exposures grow, the bank’s own funding cost changes. A desk that charged too little loses slowly; one that charges too much loses the trades.

## 20.5 Organisation

Where the desk sits decides what it optimises. Inside the markets division it prices and hedges like a trading desk; inside treasury or finance it is closer to a utility setting transfer prices. Either way it needs the exposure engine of chapter 17 on common scenarios for every [netting set](https://one-course.com/books/quant/6/en/chapter/17-counterparty-exposure#def-rc-counterparty-exposure-netting), legal data on netting and collateral agreements, credit data on every counterparty, and a capital calculator, all fast enough to answer a salesperson in seconds: the risk engine of chapter 29.

## 20.6 Tutorial: pricing a new trade

**Goal.** Compute incremental and allocated CVA in an existing [netting set](https://one-course.com/books/quant/6/en/chapter/17-counterparty-exposure#def-rc-counterparty-exposure-netting), build a proxy spread, and quote the adjustments of a new long-dated swap as running basis points. **End state:** the numbers of Examples [20.3](#ex-rc-the-valuation-adjustment-desk-incremental), [20.5](#ex-rc-the-valuation-adjustment-desk-euler), [20.7](#ex-rc-the-valuation-adjustment-desk-proxy) and [20.10](#ex-rc-the-valuation-adjustment-desk-quote) and the charts.

1. **Incremental** : `incremental_table()` , standalone against incremental.
2. **Allocation** : `allocation()` ; check that the contributions add up.
3. **Proxy** : `proxy()` and `proxy_curve()` .
4. **Quote** : `twenty_year_quote()` ; `fig_rc_xvadesk.py` writes the charts.

**What to change next.** Map the client to BBB instead and see the quote move; add a CSA and recompute every component.

## 20.7 Build: the quote service

**Purpose.** The firm’s [XVA](https://one-course.com/books/quant/6/en/chapter/18-credit-and-debit-valuation-adjustments#def-rc-credit-and-debit-valuation-adjustments-xva) quotes: incremental adjustments of a new trade against the existing [netting set](https://one-course.com/books/quant/6/en/chapter/17-counterparty-exposure#def-rc-counterparty-exposure-netting), allocations for reporting, proxy spreads, and running charges.

**Interface.** `netting_cva(values, D, times, cpty, recovery)`; `incremental(adjustment, existing, new)`; `euler_cva`; `proxy_spread(peers, target)`; `running_charge(upfront, notional, annuity)`. Reads `firm_exposure`, `firm_cva`, `firm_xvafund`.

**Rules.** All adjustments on the same scenarios as the existing [netting set](https://one-course.com/books/quant/6/en/chapter/17-counterparty-exposure#def-rc-counterparty-exposure-netting); proxies regress log spreads on credit quality, sector and region.

**Acceptance tests.** `code/firm/xvaquote/tests/`: Euler contributions add up and scale with the trade; the incremental identity; the proxy regression recovers a multiplicative table exactly; the running-charge conversion.

**Stretch.** Incremental FVA and KVA on the portfolio; allocation of capital by Euler on the SA-CCR formula; a proxy with liquidity and seniority factors; sub-second incremental runs by storing path values.

Sources and further reading

- Basel Committee on Banking Supervision, *Targeted revisions to the credit valuation adjustment risk framework* , July 2020, MAR50.

## 20.8 Exercises

**Exercise 20.1 ★.**

Why can a trade’s incremental CVA be negative while its standalone CVA is positive?

**Solution of Exercise 20.1.**

Netting: the new trade’s values offset the existing set’s on many paths, so the netted positive exposure falls. Standalone, the trade has its own positive exposure; in the set, it removes more exposure than it adds.

**Exercise 20.2 ★.**

Convert an upfront charge of USD 1 million into running basis points on a USD 100 million twenty-year swap with the chapter’s annuity.

**Solution of Exercise 20.2.**

$1\,000\,000/(100\,000\,000\times13.67) = 7.3$ basis points a year.

**Exercise 20.3 ★.**

What are the three variables a proxy spread must discriminate on under the Basel CVA framework?

**Solution of Exercise 20.3.**

A measure of credit quality (such as a rating), industry, and region.

**Exercise 20.4 ★★.**

Show that the Euler contributions to CVA add up to the [netting set](https://one-course.com/books/quant/6/en/chapter/17-counterparty-exposure#def-rc-counterparty-exposure-netting)’s CVA.

**Solution of Exercise 20.4.**

$\mathrm{CVA} = (1-R)\sum_k\E[D\max(\sum_iV_i,0)]\Delta\mathrm{PD}_k$ and $\max(\sum_iV_i,0) = \sum_iV_i\mathbf 1_{\sum_jV_j>0}$ path by path; take expectations and sum over $i$: the contributions $a_i$ add up. Equivalently, CVA is homogeneous of degree one in the sizes and Euler’s theorem applies.

**Exercise 20.5 ★★.**

Why does the order of arrival of trades matter for incremental charges, and why not for [Euler allocations](#def-rc-the-valuation-adjustment-desk-euler)?

**Solution of Exercise 20.5.**

The incremental charge of a trade depends on what is already in the set, so two offsetting trades are charged differently depending on which came first. Euler contributions are computed on the whole set at once and are symmetric in the trades, so they do not depend on history.

**Exercise 20.6 ★★.**

Which risks does index protection leave open when it hedges the CVA of an unrated client?

**Solution of Exercise 20.6.**

Basis between the client’s spread and the index, jump-to-default of the client (index protection pays little when one name defaults), [wrong-way risk](https://one-course.com/books/quant/6/en/chapter/17-counterparty-exposure#def-rc-counterparty-exposure-wwr), and the changes of the exposure itself.

**Exercise 20.7 ★★★.**

*Coding.* Compute the [CS01](https://one-course.com/books/quant/6/en/chapter/13-reduced-form-credit#def-rc-reduced-form-credit-cs01) of the twenty-year swap’s CVA on the proxy curve and the notional of ten-year protection on a peer that hedges it.

**Solution of Exercise 20.7.**

USD 2 385 per basis point; divided by the peer’s ten-year [risky annuity](https://one-course.com/books/quant/6/en/chapter/13-reduced-form-credit#def-rc-reduced-form-credit-risky) and a basis point, about USD 3.55 million of protection.

**Exercise 20.8 ★★★.**

*Find the flaw.* “Our [XVA charge](#def-rc-the-valuation-adjustment-desk-charge) covers the expected loss on the client, so the [XVA desk](#def-rc-the-valuation-adjustment-desk-desk)’s P&L should be flat over the life of the trade.”

**Solution of Exercise 20.8.**

The charge is the adjustment’s value at inception; afterwards the desk’s P&L moves with the client’s spread, the exposure’s drivers, funding and capital costs, and any unhedged gap. Only a perfect hedge of every component would keep it flat, and funding and capital cannot be hedged.

## 20.9 Problem: The Unrated Corporate

**Problem 20.1.**

Weekend problem — the all-in add-on

The company of [Example 20.10](#ex-rc-the-valuation-adjustment-desk-quote) asks for the twenty-year swap. Another bank has quoted it 20 basis points below this bank’s all-in rate.

**Part I — The quote.**

1. Give the swap’s par rate and its annuity.
2. Give the proxy spread at ten years and the internal mapping behind it.
3. Give CVA, FCA and KVA upfront and in running basis points.
4. Give the all-in add-on.
5. Which component dominates, and why for this client?

**Part II — Hedging.**

6. Which part of the add-on can the desk hedge in the market, as a share of the total?
7. Give the CVA’s [CS01](https://one-course.com/books/quant/6/en/chapter/13-reduced-form-credit#def-rc-reduced-form-credit-cs01) and the hedge in ten-year peer protection.
8. What does the hedge leave open?
9. Would a contingent default swap be available, and what would it cost?
10. How would the desk hedge the exposure’s rate risk?

**Part III — The competition.**

11. Which assumptions could make the other bank 20 basis points cheaper?
12. Map the client to BBB instead: what would change?
13. Which terms could the bank offer to close the gap?
14. Why might the other bank be wrong?
15. Is walking away a loss for the bank?

**Part IV — Judgement.**

16. Who should bear the KVA: the rates desk, the [XVA desk](#def-rc-the-valuation-adjustment-desk-desk) or the bank?
17. How should the charge be split with the salesperson’s revenue?
18. What would you tell the client about the charge?
19. State the *named result* : the all-in add-on in running basis points and the part the desk can hedge.
20. In one sentence: what does an [XVA desk](#def-rc-the-valuation-adjustment-desk-desk) sell to the rest of the bank?

**Solution of Problem 20.1.**

**1.** 4.11%; an annuity of 13.67. **2.** 337.5 basis points, from the peer regression with the client mapped to BB, industrials, US. **3.** CVA USD 1 649 494 (12.1 basis points), FCA 343 872 (2.5), KVA 3 141 826 (23.0). **4.** 37.6 basis points a year. **5.** KVA: an unrated client gets the 7% CVA risk weight and a twenty-year swap ties up capital for decades. **6.** The CVA, 32% of the total; funding and capital cannot be hedged in the market. **7.** USD 2 385 per basis point; about USD 3.55 million of ten-year peer protection. **8.** Basis to the client’s own credit, its jump-to-default, the exposure’s growth beyond ten years, and [cross-gamma](https://one-course.com/books/quant/6/en/chapter/3-rates-risk#def-rc-rates-risk-crossgamma). **9.** Rarely for an unrated corporate; if available it would cost about the CVA plus the seller’s margin and capital. **10.** With swaps and swaptions sized on the CVA’s rate delta and vega, on common scenarios. **11.** A better internal rating, a lower funding spread, a lower [hurdle rate](https://one-course.com/books/quant/6/en/chapter/19-funding-margin-and-capital-adjustments#def-rc-funding-margin-and-capital-adjustments-kva), a different CVA capital approach, or pricing only the incremental effect on a large existing relationship. **12.** With BBB and the 3% risk weight the add-on falls to 26.2 basis points (CVA 7.2, FCA 3.1, KVA 15.8). **13.** A CSA, a break clause at ten years, a shorter swap rolled over, or collateral. **14.** It may be mispricing the client’s credit or not charging capital, and it will collect the clients whose risk it underprices. **15.** Only if the trade would have paid for its risks; walking away from an underpriced trade avoids a loss. **16.** The bank sets the hurdle; the [XVA desk](#def-rc-the-valuation-adjustment-desk-desk) should charge it to the originating desk so the trade’s economics are visible where the decision is made. **17.** Sales credit should be net of the [XVA charge](#def-rc-the-valuation-adjustment-desk-charge), so that the salesperson is not paid for revenue that the bank’s risks consume. **18.** That the rate includes the cost of its credit, of funding an unsecured position and of the capital the bank holds, and that a CSA or break clause would lower it. **19.** Named result: *the unrated corporate*: an all-in add-on of 37.6 basis points a year, of which the desk can hedge the CVA’s 12.1 basis points (32%) in the market. **20.** The management of counterparty, funding and capital costs that no single desk can see or hedge alone.

## 20.10 Interview questions

**Interview question 20.1 ★ trader, bank.**

Why do banks centralise valuation adjustments in one desk?

**Solution of Interview question 20.1.**

Adjustments are properties of [netting sets](https://one-course.com/books/quant/6/en/chapter/17-counterparty-exposure#def-rc-counterparty-exposure-netting) spanning desks; only a central view can price netting and collateral, avoid double counting, hedge credit centrally, and manage funding and capital consistently.

*What the interviewer is looking for: [netting sets](https://one-course.com/books/quant/6/en/chapter/17-counterparty-exposure#def-rc-counterparty-exposure-netting) across desks and central hedging.*

**Interview question 20.2 ★★ researcher.**

What is [Euler allocation](#def-rc-the-valuation-adjustment-desk-euler), and when does it add up?

**Solution of Interview question 20.2.**

Each trade’s contribution is the derivative of the total with respect to scaling that trade; for a function homogeneous of degree one, contributions add up to the total (Euler’s theorem). CVA with fixed credit is homogeneous; capital formulas with floors or caps may not be.

*What the interviewer is looking for: the derivative definition and the homogeneity condition.*

**Interview question 20.3 ★★ trader.**

How do you price the CVA of a client with no traded credit?

**Solution of Interview question 20.3.**

Build a proxy curve from liquid peers by rating, sector and region (regression or buckets), or map to a related name; validate against bonds or loans if any; hedge with index or peer protection and hold reserves for the basis.

*What the interviewer is looking for: proxy construction and residual basis.*

**Interview question 20.4 ★★ developer.**

A salesperson needs an [incremental XVA](#def-rc-the-valuation-adjustment-desk-incremental) quote in two seconds. How do you design the service?

**Solution of Interview question 20.4.**

Keep each [netting set](https://one-course.com/books/quant/6/en/chapter/17-counterparty-exposure#def-rc-counterparty-exposure-netting)’s path values on the common scenarios in memory; price the new trade on the same paths (closed forms or precomputed grids); recompute the netted profiles and adjustments incrementally; parallelise; fall back to approximations for exotic trades.

*What the interviewer is looking for: stored paths and incremental recomputation.*

**Interview question 20.5 ★★★ trader, risk.**

What drives the [XVA desk](#def-rc-the-valuation-adjustment-desk-desk)’s P&L, and what can go wrong in a crisis?

**Solution of Interview question 20.5.**

Changes in counterparty spreads and exposures net of hedges, funding costs, capital, new-trade charges and hedge slippage. In a crisis spreads and exposures jump together ([wrong-way risk](https://one-course.com/books/quant/6/en/chapter/17-counterparty-exposure#def-rc-counterparty-exposure-wwr)), proxies decouple from clients, hedges become expensive, and CVA capital hedges can themselves be charged.

*What the interviewer is looking for: the P&L drivers and crisis dynamics.*

**Interview question 20.6 ★★★ bank.**

Should [XVA charges](#def-rc-the-valuation-adjustment-desk-charge) be passed to clients in full? Discuss competition and adverse selection.

**Solution of Interview question 20.6.**

In full, the bank loses price-sensitive clients to banks with cheaper assumptions and keeps those whose risk others price higher (adverse selection); in part, it subsidises trades. Banks pass most of it, compete on assumptions and relationship value, and must be consistent.

*What the interviewer is looking for: competition, adverse selection and consistency.*