The fundamental representation of pricing adjustments

Benedict Burnett, Ryan McCrickerd, Benjamin Piau · 2025

Read The fundamental representation of pricing adjustments by Benedict Burnett, Ryan McCrickerd, Benjamin Piau (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2503-14997.txt.

Introduction

Benedict Burnett, Ryan McCrickerd, and Benjamin Piau present a finance paper from 2025 on pricing adjustments for derivatives. Traders often need to adjust a price for costs, risks, and funding terms. Past work gave many separate formulas for these adjustments, known as XVA, over the past twenty years.

The authors give one broad formula that covers the gap between any two pricing rules. They show how older results, from a well-known option adjustment to bank-style methods, fit inside it. They then use the formula to gauge model risk and to estimate a key credit risk case. Readers get a single frame that ties a scattered field together and supports risk checks.

Catalog record

CREATORBenedict Burnett, Ryan McCrickerd, Benjamin Piau
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE53 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2503-14997

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

The fundamental representation of pricing adjustments Thanks: Disclaimer: This article represents the views of the authors alone and not the views of any firm or institution. Benedict Burnett, Ryan McCrickerd, Benjamin Piau Abstract This article consolidates and extends past work on derivative pricing adjustments, including XVA, by providing an encapsulating representation of the adjustment between any two derivative pricing functions, within an Itô SDE/parabolic PDE framework. We give examples of this representation encapsulating others from the past 20 years, ranging from a well known option pricing adjustment introduced by Gatheral, to the collection of semi-replication XVA originating from Burgard & Kjaer. We show that this fundamental representation can be applied to quantify and mitigate XVA model risk, providing a novel approach to estimating CVA wrong-way risk as an example application. Keywords: Derivatives, model risk, pricing, P&L, volatility, XVA MSC2020: 91G20, 91G30, 91G40 1 Introduction There has been a large amount of work in the literature on adjustments to ‘naive’ derivative prices, ranging from early work presented in e.g. Gatheral (2006) , through Piterbarg (2010) , Burgard and Kjaer (2013) and Green et al. (2014) to, more recently, work such as Burnett (2021) and Kenyon et al. (2023) . These articles have demonstrated the possibilities that are opened by accommodating sophisticated effects (be it volatilities, funding, hedging or climate costs) via adjustments to well-understood base calculations. What has perhaps not been explicitly addressed is the relationship between all of these adjustments, placing them within a unified view of the landscape. That is the primary aim of the present article. We will see that valuation adjustments can generally be described in terms of three types of contribution: model adjustments, discounting adjustments, and payoff adjustments. Figure 1 shows this conceptual map, highlighting the locations of some key cases from the literature. M odel adjustments \hat{{\mathcal{L}}}\neq{\mathcal{L}} D iscounting adjustments \hat{R}\neq R P ayoff adjustments \hat{F}\neq F \circ \circ \circ Gatheral (2006) Piterbarg (2010) Burgard and Kjaer (2013) Figure 1: The pricing adjustments universe as suggested by ( 1.1 ). Citations show where historic examples live, analysed in Section 3 . The location of Piterbarg (2010) here refers strictly to the fully-collateralised case. This theoretical unification has a practical purpose, which leads to the secondary aim of this article: it suggests a new and general way to estimate the impact of model changes (such as adding a new stochastic factor, or changing the assumed dynamics in another way). This approach can be applied to cases including XVAs, where such impacts can be particularly relevant (see Bénézet and Crépey (2024) ). Since XVAs are already adjustments to underlying valuations, in the below, we term any adjustments on XVAs themselves ‘meta-adjustments’. The results of this investigation have multiple potential beneficiaries. The formulae we give for general adjustments show a way to estimate future P&L bleed (important for traders, finance functions and senior management); equivalently, they suggest ways to approximate the inaccuracy in a simple calculation due to neglected effects (important for model validation and valuation control). 1.1 Structure of the article In Section 2 we present a fundamental representation of pricing adjustments. This representation reveals a natural decomposition of the pricing adjustments universe, in terms of three main categories (M, D and P) relating to models, discounting, and payoffs. 1 1 1 This set of effects is natural given the ingredients in basic pricing rules, e.g. {\mathbb{E}}[e^{-rT}G(X_{T})] . Notice the model X_{t} , discounting at rate r , and payoff G , which all stand to be interrogated. 2 2 2 We thank an anonymous reviewer for the suggestion that an alternative view would combine D and P effects into ‘cashflows’, in contrast to dynamics (M); it is possible that this approach might yield a more frequently distinct separation of adjustments than the granular approach we take here. Considering a desirable target price \hat{V} in terms of a convenient base price V plus some adjustment U , this categorisation derives from the associated P&L bleed {Z} on which U depends. Summarising what we will show in Section 2 , U={\mathbb{E}}\left[\int_{0}^{T}e^{-\int_{0}^{t}\hat{R}_{s}ds}{Z}_{t}\,dt\right],\quad{Z}=\underbrace{(\hat{{\mathcal{L}}}-{\mathcal{L}})}_{\neq 0\Leftrightarrow{\mathrm{\textbf{M}}}}V-\underbrace{(\hat{R}-R)}_{\neq 0\Leftrightarrow{\mathrm{\textbf{D}}}}V+\underbrace{(\hat{F}-F)}_{\neq 0\Leftrightarrow{\mathrm{\textbf{P}}}}. (1.1) As described in Table 1 , {\mathcal{L}} , R and F are the (naive) Itô generator of a process X_{t} , discount rate, and payoff respectively, all reflected in the base price V . Terms with hats are counterparts from \hat{V} . Although in principle the adjustments covered here might span multiple teams in financial institutions, 3 3 3 A simple model (e.g. convexity) adjustment could be handled directly by an origination desk, a payoff adjustment (e.g. CVA) could be handled by the XVA desk, etc. for concreteness we will henceforth imagine that a bank sells a derivative to a client at the target price \hat{V} , then distributes its risk management between an origination desk handling V and XVA desk handling U . But this should be taken with a pinch of salt, i.e. idealised abstraction. Moving to XVA model risk in Section 4 , U plays the role of V and we call A=\hat{U}-U a meta-adjustment. 4 4 4 Although we explicitly consider only one level of meta-adjustment in this article, the approach can in principle be taken to higher orders, with adjustments on adjustments on adjustments (and so on) without needing any amendments to the theory presented here. We thank an anonymous reviewer for pointing this out. Due to the extreme difficulty of computing these accurately in many cases, the theory developed here serves as a robust framework for constructing signals for adjustments’ magnitude and direction. 5 5 5 In 2001 Gatheral wrote it is no surprise that this approach doesn’t yield any easy results , essentially regarding U in ( 1.2 ). Easy would certainly be an understatement for meta-adjustments, depending on future XVA greeks, but this does not preclude valuable results. Based around expected P&L, these can be vital towards banks understanding risk today , given model upgrades can take significant time. We show that the general representation derived in Section 2 provides a surprisingly effective tool to approximate hard-to-calculate meta-adjustments, demonstrating its power on the classic case of estimating wrong-way risk on a CVA. 1.2 A starting example The region M of ‘model adjustments’ in Figure 1 is frequently the most challenging, being driven by changes to the model dynamics, so it serves to consider an example from Gatheral (2006) . In his equation (3.4), Gatheral writes a call option price \hat{V} on a stock S in terms of a Black-Scholes price V plus some adjustment U : {\hat{V}}=V+U,\quad U={\mathbb{E}}\left[\int_{0}^{T}\!{Z}_{t}\,dt\right],\quad{Z}=\frac{1}{2}\left(\hat{\alpha}^{2}-\alpha^{2}\right)S^{2}\partial_{SS}V. (1.2) The variable \int_{0}^{T}{Z}_{t}\,dt here represents the (pathwise) P&L after selling (and delta hedging) the option using the Black-Scholes model with volatility \alpha , while S actually evolves with volatility \hat{\alpha} . Its representation in terms of the option gamma \partial_{SS}V is often referred to as the fundamental theorem of derivative trading . 6 6 6 See Ellersgaard et al. (2017) for the history going back to El Karoui et al. (1998) , with the experiments of Ahmad and Wilmott (2005) deserving special mention. The associated pricing adjustment U was, to the best of our knowledge, first presented in Gatheral (2001) , then refined in Lee (2005) . (Comparing with ( 1.1 ), the presence of only

Read the complete text