A forum of industry leaders discusses the latest developments in valuation adjustments – known as XVAs – and the strategic, operational and technological challenges of derivatives valuation in today’s environment, including the key considerations for banks looking to move to a standardised approach and the XVA risks that will accompany the move away from Libor
- Gerard Frewen, Global XVA Product Manager, Bloomberg
- Nicki Rasmussen, Head of XVA Desk, Danske Bank Corporates and Institutions
- Stamatoula Matsoukis, Subject Matter Expert in XVA, and CCR Modelling and Regulation, Euclides Risk Solutions
- Stuart Nield, XVA Product Manager, IHS Markit
- Lucy Thomson Jones, XVA Totem, IHS Markit
- Julien Second, Managing Director, Head of XVA Trading, RBC Capital Markets
What progress have banks made in centralising XVA desks? What benefits does this bring?
Gerard Frewen, Bloomberg: More banks have been setting up centralised XVA desks in the past few years as institutions worldwide have become aware of the importance of appropriately quantifying and managing the costs to their derivatives business of counterparty credit, funding and margin. This trend has been fuelled by the realisation there is a significant overlap in the requirements for calculation of each of these costs, and that centralising the management of these risks across asset classes better captures the benefits of netting. Dealers have recognised that managing these risks centrally enables them to better price and hedge the XVAs, and to take advantage of risk mitigation techniques such as compression and margin optimisation.
In addition to becoming aware of the need to do this from a best practice perspective, accounting and regulatory changes – such as the accounting changes in certain Asian jurisdictions and the introduction of the non-cleared margin rules – have also contributed to this trend. One major incentive has been the upcoming changes to the Basel III credit valuation adjustment (CVA) market capital requirements scheduled to take effect in 2022. Having a centralised XVA desk is one of the requirements for banks to be able to gain approval to use the more risk-sensitive standardised approach (SA‑CVA). As using this approach offers significant capital savings over the alternative basic approaches to CVA (BAs‑CVA), more banks continue to move in this direction.
Julien Second, RBC Capital Markets: XVA desk mandates will vary but will include all or a combination of the valuation adjustments: CVA, unsecured funding valuation adjustment (FVA), capital valuation adjustment (KVA), discounting and collateral valuation adjustments, margin valuation adjustments (MVA) and collateral management. These risks are connected in many ways, which is one of the main reasons for centralising them. They are exposure-based, portfolio-based and often computed using the same models and simulations.
Generally speaking, regrouping risks minimises the friction costs of these risks being managed in silos in the bank. Once centralised, these risks become easier to optimise, resulting in efficiency gains and a reduction in computation and management costs, and thus allowing the business to be more competitive.
Lucy Thomson Jones, IHS Markit: As the XVAs become more established and there is greater focus on accurately accounting for more adjustments than before, there has been a definite shift in centralising the responsibility for XVA and moving towards a central desk and product function, although progress can still vary significantly from bank to bank.
Having all XVAs in a single place means banks get a complete picture of their trading costs. Without this knowledge, it can be more difficult for banks to price deals, leading to potential profit bleed.
Nicki Rasmussen, Danske Bank Corporates and Institutions: A centralised XVA desk allows for the greatest degree of consistency in pricing and risk management across credit, funding, capital, collateral and margin. Consistency comes not only from how exposures are measured, modelled and viewed, but also from the treatment of netting, variation margin and initial margin (IM) agreements, which may cover different – and potentially non-overlapping – sets of trades. A centralised XVA desk is also the best position to potentially trade off various XVAs, supporting a more holistic view while still allowing a more strict profit-and-loss (P&L) and accounting treatment.
From an organisational point of view, a centralised XVA desk brings together and collaborates strongly with more decentralised parts of the business, such as the credit department, the collateral management department, the regulatory capital department and not least the legal department. Furthermore, centralised XVA ensures a concentration of the expertise required to understand and manage the complexities involved, which for most banks is the most efficient use of human resources. Finally, from a quant and IT point of view, building the XVA desk on a single model setup supports not only pricing but also daily P&L and risk management consistently, irrespective of the level of maturity the individual XVAs have achieved.
Stamatoula Matsoukis, Euclides Risk Solutions: XVA centralisation involves the aggregation of trades across different trading desks and lines of business. As part of the aggregation process, XVA desks mitigate counterparty credit risk (CCR) by applying netting and collateral.
What are the main challenges facing XVA desks today? How does that vary by region?
Nicki Rasmussen: One of the main challenges is the revised market risk framework – specifically the SA‑CVA. While it does extend the current CVA risk charge framework and aligns better with the risk management of the CVA desk by including the other market risk factors, it remains silent on the treatment of the other XVAs. This can be a substantial challenge for an XVA desk that runs active risk management beyond just CVA, and may be punishing such active risk management by the market risk capital model not properly recognising the risks being hedged for P&L management.
Stuart Nield, IHS Markit: On the business side, there is uncertainty on which XVA to pass onto the client and in what situations to do so. For the established adjustments, such as CVA and FVA, dispersion still exists and standardisation across the market is still some way off. Take FVA as an example: the difference in dispersion observed is at least partly attributed to region – although region often plays a smaller role in XVAs than one might expect. Part of the regional difference will also be due to the impact of local regulatory requirements. A bank’s internal policy is likely to play a far greater role in the size of XVAs, if not which XVAs are accounted for.
On the XVA system side, there are three key challenges:
1. Calculating all XVAs consistently so their interactions are taken into consideration
2. Delivering this information back to the trader fast enough for them to price
3. Performing these calculations cost-effectively.
Gerard Frewen: The scope of challenges varies by the size of market participant and region. In developed markets, centralised XVA desks have assumed increased responsibilities as a result of the regulatory changes under Basel III. Desks that previously had to deal with managing CVA and FVAs have now had their mandates expanded to include managing the costs of margin, capital and sometimes leverage ratio constraints. The overall impact of the market changes has been to significantly increase the overall costs of derivatives businesses. Hence, it is important to not just quantify these at the end of the day, but also appropriately price them pre‑trade.
The advent of MVA and KVA have increased the complexity of the set of XVA calculations required, and some participants have struggled to re‑engineer legacy systems to accommodate these new metrics.
Uncertainty around how the new capital requirements will be implemented in a number of jurisdictions also poses a challenge. It is not possible for XVA desks to accurately price-in future capital costs to new transactions when those costs are not clearly defined.
Conversations between regulators are still ongoing and rules have not been finalised. In the US, for example, the rule for the standardised approach to CCR (SA‑CCR) is not yet finalised, and the industry has raised concerns around the US‑specific calibration for specific asset classes such as commodities. It is similarly unclear in Europe when SA‑CCR will take effect. It is also unclear whether the existing exemptions for end-user clients will be retained in the final implementation of the CVA market capital. Regardless, the additional capital costs are expected to be substantial, which is why there has been a recent request from the International Swaps and Derivatives Association (Isda) for recalibration of the model. The uncertainty around whether this will happen makes it more difficult for desks to plan for the future.
In developing markets, the challenges are different. In some jurisdictions, CVA and debit valuation adjustment reporting requirements have only recently been introduced, and setting up desks to calculate XVA metrics is a new development. In Asia, for example, trading books tend to have significant concentrations of exotics and there are challenges in calculating XVA for these types of transactions. Across both Asia and Latin America, the lack of liquid credit derivatives markets for pricing and hedging is a major challenge, particularly in light of the forthcoming CVA market capital regime and the associated requirements to use market observables as inputs.
Julien Second: As XVAs contribute significantly to the all-in price of derivatives, XVA desks face the same competitive pressures as any other desk. Some Asian banks adopted CVA in 2019, there is a reasonable amount of consensus on CVA and FVA, but not around KVA, where discrepancies are evident. Regulation differences between jurisdictions and business decisions are usually the reason for this, and the role of XVA desks is to understand and respond to these challenges.
XVA books are, above all, macro books and are affected by the current market environment. This year has been volatile and, with the many ongoing sources of uncertainty – low and even negative rates, pressure on some lower-rated names and reduced single-name credit default swap (CDS) liquidity – there is little reason to expect this will change any time soon. Although they will vary in each region, risk positions are generally concentrated and directional by nature. This type of risk in illiquid and volatile markets can create negative feedback loops when hedging.
Stamatoula Matsoukis: At a global level, firms are currently facing challenges implementing the Fundamental Review of the Trading Book requirements for CVA, and the projected IM into XVA calculations.
Firms operating in the US can be subject to Comprehensive Capital Analysis and Review (CCAR) and regular model risk management examinations across the different business areas, including XVA desks. Due to XVA centralisation, the impact of these additional supervisory expectations on XVA desks is substantial. Added to this, the uncertainty of CCAR rules on dividend payments associated with the breach of expectations of the qualitative test can affect the amount of KVA associated with each trade.
In the post-Brexit era, implementation of the Basel capital requirements between the UK and the European Union – including requirements on CVA capital charge – will be different. Firms conducting business in these jurisdictions will need to adjust their XVA frameworks to comply with the new version of capital requirements and any other additional national and supranational regulations.
What are the key considerations for banks moving to the SA‑CVA model?
Stamatoula Matsoukis: I would summarise the key considerations as follows:
1. An additional degree of conservativeness in the calculations. This is because diversification benefits between delta and vega sensitivity risks are not permitted in the the SA‑CVA, but also due to only partial CVA hedging recognition and aggregation across the different buckets within every asset class.
2. Computational intensity of the calculation of CVA sensitivities to underlying market factors in the portfolios.
3. Calibration of exposure models.
4. Increased model governance as the regulatory expectations are very comprehensive, requiring a separate CVA desk, independent regular model validation, control unit and audit.
Stuart Nield: For banks not currently actively managing CVA risk, the big challenge will be setting up a CVA desk. For banks that already have a CVA desk, the challenge is incorporating the new capital measure into their pricing policy. Both face a technology challenge in calculating SA-CVA at trade inception and potentially incorporating this into KVA.
Gerard Frewen: The impact of moving to a SA‑CVA model varies based on where you are moving from. For large global banks that have been using the existing advanced approach, moving to the new SA‑CVA approach is a disadvantage. They have already invested in the necessary infrastructure and obtained the necessary model approvals to calculate capital under their internal models, but are now being forced to move to a more punitive standardised model that will incur higher capital costs. Some aspects of the finalised SA‑CVA model are problematic to the industry. For example, there are concerns around how CDS index hedges are handled and the conservativeness of the model compared with the way CVA is treated under accounting rules. We have accordingly seen recent industry attempts to engage with regulators to have changes made to the SA‑CVA model to address these concerns.
For other banks that have either used the standardised model previously or not been subject to CVA capital requirements at all, the considerations are different. Banks that exceed the $100 billion threshold will have to choose between using one of the BAs‑CVA and attempting to use the new SA‑CVA model. Using the SA‑CVA model will offer significant capital savings over using a BA-CVA approach, but there are challenges for some banks in being able to do that.
The final Basel III regulations set out a number of requirements for banks to be able to use the SA‑CVA approach. As previously mentioned, they need to set up a centralised XVA desk. Furthermore, they need to be able to calculate CVA sensitivities, deltas and vegas. They need to have a transparent methodology for marking proxy credit curves – and they need to get regulatory approval. None of these challenges are insurmountable, but they do require significant investment in infrastructure and often a significant reorientation in the banks’ approach to managing their derivatives businesses.
Which XVAs offer the greatest opportunity for competitive differentiation?
Stuart Nield: Most banks have converged on a common understanding of CVA. However, differences of opinion remain on FVA, MVA and KVA. Where there is a difference of opinion, there is an opportunity to make or lose money relative to peers. For example, IHS Markit sees an opportunity for banks to correctly price the impact of asymmetric FVA.
Nicki Rasmussen: One source of differentiation between banks has been the funding levels used for FVA – these differences materialise for funding heavy or, more importantly, beneficial trades, such as long-dated uncollateralised cross-currency swaps with a significant foreign exchange forward component. These observed differences also highlight that, while FVA has been on the radar almost as long as CVA, the jury is still out on the consensus approach.
More recently, KVA has attracted more attention – not only at quant conferences or in articles published on Risk.net, but also in the understanding of how derivatives business is run and measured through return on allocated capital. The differentiation is most likely even larger than for FVA, which is due to significant differences in regulatory regimes, the use of internal models for CCR, and the different treatment by local financial services authorities. Further to these external factors, the commercial approach taken to the pricing of lifetime capital costs can differ substantially between banks. There is an obvious short-term strategy for being competitive on the KVA part, but it comes at a high risk of both ‘winner’s curse’ and ‘buyer’s remorse’ in the long term.
Stamatoula Matsoukis: KVA is an active research topic. Given its complexity and direct linkage to regulatory capital in a regulatory environment that keeps changing, it can offer the opportunity for competitive advantage to firms with the resources to calculate it properly.
Julien Second: Unlike CVA and FVA, which are risk-neutral – in other words, largely about replicating derivatives cashflows – KVA relies more on business inputs than market observables. Being the source of pricing discrepancies, this is also where competitive differentiation is possible. Business assumptions can and should be monitored over time. Depending on the bank’s capital regime, economic CVA hedges can be used to reduce the regulatory CVA, in turn reducing KVA, and so on.
How can banks best optimise their total XVA consumption?
Julien Second: Banks should centralise risks to avoid operating in silos, but achieving this requires a fair amount of computing power and infrastructure investment. That said, there is certainly a cost/benefit analysis to be made. If the size of the derivatives portfolio is small compared with the overall business, the reward might not be worth the effort of investing in a full-blown XVA capacity.
Nicki Rasmussen: Centralising XVA desks is a good foundation, but can only take you so far. Most banks must consider the link between their business model, their product and client mix, and their use of scarce resources such as credit risk appetite, funding and capital costs. This is a challenging journey some banks have already taken, while others have not started yet.
In the category of low-hanging fruit – or short-term measures – elements such as trade novations, triangulations and central counterparty (CCP) backloads is always a useful approach, but the advent of clearing obligations and uncleared margin regimes makes these strategies even more complex.
Stuart Nield: Reducing XVA by agreeing or modifying collateral terms or clearing trades is current best practice in reducing XVA. New opportunities for reducing KVA will emerge under SA-CVA as this capital framework allows hedging of CVA market risks. Reducing Isda’s standard initial margin model, IM and MVA are areas of market interest but wide adoption has yet to come.
What XVA risks does the demise of Libor introduce?
Lucy Thomson Jones: With potential Libor cessation post-2021 and a move towards new RFRs – SOFR in the US, €STR in Europe and the sterling overnight index average (Sonia) in the UK – all related contracts and collateral requirements will change. Discussions on the topic are in the early stages, but clearly any risk profiles will be affected as the new RFRs are implemented. This will present both global and regional challenges as implementation happens gradually and at different times. Within the XVA space, the challenges are magnified by the additional uncertainty currently in derivatives pricing and the additional knowledge required for the life of the instrument as it remains unclear how and when changes will be implemented. A key difficulty will be around sourcing the data required for the calculation of XVAs in an uncertain environment with potential lack of liquidity in some markets.
Nicki Rasmussen: Libor reform will have significant implications for the derivatives business in general. A particular concern around XVA is the potential risk of getting in-scope for IM when dealing with fallbacks. Here the industry – led mainly by Isda – pays great attention to these risks, and addresses them to regulators in a timely fashion.
As part of the benchmark reform, a significant number of credit support annexes will need to be updated. This may have valuation implications and require negotiation between the two parties on these agreements. This process will not be as straightforward as the shift of price alignment interest (PAI) and discounting curves planned by major CCPs for 2020.
Julien Second: CCPs are switching the PAI/discounting rate in 2020 on cleared derivatives – Eonia to the euro short-term rate (€STR) and Federal Reserve funds to the secured overnight financing rate (SOFR) for the US dollar. Some XVA desks manage the discounting risk and will be involved in the exchange of value and risks arising from this change. This is the first step on the road to Libor decommissioning.
Another effect will be on FVA – derivatives’ unsecured funding. Being an unsecured funding rate, Libor is currently the natural choice to measure FVA against. When Libor goes, FVA needs to be redefined to the new secured risk-free rate (RFR). That said, the change might just be optical because it redefines the line between present value and FVA, but probably won’t affect the external all-in derivatives price.
Stamatoula Matsoukis: The replacement of Libor with five other curves will affect the discounting used for XVA calculations. More sophisticated firms are expected to move to the new approach more quickly. Until the new standard is fully implemented by the industry, there will be mismatches in the valuation adjustments among the firms.
What impact will IM rules have on pricing?
Nicki Rasmussen: While MVA is a well-defined theoretical concept, it has yet to materialise in the pricing of derivatives trades to the same degree as CVA and FVA. This can largely be explained by the fact that phases one, two and three affected mainly the interbank community, where XVA charging is largely absent. Having an MVA measure for both cleared and bilateral IM allows for monitoring and management of the associated funding costs, even before they enter official accounts. Finally, while the exchange of IM has the potential to materially reduce both CVA and KVA, it takes a targeted review of internal models-compliant IM model implementation to get the full capital savings benefit. This is no small endeavour, and comes at a time when internal models are significantly challenged by the output floor in particular.
Stamatoula Matsoukis: IM scenario projections will reduce the net exposure levels used in the calculation of valuation adjustments.
Lucy Thomson Jones: As banks implement margining rules, the expectation has been that MVA will be an area of increased interest and there is certainly more discussion on the topic. However, most banks are not currently at the stage of pricing it into their XVA. The industry appears some way off from standardising accounting of MVA, and major impacts are yet to be seen.
To what extent is cloud computing transforming XVA?
Stuart Nield: The past five years have seen tremendous innovation in cloud-based provision, allowing banks to outsource their costly computing infrastructure. XVA systems designed to run on the cloud can spin up to handle large batch loads and then spin down to support lower volume pre-deal calculations. The bank pays for the compute time rather than provisioning hardware for the highest load, which prevents the hardware being under-utilised at other times. The elastic framework also supports ad hoc calculation requests that could not be supported previously. The cost of storing data on the cloud has plummeted, opening up the possibility of storing voluminous intermediate results rather than rerunning calculations for what-if requests.
Julien Second: One of the difficulties of managing XVA books is to anticipate risk changes arising from cross-asset market moves – so-called cross-gammas. These are usually predicted by running sets of stress scenarios, which work well but are expensive to run given the high number of dimensions. Cloud computing allows scalability in several ways: speed, flexibility and risks can be rerun on demand, at a cost. This is especially valuable in volatile environments, where risks are rapidly changing. Where having live risks on XVA books was inconceivable, the advance of cloud-supported technologies makes this more of a reality that could be achieved in the medium term.
Nicki Rasmussen: For brute-force XVA implementations there may be significant benefits and flexibility from using public cloud solutions. However, for leaner implementations based on the most efficient algorithms this may not necessarily be the case.
Which technologies will prove most valuable to banks in XVA?
Julien Second: Adjoint algorithmic differentiation (AAD) is certainly benefiting from increased computation power and scalable technologies. There have also been a few interesting artificial intelligence (AI)-based developments, improving calculation times significantly and flexibility around models. XVA is a natural beneficiary of these innovations, but other areas in the bank such as credit measurement and enterprise risk can also benefit from these technologies. Portfolio management is another area to which AI could add value and, although these are early days for XVA, a few interesting ideas have been proposed around options trading that could see further applications.
Nicki Rasmussen: AAD will continue to find its way into not only XVA, but derivatives pricing and risk management in general. This technology has a strong link to machine learning, as it may allow very complex systems to be differentiable, enabling all sorts of optimisation strategies. Equally crucial from a business point of view are technologies supporting efficient management and use of large amounts of value and risk data produced in the daily P&L and risk batches.
Stuart Nield: Systems built using open-source components will dominate future XVA technology. Open-source components benefit from a continuous loop of expert feedback and innovation as new versions, tools and features are released and incorporated. Only the specific software challenges of calculating XVA need to be tackled; other non-functional challenges, such as security and encryption, have already been solved by the open-source community.
*IHS Markit is a registered trademark of IHS Markit Ltd and/or its affiliates. Opinions, statements, estimates and projections (collectively “information”) in this article are for informational purposes only and are solely those of the individual author(s) at the time of writing and do not necessarily reflect the opinions of IHS Markit. Neither IHS Markit nor the author(s) have any obligations to update this article in the event that any information changes or subsequently becomes inaccurate. IHS Markit makes no warranty, express or implied, as to the accuracy, completeness or timeliness of any information, and shall not in any way be liable to any recipient in respect thereof. ©2019 IHS Markit Ltd. All rights reserved.
Copyright Infopro Digital Limited. All rights reserved.
You may share this content using our article tools. Printing this content is for the sole use of the Authorised User (named subscriber), as outlined in our terms and conditions - https://www.infopro-insight.com/terms-conditions/insight-subscriptions/
If you would like to purchase additional rights please email [email protected]
Copyright Infopro Digital Limited. All rights reserved.
You may share this content using our article tools. Copying this content is for the sole use of the Authorised User (named subscriber), as outlined in our terms and conditions - https://www.infopro-insight.com/terms-conditions/insight-subscriptions/
If you would like to purchase additional rights please email [email protected]