We believe that the question is ill posed in the context of Back-testing of a VaR Model. A VaR Model reprices financial instruments under market risk scenarios and should hence be benchmarked against a P&L that is adjusted (in case needed) for deficiencies in either of these two dimensions: pricers or market data.
As such adjustments like XVAs that accommodate the fair value of an instrument so that the exit price properly reflects the credit merit of the Counterparty, the Funding Costs embedded by the trade, the Capital costs, etc., while being sensitive to market risk, should not be part of the considered Adjustments.
The type of adjustments that are not considered relevant for backtesting are as follows:
• XVA in general:
o CVA (irrespective of the ad-hoc capitalization)
o Funding VA
o MVA (Initial Margin Funding Costs)
o KVA (cost of regulatory capital absorbed by a transaction)
• Write down / write offs due to the default of a Ctp (covered by CVA) or Issuer (covered by Loan loss provision)
P&L driven by changes in XVAs represent variation of a fair value component of the transaction that is not necessarily caused by market risk, but rather by other risk factors.
To give an example, a funding valuation adjustments (FuVA) is defined as a pricing adjustment used to rebate the financing cost of a derivative transaction over its whole life to the customer. FuVA is hence computed out of a simulation that spans the whole life of the product (potentially 50Y) capturing the funding costs and benefits generated from its in-flows and out-flows. The modelling of such effects is only possible via the XVA Monte Carlo Engine in which market factors are simulated along the life of the derivative that ages to its expiry. The funding cost/benefits sustained by the bank through-out the life of the trade represent the driver of the resulting FuVA.
Market risk-factors will clearly influence the value of its cash flows and, as a result, the funding cost/benefit that they originate. It is however important to point out that the effect they exert on FuVA is not comparable to a sudden change in current PV (typically captured by Market Risk metrics like VaR) but is rather related to the funding of an ageing trade until its maturity. In this respect we do not deem FuVA to be a market-risk related adjustment, because neither does it adjust the market value of a derivative for elements not captured by the pricers nor by the set of market data fed into pricers. As a result, while being included in Economic P&L, it cannot be part of APL.
The consultation paper introduces an asymmetry between the HPL and the RTPL, because the requirement to have daily computed adjustments is only included in HPL and not the RTPL. As a result, this increases the likelihood of desks failing the eligibility tests, because this adjustment raises two different problems:
• On the one hand, the extreme sensitivity of the PLAT to even small outliers will mean that this asymmetry will push the desk in the red zone. Therefore, this provision will generate a disincentive to increase frequency of the adjustments.
• On the other hand, less frequent adjustments will increase the risk that there are backtesting exceptions, which is particularly relevant in light of the fact that no smoothing is allowed. This is also true for APL.
The possibility for national authorities to grant exemptions, which is provided for in MAR32.27 of the Basel text, should also be foreseen in the EBA RTS to mitigate the risk of desks failing eligibility tests.
Moreover, the reference to the sub-paragraph 3 of Article 33(1)(c) of Regulation (EU) No 575/2013 should be removed from Articles 1, 2, 3 and 4 of the draft Delegated Regulation with regard to back-testing requirements under Article 325bf(9) and profit and loss attribution requirements under 325bg(4) of Regulation (EU) No 575/2013. Indeed, Article 33(1)(c) of Regulations (EU) 575/2013 and 2019/876 has no sub-paragraph.
Only adjustments that correct the P&L for deficiencies in pricing models or market data should be part of the Actual P&L used for Back-testing. A non-exhaustive list of examples could be
• Independent price verification (IPV) Adjustments
• Close Out Cost Adjustment (uncertainty around Mid-price)
• Less Liquid Position Adjustment (uncertainty around Bid-Ask)
• The adjustments of the model correcting the wrong valuation of the trade
• Corrections/ Market operations (MOPs)
The provision is clear and aligned with the Basel text, however it is likely to introduce structural issues in Back-testing and PLA.
Smoothing the effects generated by several days of market volatility on a single day P&L seems like a much more desirable approximation than counting back-testing exceptions triggered by the comparison of a one-day VaR metric vs adjustments updated at weekly/monthly frequency.
As discussed in Q1, daily adjustments (either computed or approximated via smoothing) should not enter HPL used in PLA anyway.
The provision is clear but not fully aligned to Basel text. We strongly disagree with the provision since it introduces purposed built P&L elements that have no recognition in accounting/economic terms and as such are not managed or steered by the banks
The requirement to compute Adjustments at Desk level is introduced with a degree of prescriptiveness not to be found in Basel market risk standards of 2019 (d457). Dispensation from the requirement is only allowed after justifying with dedicated documentation the adequacy of alternative aggregation levels.
We deem the request to have Desk level Adjustments very problematic in that it introduces a BT-only relevant P&L element that is not otherwise used by the Bank nor recognized and managed by the Traders.
It introduces the concept of a by-desk risk-management of the adjustments that is clearly neither optimal nor part of the standard practice for books and records P&L of a Bank. This inconsistency could result in P&L swings across desks causing BT/PLA failures. In addition, this may lead to further divergence from the P&L used for front office business management.
On the other hand, all the activities and documentation required to justify not using Desk-level Adjustments will necessarily generate a complexity and additional operation burden that could be more easily avoided aligning the requirement to those contained in the Basel text.
The same arguments hold also for the restriction of top-of-the-house adjustments to IMA Desks only. While the objective of harmonizing the perimeter of VaR and P&L is clearly understandable, it should not be implemented as a re-aggregation of the Adjustments on the set of trades stemming from IMA Desks only. The possibility to use the amount of Adjustment ascribed to IMA Desks by the internal back-allocation mechanism should be rather allowed to avoid proliferating P&L elements that are used for regulatory purposes only.
The EBF in principle agrees with the criteria set out in paragraph 5. However, those criteria should be relevant only for those adjustments that are market risk related.
Following the EBA 2020 EU Stress Test draft methodological note, the following definitions apply:
Day-one profit or loss: the fair value impact of new financial instruments that at the end-of-day (due to market fluctuations, passage of time and other factors) have a fair value price that differs from the transaction price;
Day-one reserve: sum of the amounts that are reserved for day-one profits that cannot be directly recognized in P&L. This is because the fair value estimation at the end-of-day of new financial instruments (mainly L3) is too uncertain, due for instance to the use of unobservable inputs
When looking at the End-Of-Day P&L of a portfolio (managerial view), the Day-1 P&L does not play a specific role and is generically part of the MOP (Market Operations) section of the P&L Explain process. This includes new trades, expires, novation, cancellation and reissuances, etc. Accordingly, the Day-One P&L is part of the End Of Day P&L and of APL. It is not part of HPL since the portfolio is kept static there.
When looking at the Accounting P&L representation however Day-1 P&L for L3 instruments is reserved and gradually released throughout its life.
As for the client margin this is not explicitly identified when it comes to OTC transaction, but it is rather part of the final price of the trade.
The provision is clear but not fully aligned to Basel text. We strongly disagree with the provision since it introduces purposed built P&L elements that have no recognition in accounting/economic terms and as such are not managed or steered by the Banks (see also Q4).
The possibility to recalculate the adjustments for top of the house back-testing only referencing positions under IMA should indeed be left as an option. However, using the back-allocation rules in place within the bank rather than by triggering recalculations. Not only a recalculation would be burdensome, but it would also suffer from the shortcomings described above and reported in Q4.
The conditions for the use of the HYP risk factor value are not really clear (in particular what is meant by paragraph 2), however the principle as such is sound and should be left as an option in those cases where it might proof useful.
We would definitely see its advantage in dealing with RNIM (RF not in Model) where – if capitalized under SES – it is allowed to use the actual return realised in HPL to help passing PLA.