We agree that the market data used or the source of information for expert based approach both need to include reliable and available data. We do not believe that all the data listed here needs to be sourced which seems unrealistic from operational point of view. We therefore call EBA to clarify that the purpose is not to multiply the sources for the sole purpose of constructing data ranges, but to undertake reasonable effort for collecting reliable data from the indicative list provided.
Yes. Having parent institutions computing AVAs on a different methodology from the one used for subsidiaries requires a double computation and therefore a heavy operating burden. The CP should rather shed clarity that it is possible to net the exposures between different legal entities, and concentrate on how the resulting AVA can be split amongst the entities contributing to the exposure. We believe therefore that the two methods must be exclusive from one another. If a subsidiary is using the simplified approach then the perimeter of this subsidiary needs to be excluded from the application of the core method.
While we understand the possible regulatory concern that the CET1 capital can be made of substantial unrealized gains, we believe that this concept virtually has no link with the valuation uncertainty and should be removed from the RTS.
A simple example may show that a very exotic trade with unrealized loss has considerably more valuation uncertainty than a plain vanilla liquid bond with unrealized gain. Accordingly, we also dislike the asymmetric treatment that is proposed between unrealized gains and losses and we further underscore that netting these gains and losses for the purpose of determining a proxy of the size of the uncertainty makes little sense.
Another remarkable drawback of this metric is its direct dependency to when the deal took place. This dependency makes unrealized P&L a rather poor proxy of a valuation risk which is (by definition) in relation to actual position and actual market conditions.
From an operational stand point, we underscore that this concept is not used by valuation and risk systems, which are the natural sources for tracking valuation uncertainties. Specifically, the initial cash proceeds are not systematically tracked by the valuation systems (for example for bonds); therefore, it is expected that the implementation of this requirement will involve significant costs.
Globally, the reference to unrealized profits is not a good proxy for valuation uncertainty, while its implementation burden will stand as pure compliance cost with no economic benefit to the assessment of the real sources of valuation uncertainty.
Therefore, we recommend the EBA to simply remove the component “25% of the net unrealised profit on financial instruments held at fair value” from the simplified approach formula.
Should the EBA nevertheless decide to keep the approach unchanged, the percentage related to unrealized gains and losses should be revised. Indeed, based on our calculations, it leads to record more than three times the average AVAs recorded by UK banks. We believe 5% instead of 25% would be more appropriate.
An approach based on (or similar to) the fair value hierarchy is definitely the right way since it would pave way for convergence between the accounting and regulatory requirements.
Liquidity ratio classification is also a possibility, but we believe that the LCR liquidity flag differs from the fair value classification because the fair value hierarchy focuses on the degree of observability (transparency and in-exchange liquidity) and its scope is wider than securities. Therefore the use of the LCR liquidity flag is at best complementary to another form of classification applicable to derivative exposures. .
Had the EBA chosen a differentiation of the positions by type and level of liquidity (in-exchange), we believe that the estimation of the valuation uncertainty should remain simple and based on either the present value or the notional (or a combination of these)
We obviously acknowledge limitation of these proxies as it is well known that the risk is not necessarily proportional to the present value, and that the notional is not sensitive to the market conditions changes. However, we think that the EBA choice must remain pragmatic and simple.
Had the present value be retained amongst these, the AVA formulation should allow for distinction between funded and unfunded valuation positions.
In substance, the simplified proxy could consist in a percentage of the present value or notional value of groups of assets and liabilities, which are specialized by valuation risk class (possibly a fair value hierarchy classification or a more granular alternative to such classification). In this formulation, we emphasise that balance sheet positions in each of the classes of instruments should be taken on netted basis, since the distinction between assets and liabilities is not meaningful from a valuation risk perspective.
Again, we dislike the use of net unrealized gains concept and further underscore that the proposed rule is such when the sign of the P&L changes, the appreciation of the valuation risk changes, while there can be virtually the same valuation risk whatever the P&L sign.
The reference to the notional value or present value makes the proxy more meaningful, and we call EBA to retain either of these as the basis of the measurement in line with the answer to question 5 above.
However, the proposed charge of 10% reads as overly punitive when compared with existing AVAs.
We understand the regulatory perspective and objective with such charge, but we underscore that there are situations where the rule might be an “over-kill” or inappropriately hamper innovations and new business undertaking.
In fact, it is customary that when positions are not material at the institution level, the institution could conclude that deploying the core approach does not pass the cost/benefit test and that it is more reasonable to retain reasonably prudent valuation adjustments (even prohibitive ones). Along the lifecycle of the business, when positions and business flow grow in size and risk, institutions may further invest in methodology refinements and adopt more sophisticated data collection processes and calculations, whose cost is in relation with the business profitability.
That reflects that there is proportionality between the size of the risk and the degree of sophistication in the IPV and valuation adjustment process. This proportionality is of utmost importance when seen from a granular level (i.e. desk level) because it directly affects the entry cost to the business. When no proportionality applies, desks are either subject to heavy investments costs on IPV and valuation adjustment process, or to a punitive capital deduction, which in turn affect their competitiveness.
In order to preserve the playing field of EU banks, we call EBA to retain this proportionality principle. The calibration of the default charge should in our view correspond to a “reasonably prohibitive charge”, but not turn to be a pure punitive one. We therefore propose to :
o at least half the charge that applies on the notional amount (to 5%),
o remove the formulation based on the 100% of the net unrealized gain (in line with our comment on this concept).
In any case, and since we expect this default charge to apply for immaterial positions for which there is no willingness to develop sophisticated measures, we urge EBA to have this default calculation be simple and to avoid reference to multiple ingredients in the calculation.
We agree on the zero AVA exception for Market Price Uncertainty and Close Out Cost, and we welcome the relevant reference to art 338.
We also welcome the recognition of risk offsetting and find the risk reduction test (P&L volatility ratio) relevant in principle.
However, as stated in our general comment, we underscore that the formulation of the articles 8 and 9 of the RTS might have the unintended consequence of imposing prescriptively the risk representation beyond what is necessary to achieve the text purpose. This is twofold:
o the prescriptive statement that the valuation exposure for a non-derivative exposure is the price, might lead to isolating securities risk positions within a portfolio, which would possible depart from the institutions’ risk representation expression. We believe that this distinction is unnecessary and brings no benefit on the methodology applied. In fact, the removal of the reference to the “non-derivative positions” would keep the level of requirement identical since valuation positions from portfolio made of derivatives and non-derivative would still need to be mapped to traded instruments.
o The requirement that the AVA calculation is made out of risk representation based on traded instruments, might also be unnecessarily prescriptive. As stated in the general comments, we would welcome that the text allows that the AVAs are still based on the sensitivities used by the institutions in their day to day risk management pending that the applicable charges are calibrated to market data. The institutions electing this option would still be required to test the validity of their risk representation through a P&L volatility test. Had such test failed, a model risk AVA would then become mandatory to the institution.
As outlined by many responses to the DP, and although significant clarification have been provided by the CP, notably through the paragraph 2 of article 9, we believe that there is still a risk of overlap between the close out and market uncertainty measures.
In fact, in most institutions, the valuation adjustment process is operationally distinct from the independent price verification (market data verification), even if they share the same datasets. In practice, there are generally two distinct steps: one consisting is determining the appropriate mid-point amongst available data which is used to value all positions, and another consisting in determining the valuation adjustment on the net exposures.
We would welcome more clarity in the RTS that when the estimation of the valuation adjustment is using exit price data, there isn’t any need to compute both Market Price Uncertainty and Close Out Cost AVA.
Put differently, we would welcome clarity that the RTS intention is not such if the IPV process is not based on the adverse bid/offer prices, the two AVAs are required.
Regarding the Unearned credit spreads AVA, we welcome the treatment of the CVA as any valuation position that is subject to uncertainty in the inputs and in the modelling. We would welcome more clarity in the RTS that according to paragraph 2 of article 10, when the Market Price Uncertainty component of the ACVA is aggregated with other Market Price AVAs, this would be eligible for diversification benefit under article 17. Similarly, the same clarification for the model risk component applies.
We agree on the approach for Model risk AVA. However we reiterate that since there is a continuum between Market Price Uncertainty on the model inputs and the model risk itself, it is of utmost importance that the text does not lend itself to arbitration such the choice of how model risk is estimated leads to different classification in the AVA categories and hence to different treatment of diversification.
A simple way of addressing this issue to allow for diversification when the model risk is being assessed through a data range approach, as explained in the general comments.
We find the article 12 about the Concentrated positions AVA too prescriptive on how the concentration risk AVA is computed. This article further requires significant amount of additional data that might not be collected in a systematic manner during the IPV process, or be hard to obtain (notably the market size); it finally remarkable that this article does not contain any provision about a possible judgmental approach.
In order to reduce the level of prescription while providing incentive to institutions to opt for a sophisticated approach to Concentrated positions AVA, we believe that the RTS should be amended such two options are offered:
o A judgmental approach where institutions would document a thought process to determine the concentrated positions AVA, without undertaking systematic data collection exercise. In this case, the regulator is notified and the resulting AVA is not eligible for diversification,
o The RTS proposed approach (paragraph 2 of article 12) which is required to be substantiated with actual data from the IPV process. When institutions opt for this sophisticated approach, the resulting AVA would be eligible to the diversification benefit under article 17.
With regard to the investing and funding AVA, we agree on the current wording of the text.
For the stake of simplicity of the RTS, and since we believe that past experience on client early termination is not predictive of the future early terminations, we believe that the text should assume zero value AVA for early termination. The RTS might however impose a yearly assessment of the non-materiality of early termination based on both past experience and judgment.
We believe that own credit risk for securities designated at fair value and derivatives’ debit valuation adjustment should be placed outside the scope of this RTS and assumed to have zero AVA.
We finally believe that the future administrative costs should be assumed to be negligible for plain vanilla instruments which would allow to put the focus on most sophisticated exposures that involve non-standard booking, trade processing, risk measurement or valuation processes. EBA might impose that institutions define a list of criteria to be met to determine the portion of their portfolios that would fall within a “non-standard” category.
We agree to the proposed approach. However we would welcome some clarification that the purpose of paragraph 3 (a) is to avoid that institutions arbitrate the diversification rule by subdividing artificially fungible AVAs.
We reiterate our call to enlarge the scope of application of the diversification benefit to some of the model risk AVAs and some of the Concentrated positions AVAs as exposed above.
We strongly disagree.
The ongoing data quality assessment process described in the Article 20 is absolutely impracticable and we recommend it to be deleted from the final RTS.
Not only the methodology proposed for the back-testing is not technically sound, the requirement to use systematically the actual prices entails a heavy system and process changes to enable to handle the storage of data.
Should EBA persists in preserving this requirement, we urge it to drastically reduce the application scope by targeting only the most exotic instruments, and to entirely delete the prescriptive methodology, leaving it up to the institutions to design the appropriate methodologies to make use of this trade data.
We believe that at very best, this test could bring value to institutions when assessing their valuation uncertainty, if restricted to an equivalent to Level 3 instruments in accounting terms.