We have not yet been able to conduct a proper assessment allowing us to adopt a clear preference with regard to the two options.
Institutions should, in principle, be free to opt for either of the two options. To avoid any cherry picking, we suggest that they opt once-only for an option depending on their derivatives portfolio and business strategy and notify supervisors of their decision. Both options should therefore be provided in the final RTS.
May we first point out that, with respect to transactions with more than one material risk driver, we consider it generally appropriate to always only map the respective transaction to the risk category associated with the most material risk driver. The computation of add-ons for several risk drivers indirectly presumes a correlation of 100% for these risk drivers. This presumption is hardly reasonable in our view and, according to our preliminary estimates, will normally lead to an unjustified further increase in SA-CCR conservativeness.
In general, delivering the relevant FRTB sensitivities, particularly the FRTB risk weights, will impose a considerable burden. This will depend on, among other things, the frequency with which transactions have to be mapped to the individual risk categories.
Neither the draft RTS nor CRR Article 277, their legal basis, specify how often such mapping has to be carried out. In this context, we call for a sound, one-time assessment of risk drivers’ materiality at product level that should take place no later than the next quarterly reporting date after close of business. Such an assessment would normally produce the same result as renewed computation on a daily basis, yet would significantly reduce the operational burden.
Mandatory daily computation of risk drivers based on FRTB sensitivities at individual transaction level, on the other hand, is disproportionately burdensome in our view. Mapping one and the same transaction several times during its term should be avoided. This could otherwise lead to the transaction jumping back and forth from one category to another, producing extremely undesirable instability of SA-CCR results and, ultimately, unwanted volatility of capital requirements.
It would also have to be made clear whether, where optional components are concerned, the options delta has to be computed for all categories or only for the category to which the optional component relates.
Furthermore, specific examples of mapping would be very helpful for cross-currency swaps. It should, for instance, be made clear in which currency mapping to the interest rate category is to be made (or which reference rate is to be chosen, both or only one?; which is then more material?). Please also provide an example of mapping of basis swaps.
We welcome it that institutions with a small trading book business that do not need to compute any sensitivities under the FRTB rules are to be given an alternative approach for determining the material risk drivers. May we point out, however, that application of the FRTB SA sensitivities approach can cause problems also for those institutions that are generally required to compute sensitivities for their trading book positions. Computation of FRTB SA sensitivities for banking book positions is not stipulated, for example. We therefore advocate generally opening the alternative approach to all institutions. Should this proposal not prove acceptable, application of the alternative approach should be possible for all institutions at least for positions for which they have no FRTB SA sensitivities.
Please note, however, that the proposed add-ons approach for determining material risk drivers is also burdensome as it is an iterative procedure: as we understand it, Article 3(1)(b) provides that the add-on is to be computed on every reporting date for every derivative transaction and for every single identified risk driver. Only afterwards can it be determined which of these risk drivers are actually material.
In general, the RTS should therefore be designed in such a way so that as many transactions as possible are directly and unambiguously mapped to a risk category without triggering a quantitative assessment according to Article 3. We would also like to reiterate that, in cases where this is not possible, mapping at product level should be possible (see under Q2 above).
In our comments of 15 March 2018 on the EBA discussion paper Implementation in the European Union of the revised market risk and counterparty credit risk frameworks (EBA/DP/2017/04)", we proposed fixed mapping of standard derivatives to risk categories. We appreciate that this has been taken up under approach 1 (referring to table 1 on page 9) for several instruments. However, we deem such fixed mapping appropriate particularly for the following standard market derivatives as well:
In line with our proposal then, mark-to-market cross-currency swaps (MtM CCS), also known as resettable cross-currency swaps, should be mapped to the interest rate risk category only, whereas CCS whose notional amount does not change over their term should be mapped to the foreign exchange risk category.
Bond forwards on investment grade bonds should be mapped to the interest rate risk category.
In our view, such a clear-cut supervisory requirement for mapping to risk categories would still make sense to, on the one hand, ensure a uniform approach and, on the other hand, simplify technical implementation. The quite burdensome quantitative assessment would normally deliver the same result in terms of mapping in any case."
We advocate a delta (δ) shift at transaction level (option 3b). This option would, firstly, be easiest to implement in our view, since to determine the lambda (λ) for a specific transaction only parameters relating to precisely that transaction would be necessary. Secondly, the distortion accompanying the lambda shift would be minimised, since the shift would have to be applied only to those trades where really necessary and, in each case, only to the minimum extent required. Furthermore, SA CCR instability and thus unwanted volatility in the resulting capital requirements would be avoided.
In the event of application at currency level, however, a (perhaps just one) critical trade would determine the lambda for all other transactions with the same reference currency. This would mean that the lambda for all other, non-critical trades would be larger than actually necessary. What is more, a single new transaction with a lower spot or strike price than the current critical trade would determine a new value for the lambda of all other trades with the same reference currency. This could result in the overall SA-CCR result changing noticeably.
We have not yet been able to accurately assess the impact on the exposure value resulting from the SA-CCR due to different thresholds. As the impact will likely vary from institution to institution, the threshold should, if possible, be chosen by the institution itself. In this case, too, the threshold should be fixed once only and notified to supervisors so as to avoid cherry picking.
Which is the preferred option depends, in our view, also on whether the lambda is applied at transaction or currency level. If it was applied at currency level, a higher threshold would be useful to serve as a buffer to prevent frequent lambda resetting and resulting instability (see our reply to Q3).
However, as we prefer application of the lambda at transaction level, we generally support – in line with our reasoning in response to Q3 – the EBA’s argument that the distortion accompanying computation should be minimised, which suggests as low a threshold as possible. This is illustrated by an example on page 45 of the consultation paper and is correct for the example in question. Yet, in our view, it is not completely clear whether this finding can be generalised. In particular, a correct delta will often not be available as a benchmark in practice. Further cases should therefore first be examined to determine which threshold would cause the least distortion.
The condition reading “where institutions apply the approach set out in Article 3(1)(a)” to determine long or short positions by assessing the cash flow or hedging purpose of the transaction should be deleted. Otherwise, the current wording means that FRTB sensitivities would have to be used for all other cases. Application of this method appears to us, firstly, to be excessive in cases where there is only one material risk driver pursuant to Article 1 and the transaction was unambiguously mapped to a risk category. Secondly, cases where the method set out in Article 3(2) (mapping based on SA-CCR add-ons) is applied would not be covered. It must be ensured that application of the method provided for under Article 6(b) is possible also in such cases.