The requirement for General approval before Specific approval is not required under FRTB and we believe that there is nothing in CRR Art.367 or Art.370 that makes specific risk permissions contingent on general permissions. The consultation paper (CP) does suggest that the CRR requires this sequencing. However, we consider that the CRR merely applies standards for specific risk modelling which include all those required for general risk. This does not logically require that a firm has to have both approvals, but rather means that both sets of standards need to be met for specific risk approval to be granted.

Further, we do not believe that the requirement for sequencing is conceptually necessary. For example, Debt-General and Debt-Specific risk categories are not variants of the same asset class, but essentially represent different asset classes (Interest Rate and Credit Spread, respectively), which are often traded on different systems, by different lines of businesses, using different risk and valuation methodologies. This means, for example, that it would be impossible under the proposed standards for a firm to seek model approval for a credit trading business (using mostly credit derivatives, with little general interest rate risk) without also having approval for its rates business – because the credit business would fail the significance test for general risk, but would be unable to seek specific risk approval without it.

Lastly, there would seem to be no downside in allowing firms to seek specific risk approvals in the absence of general risk – it would not appear to present an opportunity to ‘game’ the approval process and, as noted above, the CRR already requires firms seeking specific risk approval to have in place risk measurement and modelling standards super-equivalent to those for general risk.

On this basis we consider there should be no requirement for general risk model approval automatically to precede or accompany an application for specific risk model approval.

We believe that the threshold levels for assessing ‘significant’ and assessment method by Risk Category are too high, and go directly against the approach used in FRTB.
The threshold levels set for positions intended for IMA of a given Risk Category at 90%-95% are extremely high. The language used in the CRR requires a ‘significant share’ of positions in a risk category to be captured – this is not the same as the ‘vast majority’ or even ‘most’ of the positions, and would not appear to justify the 90-95% level proposed in the CP. Furthermore, the thresholds are much higher than the 10% of the bank’s aggregated market risk capital coming from IMA-approved desks which is viewed as a ‘significant share’ under the equivalent FRTB requirement.

We also consider that allowing differentiation by, or exclusion of, positions by product complexity or business area (perhaps temporarily to allow phasing of the model application process) would not contradict the CRR text, whilst granting firms and supervisory authorities the flexibility required to manage the approval process. This would be more consistent with FRTB (which allows differentiation, subject to the 10% threshold, by trading desk), and is also consistent with current supervisory practice, for example in the case of the UK PRA, which differentiates model permission scope by broad classes of positions within each risk category (PRA Supervisory Statement 13/13 9.4).

We find the overall proposal to be reasonable and generally in line with the existing practices. However we would like to reinforce industry comments about the prescriptive nature of certain requirements as listed in their response. The overall governance arrangements that are in place for large financial institutions would make it impractical to implement a framework with just one ‘committee’ which this proposal seems to imply.

While we agree with most of the proposed validation requirements, we would recommend that the standards recognise that elements of the validation process may be carried out by different functions. For example, the robustness of IT infrastructure is something which would often be carried out by a technology function, rather than the group primarily responsible for model review and validation.

We are supportive of industry response and the issues highlighted are largely in line with our findings. We would like to reinforce particularly the point regarding day to day management of limits by committee which we find impractical, and propose allowing delegation of limit authority to the appropriate individuals, under an agreed framework.

We acknowledge that it is desirable from both a supervisor’s and a firm’s perspective that an internal model used for calculating own funds requirements have a proven track record of reasonable accuracy in measuring risks. That said, the requirement of 250 days is overly onerous compared to the approach under FRTB which is interpreted to allow for discretion to observe model performance post submission until the required 250 day observation period is reached. Our view is that three months, which in effect is the current standard applied by some regulators (e.g. PRA, SS13/13 para 9.10) should be a sufficient time to demonstrate stability of the systems and processes supporting IMA models. A longer period might be required prior to the model ‘going live’ for the purpose of capital calculation.

VaR is looking to capture the tail risk and as such should be benchmarked against all risk factors for the actual P&L computation.

We support industry response and would like to reiterate that Net interest income (NII) exclusion should only be applied to an accrual book. However there is no reason to exclude NII from MtM securities where it presents a component of carry.

Where adjustments are calculated less frequently than daily, it may be hard to determine whether any movement occurred on a discrete date or more gradually over the readjusted period. As per the industry response, it should be a valid reason for removing the breach which is evidently caused by such adjustments.

VaR does not capture theta as a tail risk and to be consistent with this, Theta should also be removed from Hypothetical P&L for backtesting. While CRR may be not be clear on whether Theta should be included in VaR, FRTB specifically refers to applying ‘instantaneous price shocks’, implying that Theta should be excluded from both VaR and P&L.

There are significant challenges around the use of stochastic correlations and so we do not feel that stochastic correlations should be required as long as there is sufficient monitoring around the actual correlations used.