Lloyds Banking Group

LBG has no strong view on either of the two treatments.
LBG includes IRC excluding default risk of own debt ie just includes the migration risk. LBG includes own debt in VaR/SVaR.
LBG do not expect any major operational challenges if have to move own debt outside of internal model scope.
LBG is happy with the approach. We assume that when ring fencing is applied if we are allowed to grandfather our existing Market Risk Permission for the non-ring fence entity we would not be expected to run the proposed model application. Please could the approach for this UK specific issue be clarified.
LBG considers that the 4 categories outlined in section 9.4 of SS13/13 are more appropriate given this categorisation is used in our Internal Model Market Risk Permission for stating the scope of our Permission. We are concerned that the implementation of new categories would lead to LBG and other banks being made to change their Internal Model Market Risk Permissions.
The EBA approach appears to conflict with the FRTB approach, although it is possibly Article 363 (2) which is forcing the EBA to take this approach. The FRTB approach for model approval follows a desk by desk approach (overall assessment and then desk level assessment to determine which desks are in scope of internal model, to reduce the possibility of an abrupt move of all the businesses to standardised approach). An alternative approach based on setting thresholds for identifiable desks, desk definition in line with the new Accord covering FRTB, could be more consistent with FRTB.
LBG believes that an approach based on identifiable desks may work better.
As with Q’s 6 and 7 the ongoing thresholds would need to be consistent with the FRTB approach. Under FRTB the assessment of the desk level model performance may require desk(s) to be moved to the standardised approach. This could lead to the ongoing thresholds being breached and hence require the whole trading book to move to the standardised approach under the new framework. Hence would suggest thresholds based at desk level.
The situation under the new internal model framework could be handled by including desks that have reverted to standardised for “technical” reasons. They could still be considered as covered by the internal model approach.
Article 16, point 3 (a) assumes institutions calculate standardised approach for desks which are within the scope of their internal model. This is unlikely to be the case in advance of implementation of FRTB.
Note that for using own funds requirements as described in 3 (b) as at 31/12/14 for LBG share of standardised approach was 35%, excluding RNIVs it was 40% (This includes all standardised approach positions including TB securitisation positions, without these they would be slightly lower).
Article 16 (3) states “for which permission is sought”, but Article 16 is about subsequent reviews - presumably this is an error.
Article 16 (3) b is not sufficiently precise in two aspects: 1. “indication of growth” and 2. Use of “P&L”. Could we have more guidance?
LBG have no issues with these proposals.
Institutions should consider all possible means of validation and do what they consider are relevant for their trading books. They should be able to demonstrate why they consider some types of validation are not appropriate for their trading books.
LBG has the following concerns:
Art 23 (2) c. assessing empirical correlations is the right thing to do but monthly is potentially onerous (Art 46(2)b).
Art 23 (2) h. A bit vague and potentially onerous/inconclusive.
LBG considers that they do allow sufficient flexibility.
LBG considers this to be too onerous/inflexible to have to re-state up to 250 day backtesting history in order to introduce improvements to the model. It also considers this proposal will become an entry barrier for new banks.
Model accuracy track record requirements look reasonable. However as per question 4 we assume that when ring fencing is applied if we are allowed to grandfather our existing Internal Model Market Risk Permission for the non-ring fence entity we would not be expected to perform 250 business days of backtesting. Could the approach for this UK specific issue be clarified?
We also assume that requirements for model changes under the existing Permission are covered under the current RTS for assessing the materiality of extensions and changes of internal approaches.
For Article 34 1. The wording should be relaxed to remove: “no major system breakdowns shall occur” and replace with a Support Level Agreement on recovery of major system breakdown.
Not relevant to LBG
Not relevant to LBG
For LBG there would be no additional burden
LBG include in the P&L all pricing risk factors of the portfolio. We consider attempting to break out all risk factors which are not being used by the VaR model itself as impractical.
NA given above answer was yes.
As with our answer to 22 there is no justification with stripping out risk factors.
LBG does not consider this to be an issue as the trading book P&L should be valuation based not accrual based.
For LBG the intra-day P&L should be relatively small compared to moves based on close of business exposures.
Valuation adjustments will feed through into P&L on the day the adjustment made.
LBG consider Theta should be in both definitions of P&L used for backtesting.
LBG does not believe that they should be required. The proposal seems to suggest the use of stochastic correlations regardless of expertise within the institution. Institutions should be able to demonstrate that their VaR model appropriately reflects correlation risk.
LBG uses historical simulation and therefore implicitly uses empirical correlations. To review implicit correlations from historical simulation at least monthly would be very difficult to perform.
LBG does not disagree with the individual elements. Note that RNIVs can be implemented to capitalise for a risk not capture or not adequately captured by the VaR model. In such cases where an RNIV has been implemented then the regulator should have the autonomy to consider the additional capital required through RNIVs whilst considering the multiplier factor.
We welcome an explicit list for assessing the SVaR multiplier. We do not see why the SVaR multiplier should always be required to be higher than the standard VaR multiplier. The VaR and SVaR should be judged on their own ability to capture the risks adequately
We welcome an explicit framework. Our concern is that if the regulator had concerns across a number flaws/shortcomings resulting in an increase in the multiplier by several points.
Also how would it be possible to ensure the regulators apply the review framework proposed for multipliers uniformly and consistently across EU?
No additional comments.
No issues from LBG’s perspective.
No issues from LBG’s perspective under our current methodology.
LBG does not agree with this approach. LBG does not use internally derived ratings and considers these not to be practical.
We would like to point out that each institution could have quite different PDs dependent on their IRB model which given Supervisors’ concerns on the variability of internal models may be the wrong direction to proceed.
It would provide a more consistent approach between different institutions if external based PDs were sourced by all institutions.
Yes, but would need guidance on levels.
Could use both.
LBG has the following comments:
Article 69 2. b. to consider different copula candidates. We suggest that this should be less explicit and require institutions to demonstrate that the copula used in their model is appropriate.
Article 64 2,3 we would like to see more explanation why points 2 and 3 are being proposed.
Article 65 2. Would it be possible to provide more guidance on how institutions should assess the robustness of transition matrices plus what actions could be envisaged to
Article 72 requirement to have procedures to ensure that on the day the IRC is calculated the portfolio is representative of the portfolio held during the week.
Paul Fairhurst