15 April 2014 Mr Ju Quan Tan BCBS Secretariat Basel Committee on Banking Supervision Bank for International Settlements CH-4002 Basel Switzerland Doc Ref: Your ref: Direct : +27 11 645 6708 E- : garyh@banking.org.za Via Email: Juquan.tan@bis.org Dear Mr Quan Tan Comments on: The revised Standardised Approach to Market Risk - Update on revised Accord texts Our member s thank you for the opportunity to provide comments on the two alternative frameworks proposed in your paper of 10 March 2014. 1. Executive summary In general: 1.1 We support the Sensitivity Based Approach (SBA) for the General Interest Rate Risk (GIRR) and Credit Spread Risk (CSR) for the following reasons: o Simplifies the original proposed calculation; o Possible leverage off established risk infrastructure in which the sensitivities are already calculated in an existing risk framework or could easily be extended in this regard; o SBA has a more logical and meaningful risk interpretation relative to the cash flow mapping alternative, i.e. the sensitivities could be utilised appropriately to manage market risk. 1.2 We welcome the addition of the allowance of full netting-off of either sensitivities or adjusted cash flows (depending upon the final Standardised Approach) at an appropriate instrument level. Notwithstanding the above, we are still of the view that the consultative document (issued for comment by 31 January 2014) and the revised Standardised Approach for Market Risk proposal (issued for comment by 16 April 2014) remains a complex model for small emerging market banks to implement. Building a new standardised model that requires substantial cash-flow data inputs for each asset will not be cost effective, let alone trying to find resources capable of building such a model. Considering that the modelling Registration Number: 1992/001350/08 An Association incorporated under Section 21 of the Companies Act 1973 Directors: SK Tshabalala (Chairman), C Coovadia (Managing), M Brown, S Koseff, T Sokutu, E Essoka**, S Nxasana, Ms D Oosthuyse*, Ms M Ramos, L Vutula Company Secretary: N Lala-Mohan (*American)(**Cameroonian)
Page 2 approach bears little resemblance to current trading desk management techniques, we propose that these banks, at the discretion of their national regulator, be allowed to retain the use of the existing Standardised Approach. Banks in the South African context are typically on an internal model and to make a standardised approach so complex that it can be used to calibrate a model, undermines the robustness of the model licence award process. You either have a model that is reviewed by the regulator in the region or use a simplified standardised approach that is capital expensive but simple to implement. Making it more complex to achieve similar results in the two models is confusing as, under the framework for internal models, on any specific desk you will need to be able to switch between the two as the floor removes any incentive to have an internal model. Some emerging market banks with sub branches in other remote jurisdictions will not have any internal models approved and hence will be required to go via the simple standardised approach as the have simple portfolios. 2. Technical Comments and Feedback 2.1 Annex 2 - Sensitivities Based Approach - Draft Accord Section 2 Derivation of notional positions Line 52 56 (Page 1of 27) With regards to the SBA model, the draft proposal indicates that Market Rates or PAR PV01 s are to be used. The use of Zero PV01 s may be more suited given that all of the exposures will need to be bucketed into a vertex of 10 points. We question if there is any specific motive or benefit that The Committee has specifically selected PAR-based bucket exposures and request clarification on this. 2.2 Annex 2 - Sensitivity Based Approach Draft Accord Section 2.2 Other asset classes Line 75 and 106 (Page 3/27) - Treatment of Options and Warrants Both paragraphs prescribe a similar treatment; with the exception that paragraph 106 prescribes an additional treatise on default risk. The main topic of this discussion pertains to the market risk framework, in this context these paragraphs are treated as equivalents. The market risk treatment suggests a two-phased approach to the market risk of option positions: (i) The delta position of an option position is treated as Equity Risk for which section (D) Equity Risk of the document applies, and (ii) non delta risks are captured through the scenario matrix approach discussed in section (I) Options non-delta Risk The most plausible reason for the two-phased treatment is to offset equity positions which are held as hedges for option positions. If this is not the case, then a single phase treatment which entails the recalculation of the value of an option position under different conditions describes market risk more accurately. By having the two-phase treatment, two species of systematic error may be introduced. These are (i) double counting and (ii) unlike comparisons.
Page 3 2.2.1 Double counting Double counting entails that the risks that are accounted for as equity risk is accounted for again as non-delta risk. Section (3) of (I) Options non-delta Risk describes the process of stripping out of the effect of delta from the scenario matrix. The method entails that the specific delta for a set of perturbations be stripped out for that scenario. Mathematically this may be written as [ ] Where denotes the non-delta risk for scenario i, denotes the option value for scenario i, denotes the option value priced with the current market conditions, and denotes the delta, priced with current market conditions and multiplied by the shift in the underlying under scenario i, and the number of underlying specified by the option contract. As stated before, the ultimate value that one wishes to achieve is simply Where denotes the market risk of an option position given the i th scenario. From these two equations we have In order to avoid double counting, one has to account for elsewhere. Apparently the intention is to account for it as equity risk; however the equity risk treatise is of a (potentially) different scale. The risk weights for the various buckets are between 30% and 70%, and can only coincidentally offset the general 40% or 50% perturbation of the scenario matrix. It is therefore virtually unavoidable to encounter at least a portion of double counting for option positions. 2.2.2 Unlike comparisons By not comparing similar instruments will become problematic when dissecting an option position into equity and non-delta risks, with the intention of combining these risks later on. In its simplest form, and ignoring offsets, the equity risk of an option assumes a market perturbed by the risk weight. Since only net positions (after longshort offsets) are noted, this shift is to be interpreted as to be in the direction of the most adverse effect. It could be either up or down; the true direction is obscured. To some degree it makes sense to compensate for this move by eliminating it from the scenario matrix, but only insofar the direction and magnitude of the spot perturbation are matched. Currently equity risk shifts are to be added to the worst performing non-delta position, even if one implies an equity shift in one direction while the other
Page 4 implies an equity shift in the other direct, or even a stagnant underlying market. Please refer Annexure 1 for a detailed example. 2.3 Annex 2 - Sensitivities Based Approach - Draft Accord Section 3(a) General Interest Rate Risk Line 114 (Page 8of 27) With reference to the addition of the allowance of full netting-off of either sensitivities or adjusted cash flows (depending upon the final Standardised Approach standard) at an appropriate instrument level, we would suggest the allowance for netting off at a unique risk driver level as well, in order to avoid capital requirements for perfectly hedged (i.e. no basis risk) exposures. As an example, consider the simple case of a 1 year interest rate swap with quarterly resets; this position is fully hedged through a replication strategy consisting of Forward Rate Agreements (3x6, 6x9, 9x12) with reference to the same floating rate, the hedged position would have no market risk exposure, yet would attract capitalisation (differing instruments). The allowance for netting of exposures that are priced off a unique risk driver (e.g. the Rand Swap curve) would recognise the effective hedging across instruments more appropriately. 3. Conclusion The complexity as well as associated implementation difficulty of the new standardised approach is not a good fit for emerging markets and we would request that the BCBS consider the discretionary use of the current Standardised Approach under local Regulator supervision. We trust that our comments have added value to your deliberations. Should you require any further information, please do not hesitate to contact us. Yours faithfully Gary Haylett General Manager Strategic Projects
Page 5 ANNEXURE 1 Example: Consider a long call option on the equity price of a small emerging market company (bucket 9). To simplify matters, assume there is only one option on a notional of ZAR100.00, and that there are no interest and dividends, and that we operate in a flat volatility environment; the implied volatility being 25% for this particular option. Assume the option matures in 3 months and it is struck at-the-money. The current fair market price of the option is R4.98, and the delta is 0.525. The equity risk capital, according to the formula for paragraph 83 is thus The capital requirement for equity risk is ZAR36.74, which is more than 7 times the value of the option. The following table summarises the non-delta risk: Scenario Spot Volatility Spot Volatility Value Delta Risk P/L 1 50% 20% 150 30% 50.02 26.25 18.79 2 50% -25% 150 19% 50.00 26.25 18.77 3-50% 30% 50 33% 0.00-26.25 21.26 4-50% -10% 50 23% 0.00-26.25 21.26 5 0% 20% 100 30% 5.98-0.99 6 0% -25% 100 19% 3.74 - -1.24 The capital requirement for non-delta risk is ZAR1.24. Added to the capital requirement of ZAR36.74 equals ZAR37.99, which is almost 8 times the value of the option. Clearly, the maximum loss for this particular option is its market value, ZAR4.98. If one was to use the delta implication accounted for by the equity risk, i.e. ZAR36.74 as the delta risk for the option matrix, the non-delta risk would be summarized as follows: Scenario Spot Volatility Spot Volatility Value Delta Risk P/L 1 50% 20% 150 30% 50.02 36.74 81.78 2 50% -25% 150 19% 50.00 36.74 81.76 3-50% 30% 50 33% 0.00 36.74 31.76 4-50% -10% 50 23% 0.00 36.74 31.76 5 0% 20% 100 30% 5.98 36.74 37.74 5 0% -25% 100 19% 3.74 36.74 35.50
Page 6 The worst performing scenarios are 3 and 4, which are both positive. Adding ZAR31.76 to the drawdown for equity risk (ZAR36.74) equals ZAR4.98, which is precisely the premium of the option. Naturally, accounting delta risk in this way is also not a good idea as the option matrix now suggests that delta risk is the only risk (because all the P/L outcomes after delta risk are positive).