Model elements

B.3.1 The IM model should capture all relevant risk factors which materially influence the non-centrally cleared derivatives contracts in a netting set. As a minimum, risk factors should include foreign exchange or interest rate risk, equity risk, credit risk and commodity risk.

B.3.2 The model should appropriately assess other risks if material, arising from imperfect correlations, idiosyncratic risk for credit underlying, market liquidity, and main non-linear dependencies.

B.3.3 The yield curve should be divided into a sufficient number of buckets for capturing interest-rate risk to adequately account for different conditions along the yield curve.

B.3.4 If data is insufficient to capture certain risks accurately an AI may use proxies or approximations in its IM model, provided they are sufficiently conservative.

B.3.5 An IM model may incorporate diversification, hedging, and risk offsets for non-centrally cleared derivatives in the same legally enforceable netting set within one of the following asset classes:

64 As an example, consider a European call option on a single stock. Suppose that an AI agrees to sell a fixed number of shares to another party, the option buyer, at a predetermined price at some specific future date (the contract’s expiry) and the option buyer wishes to buy the shares. Suppose further that the option buyer makes a payment to the AI at the outset of the transaction that fully compensates the option writer for the possibility that it will have to sell shares at contract expiry at the predetermined price. In this case, the AI faces zero counterparty risk while the option buyer faces counterparty risk. The AI has received the full value of the option at the outset of the transaction. The option buyer, on the other hand, faces counterparty risk since the AI may not be willing or able to sell shares to the option buyer at the predetermined price at the expiry of the contract. In this case, the AI would not be obliged to collect any IM from the option buyer and the call option could be excluded from the IM calculation. Since the option buyer faces counterparty risk, the option buyer needs to collect IM from the option writer in a manner consistent with the standards in this module.

Supervisory Policy Manual

CR-G-14 Non-centrally Cleared OTC Derivatives Transactions – Margin and Other Risk Mitigation Standards

V.2 – ConsultationV.1 –

27.01.17

• interest rates and currency⁶⁵;

• equity;

• credit;

• commodities⁶⁶;

• other.

B.3.6 These risk-offsetting features should only be recognised within the same asset class and not across different asset classes.

B.3.7 In cases where the allocation of a particular non-centrally cleared derivative to a specific asset class is not straightforward, an allocation should be made based on whichever asset class represents the preponderance of the derivative’s overall risk profile.

Only non-centrally cleared derivatives that do not fit in any asset class on this basis should be classified as

“other”.

B.3.8 The total IM for a netting set is the sum of IMs calculated separately for each asset class within the netting set.

B.3.9 The IM amount calculated by the model should not be offset by any IM that may be owed by the AI to its counterparty, i.e. IM should be exchanged on a gross basis.

B.4 Model performance^{67, 68}

B.4.1 An AI has to ensure that the data used in the model are subject to a process that ensures their quality.

65 Inflation swaps, which transfer inflation risk between counterparties, are to be considered as part of the interest rates and currency asset class.

66 The asset class “commodities” includes gold and other precious metals such as silver and platinum.

67 This section applies to AIs adopting the internal model approach for calculating IM, regardless of whether the model is developed internally or sourced from a third party (including but not limited to an industry-wide standard model). Even when an AI’s IM model is centrally developed by its group or a third party, the initial and periodic validations should include, inter alia, an evaluation of the applicability of the model to the AI’s relevant derivatives portfolio.

68 With regard to the initial validation and ongoing monitoring of IM model, an AI is allowed to develop its own statistical tests, provided that they are theoretically sound, well-documented and consistently applied.

Supervisory Policy Manual

CR-G-14 Non-centrally Cleared OTC Derivatives Transactions – Margin and Other Risk Mitigation Standards

V.2 – ConsultationV.1 –

27.01.17

B.4.2 An AI should have written policies and procedures in relation to the recalibration of the IM model, including:

(i) that the AI should recalibrate its internal model regularly but at least once a year;

(ii) that the AI should ensure that the data used to recalibrate the model incorporates a period of significant financial stress subject to paragraph B.2.5;

(iii) the circumstances that would trigger an earlier recalibration than stipulated under (i);

(iv) procedures for adjusting the margin amount to be exchanged resulting from the recalibration of the model in response to changing market conditions. These procedures should ensure that an AI agrees on a specific period with its counterparties during which the additional IM resulting from the recalibration of the model has to be posted.

B.4.3 The performance of the model should be monitored on a continuous basis which includes testing the model’s assessments against realised data and experience.

Adequate documentation of backtesting results should be maintained.

B.4.4 An AI’s policies and procedures should describe the methodologies used for backtesting and the results which would necessitate a recalibration of the model.

B.4.5 The IM model should be subject to initial and periodical validation to validate its conceptual soundness and the applicability of the model to the derivatives for which it is being used. The validation should also include a review of the IM model in light of developments in the financial markets and advances in model techniques.

B.4.6 Initial validation should be conducted before the model is implemented or whenever any significant changes to the model are made. Periodical validation should be conducted at least once a year.

B.4.7 An AI should take remedial action in a timely manner if the validation reveals any material problems with the model. The MA should be notified as soon as the AI becomes aware of any material problems.

Supervisory Policy Manual

CR-G-14 Non-centrally Cleared OTC Derivatives Transactions – Margin and Other Risk Mitigation Standards

V.2 – ConsultationV.1 –

27.01.17

B.4.8 The IM model should be validated by an independent qualified party. Validation by an internal party is acceptable if the internal party is not involved in derivatives trading or the development and operations of the IM model and has the required knowledge and expertise to carry out the validation.