Response to consultation on Regulatory Technical Standards on the authorisation of initial margin models
1. Are there additional elements of either pro forma based models or models not based on pro forma models that these draft RTS should identify validation techniques for?
See attached document for more details
[CG0.1] We propose that the RTS formally recognizes that a pro forma model consists of both a methodology for IM calculation and a governance framework for the maintenance and monitoring of the IM model and its performance.
The current draft RTS does not recognise that a pro forma model should have a robust governance framework to maintain the model and ensure regulatory-compliant IM amounts are exchanged. Furthermore, it does not set out validation approaches that take into account that the primary means of monitoring and addressing margin coverage issues should be the established governance framework in the context of the relevant pro forma model.
The application of the SIMM governance framework would remove the need for unilateral, entity‑level remediation actions as set out in Articles 18, 19, 21 and 24, and instead support remediation at the bilateral level in a practical manner, consistent with the existing contractual framework.
[CG0.2] We request clarification on the use of the term “remediation” within the RTS, and recommend distinguishing clearly between its use in different contexts.
The RTS contains multiple references to “remediation”, including in Articles 8, 9 and 12 (in the context of audit, validation and IT matters), and in Articles 21 and 23 (in relation to model parameters and proxies). However, the term is not defined and appears to be used to describe a range of potentially distinct actions. This creates a risk of inconsistent interpretation across firms and jurisdictions.
Under the ISDA SIMM Governance Framework, “remediation” has a specific and well-established meaning. It refers to the bilateral process whereby counterparties agree the exchange of additional initial margin following backtesting results that have been validated and deemed to breach predefined thresholds. This is a formal, outcome-based process with clear triggers, governance and financial consequences, and we recommend that this established usage be preserved.
Where the RTS intends to refer to other forms of follow-up action such as:
- further thinking or analysis; or
- internal escalation; or
- engagement with a model provider;
We recommend either clarifying the definition of “remediation” in the RTS or adopting a distinct term instead of remediation to avoid conflating these activities. Clarifying the use of the term would reduce the ambiguity in supervisory expectations, support consistent implementation across firms, and better align the RTS with established industry governance frameworks.
[CG0.3] We request confirmation that none of the provisions in Chapter 3 are relevant to a pro forma model.
Please explicitly confirm that the provisions in Chapter 3 are solely provided for the benefit of the Competent Authorities (CAs) to assess a non-pro forma model and should not be applied to CA consideration of a pro forma model since assessment of the pro forma model will already have been conducted by the EBA. In addition, we would appreciate confirmation that Chapter 3 provisions are not directly relevant to the EBA’s assessment of a pro-forma model; instead that assessment is governed solely by the provisions in the EBA Decision EBA/DC/610[1] concerning arrangements for the Initial Margin Model Validation function under EMIR for ISDA SIMM, January 27, 2026.
[1] EBA DC 610 - Decision on arrangements for ISDA SIMM validation.pdf
2. Are there elements of either pro forma based models or models not based on proforma models that these draft RTS should include to add further proportionality in the supervisory procedures used for authorisation?
See attached document for more details
We do not consider that there are additional elements specific to either pro forma–based models or models not based on pro forma approaches that need to be incorporated into the draft RTS to enhance proportionality in supervisory authorisation procedures.
The existing framework is already sufficiently flexible to accommodate differences in modelling approaches, and further differentiation is not necessary to achieve a proportionate supervisory process.
3. Do you have any comments on the qualitative requirements for pro forma models?
See attached document for more details
[CA7.1] We propose that Article 7(a) be revised to make this a responsibility of the counterparty and not the model development unit, which affords flexibility for firms in their implementation of the control
The current draft is too prescriptive regarding the firm-level allocation of responsibility for controls over segregation of duties. There may be different established controls or responsibilities for the control over global model change at each EU counterparty, that may not be readily altered to meet a jurisdiction-specific vision of a model development unit.
[CA7.2] We propose that Article 7(c) be deleted
Article 7(c) seems to duplicate Article 6. We suggest that the responsibility for adequate staffing is retained in Article 6 but deleted in Article 7(c) as Senior management would oversee the sufficiency of staffing overall.
[CA7.3] We propose that Article 7(f) be deleted
The meaning of 'the risk unit' is not defined but in practice, ensuring the integrity of the data inputs on a daily basis may be overseen in functions other than the model development unit. In addition, the nature of the reports and how they might be analysed is vague, and it is unclear how the outcome of such analyses will be evidenced to Competent Authorities (CAs).
Abnormal outputs or swings may be reviewed as part of the business as usual process but that is across margin operations and model owning teams, and dispute resolution is an important part of this.
[CA10.1] We propose that Article 10(c)(i) be revised to require that the counterparty has a documented process for the investigation of material discrepancies including identification of root cause, which may consider (and not require consideration of) discrepancies in trades, mapping of risk factors, valuation methods, or calculation of sensitivities.
The article requires that the source of the discrepancy is identified down to the mapping of risk factors, valuation methods and calculation of sensitivities. This level of investigation is only possible where both parties use the same third-party vendor capable of easily reconciling all inputs to the IM calculation (e.g. Acadia). However, not all market participants subscribe to these services. The SIMM Common Risk Interchange Format files are also generally not exchanged.
[CA13.1] We propose in Article 13(1) that if the continued use of a model based on a proforma model framework is no longer viable, then the counterparty require adherence to existing contractual arrangements, which would require appropriate notice provisions and an agreed adjustment period to switch initial margin methodologies
It is important to recognise that the initial margin (IM) calculation methodology is typically agreed between counterparties within the bilateral Credit Support Annex (CSA). As such, any change to the IM methodology is not purely a technical or supervisory matter, but also a contractual one.
We therefore propose that competent authorities explicitly acknowledge that a transition from one IM calculation method to another would require adherence to existing contractual arrangements. In practice, this would involve appropriate notice provisions and an agreed adjustment period to allow counterparties to implement the change in a controlled and orderly manner.
Such clarification would ensure that regulatory expectations are aligned with market practice and legal frameworks, and would help to avoid unintended disruption arising from changes implemented without sufficient time for operational, legal, and risk management adjustments.
See attached document for more details
Accordingly, we recommend that any requirement or expectation relating to changes in IM methodology should explicitly take into account existing CSA agreements and provide for a reasonable transition period, subject to bilateral agreement between counterparties.
[CA13.2] We request clarification on Article 13(1) regarding the expectations for contingency planning, specifically that these should not be interpreted as requiring firms to hold capital or liquidity reserves equivalent to the potential difference between a pro forma model and an alternative approach.
We recommend that the RTS clarify expectations regarding contingency planning in this context. While firms should maintain appropriate operational, legal, and governance readiness to implement a transition to an alternative IM methodology if required, this should not be interpreted as requiring firms to pre-fund or hold capital or liquidity reserves equivalent to the potential difference between a pro forma model and an alternative approach (e.g. the standardised grid).
Accordingly, any reference to contingency expectations should be understood as referring to being prepared in terms of processes, infrastructure, and contractual readiness, rather than a requirement to hold contingent liquidity or capital buffers. Clarifying this point would avoid unintended interpretations that could impose disproportionate liquidity burdens on firms without enhancing prudential outcomes.
[CA13.3] We propose that Article 13(2) be revised to allow for a degree of flexibility in how such outsourcing arrangements are defined and documented where elements of a pro forma IM model are outsourced to affiliates within the same group.
While it is appropriate for the counterparty to retain oversight of outsourced model components, in a group context the allocation of responsibilities across affiliates may evolve over time. In practice, however, governance remains stable where functional reporting lines are consistent and senior management (or the relevant management body or risk committee) receives regular and comprehensive management information (MI) on delegated activities.
Accordingly, we recommend that the RTS recognise that, in cases where model management is conducted at a group level, outsourcing arrangements may be articulated at a level that accommodates this operational flexibility, provided that effective oversight, accountability, and reporting are maintained.
We suggest that Article 13.2.(d) is amended to delete the reference to design and calibration as follows:
“Where a counterparty has outsourced some or all important or critical functions, activities or services related to the design,calibration, implementation, internal validation and audit of a pro forma model and subsequent model and calibration updates or changes, competent authorities shall verify that the outsourcing does not hinder the application of the assessment methodology referred to in this Regulation […].”
The outsourcing to a model provider of design and calibration is already captured in Article 13(1), therefore removing these elements from Article 13(2) clarifies the counterparty’s differing obligations towards model providers and other vendors.
4. Do you have any comments on the quantitative requirements for pro forma models?
See attached document for more details
[CA18.1] We propose that Article 18(c) be re-written to recognise that the governance for a pro forma model should be designed to identify and remediate risk not in the model that may be driving material shortfalls in portfolios.
Article 14(2) of Delegated Regulation (EU) No 2016/2251 is concerned with ensuring that initial margin is sufficient at the portfolio level, rather than requiring each individual risk factor to be margined precisely in isolation.
Under section 5 of the SIMM Governance Framework, material shortfalls arising from risks not captured in SIMM are already addressed through bilateral remediation where remediation thresholds are breached, using a range of recognized remediation options, including the exchange of additional initial margin or the margining of the relevant risk factor outside SIMM (e.g. via the Grid).
In its current form, Article 18(c) does not allow firms sufficient flexibility to determine the most appropriate remediation approach within this established framework.
[CA19.1] We propose that Article 19(b) be re-written to not introduce additional requirements mandating the transfer of entire netting sets to a standardised calculation approach where the governance of the pro forma model is designed to identify and remediate insufficiencies.
Introducing additional requirements mandating the transfer of entire netting sets to a standardised calculation approach would be unnecessary given the effectiveness of existing monitoring and remediation processes, and would introduce practical and operational challenges.
In particular, transitioning to a standardised calculation requires counterparty consent and cannot be implemented unilaterally, limiting its feasibility in a bilateral IM framework.
We propose that the RTS recognise that the existing SIMM backtesting processes, including both historical and realised PnL backtesting, are sufficient to identify modelling issues, including those arising from non‑linearities.
Where such issues are identified, the ISDA SIMM Governance Framework already provides a robust and well‑established remediation mechanism to address material deficiencies.
[CA20.1] We propose that Articles 20(a) and 20(b) be deleted, as this requirement is already addressed by the EBA’s approval of the pro-forma model incorporating a 10-day MPOR.
The regulatory rule referenced in Article 20 requires the Margin Period of Risk (MPOR) to be at least ten days, but does not prescribe any cases where a longer period may be required.
We believe that ten days is already a conservative assumption, and in addition, the presence of the concentration threshold in the SIMM model provides additional mitigation to account for positions that may be difficult to liquidate over the ten-day period. As a result, there should be no need for firms to put in additional mitigation measures to cater to longer liquidation periods.
[CA21.1] We propose that Article 21(2) be deleted
Article 21(2) extends beyond the requirements set out in Article 16(2) and Article 16(4) of Delegated Regulation (EU) No 2016/2251. Article 16(2) requires calibration data to include a stress period, and Article 16(4) requires that the stress period used for calibration be identified and applied separately, at least by asset class.
Neither provision requires the calibration stress period to be representative of every individual portfolio. Accordingly, portfolio-level deviations from the calibration stress period do not, in themselves, undermine the soundness or regulatory compliance of the model.
Model performance at portfolio level is already addressed through the SIMM governance framework, including the 1+3 Backtest and the Actual P&L Test, with clear remediation mechanisms where thresholds are breached. A literal reading of Article 21(2) would instead imply firm-level recalibration of an industry proforma model based on individual portfolio information and/or remediation outside the established governance framework.
Such an interpretation would raise several concerns:
Operational impracticality: A single firm cannot mandate the recalibration of a globally applied pro forma model solely due to differences between its portfolio stress characteristics and the calibration stress period.
Inconsistent margin outcomes: Requiring recalibration or additional margin collection for each portfolio-level deviation would lead to variability and unpredictability in margin outcomes, even where existing margin levels remain adequate.
- Undermining model consistency: This approach would erode the consistency and comparability of a globally calibrated initial margin model.
[CA23.1] We propose that in Article 23(b) the authorities clarify that the use of a proxy methodology is considered acceptable where it represents a reasonable and appropriate method, as evidenced through existing backtesting methodology.
Article 23(b) appears to complement Article 16(10)(b) of Delegated Regulation (EU) No 2016/2251, with the objective of ensuring that the use of proxies results in a conservative level of initial margin. The sufficiency of initial margin outcomes is already assessed through the established 1+3 Backtesting and Actual P&L Test.
However, there is a lack of clarity as to what constitutes a “conservative” level of margin in this context. It may therefore be helpful for competent authorities to clarify that a proxy methodology is considered acceptable where it represents a reasonable and appropriate method, as evidenced through existing backtesting methodology. Such clarification would promote consistency in supervisory expectations while avoiding the introduction of duplicative or subjective assessments of what constitutes a “conservative” level of margin.
[CA24.1] We propose that Article 24(1)(f)(i) specify that as part of the robust process to assess results of backtesting, overshooting analysis should be conducted at an aggregated or portfolio level rather than at the level of individual exceedances
There is a concern that Article 24(1)(f)(i) could be interpreted as mandating a detailed assessment of every individual overshooting event, which would be impractical, disproportionate and inconsistent with a risk-based supervisory approach focused on identifying material weaknesses.
By way of illustration, a firm may report approximately 240 portfolios in a 1+3 backtesting exercise, with no Red portfolios, 12 Amber portfolios, and 228 Green portfolios. Across the Green portfolios alone, this could result in up to 5,016 overshooting events (228 portfolios × 22 exceedances) that would technically require analysis, despite the absence of any indication of model underperformance. Requiring assessment of all such events would not generate meaningful supervisory insight and would divert resources away from the investigation of genuinely material issues.
[CA24.2] We propose that Articles 24(1)(f)(iii) and 24(1)(g) be assessed at the industry level and not at the counterparty level
This proposal is based on the following observations:
The fact that the realised number of netting sets classified as yellow or red exceeds the expected number does not, in itself, demonstrate that the model is not performing as intended.
- The presence of correlations across portfolios means that outcomes are not independent
- Any attempt to assess and remediate perceived deviations at the counterparty level would be fundamentally constrained. Remediation actions would necessarily rely on unilateral analysis based on firm-specific data, which counterparties cannot independently validate. This undermines the robustness, consistency, and credibility of any such remediation process.
Taken together, these factors indicate that the proposed framework is not well suited to generating reliable or decision-useful insights when applied at the counterparty level. Instead, any meaningful interpretation of these statistics requires aggregation across a sufficiently broad dataset, where interdependencies can be properly accounted for.
Accordingly, we recommend that:
- The assessment of realised versus expected classifications should be performed at the industry level, where the analysis is statistically meaningful; and
- Article 24(1)(g) should be aligned accordingly, such that any consideration of remedial actions is based on systematic, industry-level patterns, rather than counterparty-level outcomes.
- The authorities recognise that, where bilaterally validated shortfalls breach defined thresholds under the SIMM governance framework, there are established processes requiring firms to investigate the issue and exchange additional initial margin on a bilateral basis.
Without such a clarification, the requirements risk creating operational burden without improving supervisory insight, and may lead to inconsistent and unimplementable remediation expectations.
[CA24.3] We request clarification on the meaning of “…defined appropriate actions in case…” in Article 24(1)(g).
There is a risk that this phrase may be interpreted as requiring concrete remedial actions beyond analytical assessment. However, our understanding is that such events should primarily serve as triggers for further analysis and consideration, rather than automatically necessitating specific actions (such as the exchange of additional initial margin).
See attached document for more details
To avoid ambiguity, we recommend that the RTS explicitly clarify that “appropriate actions” in this context refer to the initiation of analysis and internal review, rather than a prescriptive requirement to implement specific remedial measures. Such clarification would help ensure consistent interpretation across counterparties and competent authorities, and avoid the risk that analytical triggers are inadvertently treated as automatic triggers for additional margin exchange.
[CA24.4] We propose that the requirement in Article 24(1)(h) to use the Margin Average Shortfall (MAS) metric be replaced with an approach aligned with the SIMM Green Shortfall (SGS), as defined under the ISDA SIMM Governance Framework.
The use of the MAS metric goes beyond the requirements of the rules, with the authorities suggesting that firms need to take action based on aggregate performance of portfolios. This is not consistent with established approaches for assessing the adequacy of initial margin models that have been designed in line with the regulatory standards. The Margin Average Shortfall (MAS) metric, while useful in other contexts is out-of-scope and out-of-place for assessing the performance of an IM model against the regulations.
By contrast, the SIMM Green Shortfall (SGS), as defined in section 6 of the Remediation Annex to the ISDA SIMM Governance Framework, provides a more appropriate measure. It is a robust and intuitive measure that can be broadly applied to assess initial margin models, and not only SIMM. It is a well-understood measure which respects statistical significance and the BCBS traffic-light system. It is well known to the industry and the community of global regulators and reflects the amount of additional margin which is required to make the portfolio comply with the rule.
More broadly, the RTS should recognise the existing governance framework rather than prescribing a specific metric. The SIMM Governance Framework already provides a well‑established and effective approach for monitoring model performance, including clearly defined criteria for identifying and addressing under‑margining. Embedding this framework would promote consistency and avoid introducing additional or potentially conflicting requirements.
[CA24.5] We propose that Red/Yellow Threshold stated as 0.1% in Article 24(1)(f)(ii)(b) be corrected to 0.01%.
There is a suggestion that there may be an error in the Red/Yellow Threshold in the RTS. In the previous draft RTS, a 99.99% threshold (i.e., 0.01%) was used, consistent with the market risk capital regulation.
We note that the existing definition of the red/yellow threshold is very long-standing, going back to a definitive BCBS document from 1996 entitled “SUPERVISORY FRAMEWORK FOR THE USE OF “BACKTESTING” IN CONJUNCTION WITH THE INTERNAL MODELS APPROACH TO MARKET RISK CAPITAL REQUIREMENTS (January 1996)[1]”, and is very established throughout the regulatory community and industry firms. The original definition specifies that “the red zone begins at the point where the cumulative probability equals or exceeds 99.99%”.
[1] https://www.bis.org/publ/bcbsc223.pdf
5. Do you have any comments on the qualitative requirements for models not based on pro forma models?
We do not have any comments on the qualitative requirements for models not based on pro forma models.
6. Do you have any comments on the quantitative requirements for models not based on pro forma models?
See attached document for more details
[CA46.1] We request that Article 46(2)(c), (d), (f) and (h) be deleted:
In line with [CA21.1], Article 46(h) extends beyond the requirements set out in Article 16(2) and Article 16(4) of Delegated Regulation (EU) No 2016/2251. Article 16(2) requires calibration data to include a stress period, and Article 16(4) requires that the stress period used for calibration be identified and applied separately, at least by asset class.
Regarding Articles 46(c), (d) and (f), adopting a more streamlined approach would not be meaningful, as any perceived simplification would ultimately require additional granular analysis to satisfy the underlying requirements, thereby negating any practical benefit and introducing unnecessary complexity and duplication of effort.