Response to consultation Paper on draft RTS on criteria for assessing risk factors modellability under the IMA
Go back
We want to emphasize that while in the current state our risk model in many cases shocks inputs used to calibrate such parameterized curves, surfaces and cubes, we are considering using these parameterisations more systematically in our risk model in future. We therefore consider the regulatory provision referred to highly relevant.
We think Option 1 is conceptually valid and aligns with the spirit of NMRF as outlined in the Basel text. For some types of risk factors, we think Option 1 is elegant approach as it tends to avoid arbitrage situation. However, we acknowledge that it also is operationally intensive and may therefore not be suitable for all risk factors (for further details see Question 9).
See Appendix 1 for a detailed example that outlines our understanding of Option 1. By way of this example, we provide evidence of how the guidelines produced by Basel and the Regulatory Technical Standards defined by EBA could actually be implemented. The example could also serve to further highlight to the Regulatory community the commonalities of requirements between the FRTB NMRF guidelines (to prove the “modellability” of a risk factor) and the IFRS13 requirements on “observability” of market data inputs. We consider the terms modellability and observability as well as risk factor and market data inputs almost synonyms from conceptual perspective.
Wording of the option 2 is quite ambiguous therefore we find it difficult to respond to it. One interpretation of the wording implies building an alternative model that takes as inputs, the redefined risk factors. Additionally this will lead to divergence between End-Of-Day pricing/risk management and the Internal Model Approach. This, in turn can potentially have an adverse impact on the PLA test. If internal model risk factors were to be redefined in accordance with Option 2, while daily risk management remains based on the parametric approach, it will be in violation of the spirit of the qualitative standards, expressed both in the Basel text (MAR 30.10(3)) and CRR2 (Articles 325bi-1(f) ).
We would appreciate further dialogue and examples on what the EBA’s proposition is.
Give the challenges raised in the text above we re-emphasize that neither Option 1 nor Option 2 would work in all required circumstances. We therefore propose a complementary framework, which adds a 3rd option (“Option 3”). See Q10 for further details.
If {x1,x2,x3,x4} are the “output risk factors” chosen to discretize the curve, surface or cube, then the pricing function ϕ(a,b,c) has to be replaced by an equivalent pricing function of the form Ψ(x_1,x_2,x_3,x_4) . Otherwise, it would be impossible to unshock x1 separately from the other risk factors in the ES, or shock it separately from the other in the SSRM, should it be NMRF.
Ultimately, the build of new pricing functions makes the parametric function almost useless in the risk engine. Furthermore, if internal model risk factors were to be redefined in accordance with Option 2, while daily risk management is still based on the parametric approach, it will be in violation of the spirit of the qualitative standards, expressed both in the Basel text (MAR 30.10(3) ) and CRR2 (Articles 325bi-1(f) ).
B) The implementation of Option 1, largely depends on the IMA model framework employed in the institution and on the specific nature of the function parameters. Specifically Option 1 requires recalibration of historical parameters, as highlighted already by EBA in describing the option itself.
A necessary pre-requisite to recalibration of historical parameters is the availability of historical market information. For some risk factors, industry practice is to store the historical data in the form of function parameters instead of market information (e.g. SABR parameters for interest rate volatilities). In such cases, it will be impractical to re-calibrate the function parameters based on RFET qualifying market information. If historical market information (e.g. for the underlying trade prices or implied volatilities, rather than model parameters) is available, however, Option 1 is feasible if the implementation of the IMA model in an institution allows for “on-the-fly” calibration of function parameters which will be challenging to implement. In such a case Option 1 will not be a practical approach as:
it will force institutions to maintain multiple timeseries for the same set of parameters
the entire history of time series will need to be regenerated every quarter based on the current modellability / non-modellability
From an operational standpoint, the marking of a set of parameters {a,b,c} is driven by the market information available at a given point in time, and completed if needed by human expertise. Stripping a set of alternative parameters {a’,b’,c’} only from RFET qualifying data is possible only if a full history of RFET qualifying data (since 2007) is still available and if the human expertise is replaced by some algorithmic intelligence to solve operational issues.
The modellability of the underlying instrument buckets evolve through time. If N is the number of buckets supporting the modellability of the parameters {a,b,c}, then there are 2^N-1 versions of possible alternative sets {a’,b’,c’} to maintain.
We recommend (as indicated in the response to Q8), to allow Institutions to use more than one approach. We share the Industry view in proposing Option 3 as additional option based on the argument that the principle for assessing modellability should be aligned to the spirit of the Basel text - the modellable parameter is one that could in principle, be calibrated solely by reference to modellable buckets. In other words the approach would be to allow modellability of function parameters to be assessed independently of each other. For instance, the ATM volatility parameter plays a central role in the calibration of a SABR model. Assessing its modellability based on the ATM bucket makes sense, whether or not DITM bucket passes the RFET.
For further clarity, we illustrate by means of a stylised example in Appendix 2. By way of this example, we provide evidence of how the guidelines produced by Basel and the Regulatory Technical Standards defined by EBA could actually be implemented.
We would also like to take the opportunity to comment on Article 3: the criteria on how a verifiable price shall be considered representative of a risk factor.
A verifiable price shall be considered representative of a risk factor as of its observation date where both the following conditions are met:
(a) the institution has demonstrated that there is a close relationship between the risk factor and the verifiable price;
(b) the institution has specified a conceptually sound methodology to extract the value of the risk factor from the verifiable price. Any input data or risk factor used in that methodology other than that verifiable price shall be based on objective data.
While the second criteria can be easily applied to vanilla products, it can be particularly difficult to apply to exotic derivatives. For example, a swap rate can be relatively easily implied from an interest rate swap; however, it is difficult to imply parametric volatility from a barrier option. More particularly the price of an exotic derivative is a non-linear function of multiple market data inputs/parameters hence there could be many-to-one mapping between function parameters and price making it extremely challenging to infer the level of the function parameter(s) from the price. Beside the value (level) of the same risk factor derived from different exotic products for the same observation dates could be materially different. Finally the price of a transaction is also dependent on non-model considerations for example client relationships; extracting just the model related component of such risk factor could be extremely difficult.
We propose the following wording:
A verifiable price shall be considered representative of a risk factor as of its observation date where:
(a) the institution has demonstrated that there is a close relationship between the risk factor and the verifiable price;
(b) where possible the institution has specified a conceptually sound methodology to extract the value of the risk factor from the verifiable price. Any input data or risk factor used in that methodology other than that verifiable price shall be based on objective data, e.g. market data inputs/risk factors used in the daily market to market.
c) where it is not possible to extract the value of the risk factor from the verifiable price the Institution shall demonstrate that the value of the risk factor on the observation date used in the risk model and the risk P&L thereof can be reconciled to the economic P&L
Q5. Do you see any problems with requiring that institutions are allowed to use data from external data providers as input to the modellability assessment only where the external data providers are regularly subject to an independent audit (independent of whether the price is shared with the institution or not)? If so, please describe them thoroughly (i.e. for which data providers and the reasons for it).
NoQ6. Do you have any proposals on additional specifications that could be included in the legal text in order to ensure that verifiable prices provided by third-party vendors meet the requirements of this Regulation?
No further proposals in addition to those included in the other responses to this questionnaireQ7. How relevant are the provisions outlined above for your institution? How many and which curves, surfaces or cubes are (planned to be) represented by a mathematical function with function parameters chosen as risk factors in your (future) internal model?
Yes we confirm the relevance of the provision for our institution. Currently we use parametric functions in particular for the volatility modelling for Interest Rates, Foreign Exchange, Equities and Commodities. Models we use are mainly SABR, local volatility and models where IV is parametrized by ATM Volatility, Skew and Smile functions.We want to emphasize that while in the current state our risk model in many cases shocks inputs used to calibrate such parameterized curves, surfaces and cubes, we are considering using these parameterisations more systematically in our risk model in future. We therefore consider the regulatory provision referred to highly relevant.
Q8. Do you have a preference for any of the options outlined above? For which reasons? Please motivate your response.
We appreciate the flexibility provided in the RTS when assessing modellability of function parameters. This is aligned with our view that restricting possible approaches to one specific option is counter-productive. Rules for general eligibility criteria should be adaptive to the specific nature of parameters and the implementation of the IMA model in an institution. We would welcome an approach where multiple options are provided in the final RTS allowing a bank to model a modellable risk factor using the best available data, in line with MAR21.23.We think Option 1 is conceptually valid and aligns with the spirit of NMRF as outlined in the Basel text. For some types of risk factors, we think Option 1 is elegant approach as it tends to avoid arbitrage situation. However, we acknowledge that it also is operationally intensive and may therefore not be suitable for all risk factors (for further details see Question 9).
See Appendix 1 for a detailed example that outlines our understanding of Option 1. By way of this example, we provide evidence of how the guidelines produced by Basel and the Regulatory Technical Standards defined by EBA could actually be implemented. The example could also serve to further highlight to the Regulatory community the commonalities of requirements between the FRTB NMRF guidelines (to prove the “modellability” of a risk factor) and the IFRS13 requirements on “observability” of market data inputs. We consider the terms modellability and observability as well as risk factor and market data inputs almost synonyms from conceptual perspective.
Wording of the option 2 is quite ambiguous therefore we find it difficult to respond to it. One interpretation of the wording implies building an alternative model that takes as inputs, the redefined risk factors. Additionally this will lead to divergence between End-Of-Day pricing/risk management and the Internal Model Approach. This, in turn can potentially have an adverse impact on the PLA test. If internal model risk factors were to be redefined in accordance with Option 2, while daily risk management remains based on the parametric approach, it will be in violation of the spirit of the qualitative standards, expressed both in the Basel text (MAR 30.10(3)) and CRR2 (Articles 325bi-1(f) ).
We would appreciate further dialogue and examples on what the EBA’s proposition is.
Give the challenges raised in the text above we re-emphasize that neither Option 1 nor Option 2 would work in all required circumstances. We therefore propose a complementary framework, which adds a 3rd option (“Option 3”). See Q10 for further details.
Q9. Do you consider any of the options outlined above as impossible or impractical? For which reasons? Please motivate your response.
A) Option 2 is materially impractical. See response to Q8 as well as further details provided below. Option 2 requires alternative (parametric) data models to be built in the risk engineIf {x1,x2,x3,x4} are the “output risk factors” chosen to discretize the curve, surface or cube, then the pricing function ϕ(a,b,c) has to be replaced by an equivalent pricing function of the form Ψ(x_1,x_2,x_3,x_4) . Otherwise, it would be impossible to unshock x1 separately from the other risk factors in the ES, or shock it separately from the other in the SSRM, should it be NMRF.
Ultimately, the build of new pricing functions makes the parametric function almost useless in the risk engine. Furthermore, if internal model risk factors were to be redefined in accordance with Option 2, while daily risk management is still based on the parametric approach, it will be in violation of the spirit of the qualitative standards, expressed both in the Basel text (MAR 30.10(3) ) and CRR2 (Articles 325bi-1(f) ).
B) The implementation of Option 1, largely depends on the IMA model framework employed in the institution and on the specific nature of the function parameters. Specifically Option 1 requires recalibration of historical parameters, as highlighted already by EBA in describing the option itself.
A necessary pre-requisite to recalibration of historical parameters is the availability of historical market information. For some risk factors, industry practice is to store the historical data in the form of function parameters instead of market information (e.g. SABR parameters for interest rate volatilities). In such cases, it will be impractical to re-calibrate the function parameters based on RFET qualifying market information. If historical market information (e.g. for the underlying trade prices or implied volatilities, rather than model parameters) is available, however, Option 1 is feasible if the implementation of the IMA model in an institution allows for “on-the-fly” calibration of function parameters which will be challenging to implement. In such a case Option 1 will not be a practical approach as:
it will force institutions to maintain multiple timeseries for the same set of parameters
the entire history of time series will need to be regenerated every quarter based on the current modellability / non-modellability
From an operational standpoint, the marking of a set of parameters {a,b,c} is driven by the market information available at a given point in time, and completed if needed by human expertise. Stripping a set of alternative parameters {a’,b’,c’} only from RFET qualifying data is possible only if a full history of RFET qualifying data (since 2007) is still available and if the human expertise is replaced by some algorithmic intelligence to solve operational issues.
The modellability of the underlying instrument buckets evolve through time. If N is the number of buckets supporting the modellability of the parameters {a,b,c}, then there are 2^N-1 versions of possible alternative sets {a’,b’,c’} to maintain.
Q10. Do you have alternative proposals to define the consequence on the modellability of the parameters where some buckets of a curve, surface or cube are modellable whilst others are nonmodellable?
We believe that the approach here should be principles-based rather than prescriptive, in line with the requirement for real prices to be representative of risk factors in MAR31.15.We recommend (as indicated in the response to Q8), to allow Institutions to use more than one approach. We share the Industry view in proposing Option 3 as additional option based on the argument that the principle for assessing modellability should be aligned to the spirit of the Basel text - the modellable parameter is one that could in principle, be calibrated solely by reference to modellable buckets. In other words the approach would be to allow modellability of function parameters to be assessed independently of each other. For instance, the ATM volatility parameter plays a central role in the calibration of a SABR model. Assessing its modellability based on the ATM bucket makes sense, whether or not DITM bucket passes the RFET.
For further clarity, we illustrate by means of a stylised example in Appendix 2. By way of this example, we provide evidence of how the guidelines produced by Basel and the Regulatory Technical Standards defined by EBA could actually be implemented.
Q11. Do you intend to apply paragraph 4? If so, for which risk factors will it be relevant? Do you expect any implementation issues related to it? Please explain expected issues thoroughly.
At present we do not rule out the option to apply the paragraph 4.Q12. Do you agree with the outlined methodology for the assessment of modellability of risk factors? If not, please explain why.
As a general principle, we favour a framework where Institutions are allowed to leverage off a flexible framework which allows the use of various options according to the specific nature of the function parameters. These can then adapt to requirements arising from their implementation of the internal model with reference to the parametric risk factors. We have explained our arguments throughout our response, by also means of illustrative example to Q8, Q9 and Q10.We would also like to take the opportunity to comment on Article 3: the criteria on how a verifiable price shall be considered representative of a risk factor.
A verifiable price shall be considered representative of a risk factor as of its observation date where both the following conditions are met:
(a) the institution has demonstrated that there is a close relationship between the risk factor and the verifiable price;
(b) the institution has specified a conceptually sound methodology to extract the value of the risk factor from the verifiable price. Any input data or risk factor used in that methodology other than that verifiable price shall be based on objective data.
While the second criteria can be easily applied to vanilla products, it can be particularly difficult to apply to exotic derivatives. For example, a swap rate can be relatively easily implied from an interest rate swap; however, it is difficult to imply parametric volatility from a barrier option. More particularly the price of an exotic derivative is a non-linear function of multiple market data inputs/parameters hence there could be many-to-one mapping between function parameters and price making it extremely challenging to infer the level of the function parameter(s) from the price. Beside the value (level) of the same risk factor derived from different exotic products for the same observation dates could be materially different. Finally the price of a transaction is also dependent on non-model considerations for example client relationships; extracting just the model related component of such risk factor could be extremely difficult.
We propose the following wording:
A verifiable price shall be considered representative of a risk factor as of its observation date where:
(a) the institution has demonstrated that there is a close relationship between the risk factor and the verifiable price;
(b) where possible the institution has specified a conceptually sound methodology to extract the value of the risk factor from the verifiable price. Any input data or risk factor used in that methodology other than that verifiable price shall be based on objective data, e.g. market data inputs/risk factors used in the daily market to market.
c) where it is not possible to extract the value of the risk factor from the verifiable price the Institution shall demonstrate that the value of the risk factor on the observation date used in the risk model and the risk P&L thereof can be reconciled to the economic P&L