Response to consultation Paper on draft RTS on back-testing and PLA attribution requirements
Go back
The industry of the provision in paragraph 5(a)) “that adjustment is computed this way due to its nature” ambiguous or not clear, and seeks more clarity in order to implement in practice. The Industry’s interpretation was that adjustments that aggregate linearly would not meet this criteria (e.g. an IPV adjustment) and adjustments that reflect non-linear effects would meet this criteria (e.g. a delta bid-offer calculation on a portfolio of derivatives).
In reality, a bank may choose to take adjustments at different levels of reporting hierarchy in a way that reflects how it seeks to manage its business. This may result in adjustments being made at a higher level than the FRTB desk. Requiring banks to calculate the adjustment at a lower level only for the purpose of regulatory back-testing would be an additional process to adjust the actual P&L to include a factor not in the desks' reported P&L. This is undesirable, not just due to the additional process but also due to the lack of business use case that would render this allocation liable to error.
The majority of industry respondents confirm that client margins and commissions are taken into account in the end-of-day valuation process and primarily not in the actual P&L. The overwhelming number of respondents confirm Day-one profits/losses are taken into account in both the end-of-day valuation and also in the actual P&L. As a general practice, the industry does not include client margin, commissions or day-one profits/losses in their hypothetical P&L.
The industry acknowledges that client margins corresponds to the P&L generated at the initiation of the transaction. New transactions are part of the EOD Valuation process and Actual P&L but not part of the Hypothetical P&L (CRR Article 366(3)). Currently the margin generated at the initiation of the trade is determined on a declaratory basis. Consequently, day-one profits/losses reserve is computed from the declared client margin.
The industry also recognizes that risk commission is included in the actual P&L for back-testing by some banks. The industry defines the risk commission as a commission on a transaction where a bank agrees to procure an instrument for a client with a price agreed up front. The bank takes on the risk on entering the market (committing capital to fill the order) that the price is different than that agreed with the client.
However, industry members noted that the definition for day-one and client margin P&L may differ across institutions, therefore would urge EBA to retain some level of flexibility in the context of back-testing and P&L attribution. Although current firm-wide definition is subject to ongoing governance and control and often subject to scrutiny by local regulators/supervisors, as part of the Basel II back-testing process. Therefore, industry is in favour of some degree of consistency with the current capital and back-testing regime.
“An entity shall measure the fair value of the group of financial assets and financial liabilities consistently with how market participants would price the net risk exposure at the measurement date".
If the bank would need to close all positions it will net off positions across all trading desks (IMA and SA both), before going externally. For example, P&L volatility for the bank is calculated based on net positions across all trading desks. As such a calculation based on all positions should be possible, with allocation to individual desks.
The industry respondents also noted that it will be overly burdensome to run valuation adjustment calculation based on the IMA population only. The industry compute valuation adjustments (VA) not for the purpose of the back-testing, but VAs are calculated to account for factors not explicitly addressed in models to ensure fair value compliance for all relevant products across trading desks, which is not specific to IMA population. There is no economic meaning to calculate valuation adjustment only for the IMA population."
As a practical expedient, and to align the calculation of adjustments with business practice, the industry believes the approach of including all positions subject to own funds requirements for market risk is appropriate.
The provision also leads to the possibility of alignment with the actual calculations in the banks. Some banks might calculate adjustments on a portfolio basis and allocate to individual desks, other banks might calculate on a desk basis and take the diversification centrally.
The industry also notes that computing adjustments on the positions in trading desks that are calculating the own funds requirements using the internal model approach only is overly burdensome because it will deviate from the actual calculations which do not make a distinction between internal model and standardised approach. Valuation adjustment is calculated to account for factors not explicitly addressed in models to ensure fair value compliance for all relevant products across trading desks, which is not specific to IMA population. There is no economic meaning to calculate valuation adjustments only for the IMA population.
“2. For the purpose of Article 325bg of Regulation (EU) No 575/2013, an institution may replace the value of a risk factor used in the calculation of the theoretical changes in the trading desk portfolio’s value by the value of the same risk factor used in the calculation of the hypothetical changes in the trading desk portfolio’s value (or derived from input data used in the calculation of the hypothetical changes in the trading desk portfolio’s value), where all of the following conditions are met:
(a) the risk factor used in the calculation of the hypothetical changes in the trading desk portfolio’s value does not directly correspond to an input data;
(b) the risk factor has been derived from input data using any techniques of the valuation systems used for the hypothetical changes in the trading desk portfolio’s value;
(c) not all of the techniques of the valuation systems referred to in (b) have been rebuilt in the valuation systems used in the risk measurement model to derive the value of the risk factor used in the calculation of the theoretical changes in the trading desk portfolio’s value.”
The rephrasing will create more transparency on eligible adjustments – e.g. in case where the RTPL is using more risk factors than a particular trading desk marking (example below) it’s not clear if these can be aligned, whereas the alignment would be allowed if the data from vendors are used (as per the example given on the page 19). We also point out to the Article 16 that stipulates the requirement on the documentation and provision of the rational of any adjustment done within article 15 to ensure appropriate control framework. Example:
Risk Factors (Hypothetical): Swaps(1Y, 2Y) (Bootstrapping method used: A)
Risk Factors (Risk Theoretical): Swaps(1Y, 1.5Y, 2Y) (Bootstrapping method used: B)
For theoretical changes in the trading desk’s portfolio, the value of the 1.5Y Swap rate could be derived from FO marks (Swap rate 1Y & 2Y) and the FO bootstrapping method (A).
Illustration:
For the needs of end-of-day valuation, interest yield curve are usually stripped from the most liquid instruments and corresponding market prices. When instruments with fixed maturity date such as IR futures contracts are involved in the stripping for some segment of the curve, it is common practice to transform these market observations into equivalent price of “synthetic instrument” with fixed to maturity to exclude time decay effects in the calibration of risk factors and thus enable consistent shock distributions in the risk engine.
The right hand table shows a situation where an institution uses an outright USD-LIBOR 3M to perform end-of-day valuation but chooses to model this curve as a spread over the OIS curve, in order to get a more orthogonal representation of risk factors in the risk engine.
This transformation may be done by defining synthetic mono-currency basis swaps and derive related price based on the same pricing functions and input data as in the HPL. It also allows transforming data points corresponding to IR futures into equivalent data points with constant time to maturity. This transformation is not meant to reduce the granularity of the market information captured in the risk engine but transform into a tractable format, fit for the purpose of risk modelling.
Here, the 32 equivalent data points would be used as the input of reference price in the risk engine (ie reference price before shock), notwithstanding of how the institution decide to shock the data points (either with direct calibration or using a proxy).
The industry considers that data alignment should be allowed in such situation as long as institutions can justify to the satisfaction of supervisory authorities that these inherent transformations of market data, performed for the purpose of risk modelling, do not distort the representation/dynamic of the PL function and provide documentation thereof as required in Article 16(2).
Of course, allowing data alignment on these transformations would not mean that the institution would be able to disregard the quality of its shock model. If for instance it is chosen in the risk model to use historical data of the 10Y data point (#24) to calibrate the shocks for the longer segment of the curve (data points #25 to #32), then these proxies should inevitably be captured in the PLA test, that will assess the materiality of the approximation.
Q5. Do you agree with the criteria in paragraph 5 allowing institutions to exclude an adjustment from the changes in the trading desk’s portfolio value? Are there any other criteria you deem useful for this purpose?
The industry survey participants strongly agree to the criteria sets out in paragraph 5 allowing institutions to exclude an adjustment from the changes in the trading desk’s portfolio value and also suggest that the provisional criteria should only apply to market risk relevant adjustments.The industry of the provision in paragraph 5(a)) “that adjustment is computed this way due to its nature” ambiguous or not clear, and seeks more clarity in order to implement in practice. The Industry’s interpretation was that adjustments that aggregate linearly would not meet this criteria (e.g. an IPV adjustment) and adjustments that reflect non-linear effects would meet this criteria (e.g. a delta bid-offer calculation on a portfolio of derivatives).
In reality, a bank may choose to take adjustments at different levels of reporting hierarchy in a way that reflects how it seeks to manage its business. This may result in adjustments being made at a higher level than the FRTB desk. Requiring banks to calculate the adjustment at a lower level only for the purpose of regulatory back-testing would be an additional process to adjust the actual P&L to include a factor not in the desks' reported P&L. This is undesirable, not just due to the additional process but also due to the lack of business use case that would render this allocation liable to error.
Q6.How do institutions identify client margins and day-one profits/losses in the systems (e.g. as commissions, margins)? Please specify if currently they are taken into account in the end-of-day valuation process, in the actual P and L and in the hypothetical P and L.
The industry conducted a survey requesting banks to confirm if the client margins and day-one profits/losses are taken into account in the end-of-day valuation process, in the actual P&L and in the hypothetical P&L with the subsequent results shown in the figures above.The majority of industry respondents confirm that client margins and commissions are taken into account in the end-of-day valuation process and primarily not in the actual P&L. The overwhelming number of respondents confirm Day-one profits/losses are taken into account in both the end-of-day valuation and also in the actual P&L. As a general practice, the industry does not include client margin, commissions or day-one profits/losses in their hypothetical P&L.
The industry acknowledges that client margins corresponds to the P&L generated at the initiation of the transaction. New transactions are part of the EOD Valuation process and Actual P&L but not part of the Hypothetical P&L (CRR Article 366(3)). Currently the margin generated at the initiation of the trade is determined on a declaratory basis. Consequently, day-one profits/losses reserve is computed from the declared client margin.
The industry also recognizes that risk commission is included in the actual P&L for back-testing by some banks. The industry defines the risk commission as a commission on a transaction where a bank agrees to procure an instrument for a client with a price agreed up front. The bank takes on the risk on entering the market (committing capital to fill the order) that the price is different than that agreed with the client.
However, industry members noted that the definition for day-one and client margin P&L may differ across institutions, therefore would urge EBA to retain some level of flexibility in the context of back-testing and P&L attribution. Although current firm-wide definition is subject to ongoing governance and control and often subject to scrutiny by local regulators/supervisors, as part of the Basel II back-testing process. Therefore, industry is in favour of some degree of consistency with the current capital and back-testing regime.
Q7. Paragraph 4 requires institutions to compute (for the purpose of the backtesting) the value of an adjustment (that is included in the changes in the portfolio’s value) performing a stand-alone calculation, i.e. considering only the positions in trading desks that are calculating the own funds requirements using the internal model approach (i.e. desks meeting all conditions in article 325az(2)). Do you agree with the provision? Do you consider the provision clear?
The industry considered the provision to be clear, however the majority of industry survey respondents disagreed with the provision and are concerned that by calculating adjustments only for the trading desks of the Internal Model Approach (‘IMA’) perimeter, without taking into account the netting with Standardised Approach (‘SA’) trading desks, will considerably deviate from the fair value" which is the foundation on which P&L is based according to IFRS 13:48:“An entity shall measure the fair value of the group of financial assets and financial liabilities consistently with how market participants would price the net risk exposure at the measurement date".
If the bank would need to close all positions it will net off positions across all trading desks (IMA and SA both), before going externally. For example, P&L volatility for the bank is calculated based on net positions across all trading desks. As such a calculation based on all positions should be possible, with allocation to individual desks.
The industry respondents also noted that it will be overly burdensome to run valuation adjustment calculation based on the IMA population only. The industry compute valuation adjustments (VA) not for the purpose of the back-testing, but VAs are calculated to account for factors not explicitly addressed in models to ensure fair value compliance for all relevant products across trading desks, which is not specific to IMA population. There is no economic meaning to calculate valuation adjustment only for the IMA population."
Q8. Do you agree with the possibility outlined in paragraph 5 to include in the portfolio’s changes the value of an adjustment stemming from the entire portfolio of positions subject to own funds requirements (i.e. both positions in standard-approach desks and positions in internal model approach desks)? Or do you think it would not be overly burdensome for institutions to compute adjustments on the positions in trading desks that are calculating the own funds requirements using the internal model approach only?
The industry agrees with the provision outlined in paragraph 5 to include in the portfolio’s changes the value of an adjustment stemming from the entire portfolio of positions subject to own funds requirements (i.e. both positions in standard-approach desks and positions in internal model approach desks).As a practical expedient, and to align the calculation of adjustments with business practice, the industry believes the approach of including all positions subject to own funds requirements for market risk is appropriate.
The provision also leads to the possibility of alignment with the actual calculations in the banks. Some banks might calculate adjustments on a portfolio basis and allocate to individual desks, other banks might calculate on a desk basis and take the diversification centrally.
The industry also notes that computing adjustments on the positions in trading desks that are calculating the own funds requirements using the internal model approach only is overly burdensome because it will deviate from the actual calculations which do not make a distinction between internal model and standardised approach. Valuation adjustment is calculated to account for factors not explicitly addressed in models to ensure fair value compliance for all relevant products across trading desks, which is not specific to IMA population. There is no economic meaning to calculate valuation adjustments only for the IMA population.
Q8. Do you agree with the possibility outlined in paragraph 5 to include in the portfolio’s changes the value of an adjustment stemming from the entire portfolio of positions subject to own funds requirements (i.e. both positions in standard-approach desks and positions in internal model approach desks)? Or do you think it would not be overly burdensome for institutions to compute adjustments on the positions in trading desks that are calculating the own funds requirements using the internal model approach only?
This question has been repeated - see above response for Q8.Q9. Do you agree with the criteria outlined in this article for the alignment of input data? Please provide some examples where an institution could use the provision set out in paragraph 2.
Industry welcome EBA’s work to allow alignment of RTPL / HPL due to the differences in the risk factor derivation due to data sources and use of different techniques to derive the risk factors “at source” of this information, as described in particular in the explanatory part of the EBA consultation paper p19. Industry observed however that the text of the article 15 (Section 2) leaves room for interpretation. To clarify, industry suggests to reword the text of the section 2 as follows:“2. For the purpose of Article 325bg of Regulation (EU) No 575/2013, an institution may replace the value of a risk factor used in the calculation of the theoretical changes in the trading desk portfolio’s value by the value of the same risk factor used in the calculation of the hypothetical changes in the trading desk portfolio’s value (or derived from input data used in the calculation of the hypothetical changes in the trading desk portfolio’s value), where all of the following conditions are met:
(a) the risk factor used in the calculation of the hypothetical changes in the trading desk portfolio’s value does not directly correspond to an input data;
(b) the risk factor has been derived from input data using any techniques of the valuation systems used for the hypothetical changes in the trading desk portfolio’s value;
(c) not all of the techniques of the valuation systems referred to in (b) have been rebuilt in the valuation systems used in the risk measurement model to derive the value of the risk factor used in the calculation of the theoretical changes in the trading desk portfolio’s value.”
The rephrasing will create more transparency on eligible adjustments – e.g. in case where the RTPL is using more risk factors than a particular trading desk marking (example below) it’s not clear if these can be aligned, whereas the alignment would be allowed if the data from vendors are used (as per the example given on the page 19). We also point out to the Article 16 that stipulates the requirement on the documentation and provision of the rational of any adjustment done within article 15 to ensure appropriate control framework. Example:
Risk Factors (Hypothetical): Swaps(1Y, 2Y) (Bootstrapping method used: A)
Risk Factors (Risk Theoretical): Swaps(1Y, 1.5Y, 2Y) (Bootstrapping method used: B)
For theoretical changes in the trading desk’s portfolio, the value of the 1.5Y Swap rate could be derived from FO marks (Swap rate 1Y & 2Y) and the FO bootstrapping method (A).
Illustration:
For the needs of end-of-day valuation, interest yield curve are usually stripped from the most liquid instruments and corresponding market prices. When instruments with fixed maturity date such as IR futures contracts are involved in the stripping for some segment of the curve, it is common practice to transform these market observations into equivalent price of “synthetic instrument” with fixed to maturity to exclude time decay effects in the calibration of risk factors and thus enable consistent shock distributions in the risk engine.
The right hand table shows a situation where an institution uses an outright USD-LIBOR 3M to perform end-of-day valuation but chooses to model this curve as a spread over the OIS curve, in order to get a more orthogonal representation of risk factors in the risk engine.
This transformation may be done by defining synthetic mono-currency basis swaps and derive related price based on the same pricing functions and input data as in the HPL. It also allows transforming data points corresponding to IR futures into equivalent data points with constant time to maturity. This transformation is not meant to reduce the granularity of the market information captured in the risk engine but transform into a tractable format, fit for the purpose of risk modelling.
Here, the 32 equivalent data points would be used as the input of reference price in the risk engine (ie reference price before shock), notwithstanding of how the institution decide to shock the data points (either with direct calibration or using a proxy).
The industry considers that data alignment should be allowed in such situation as long as institutions can justify to the satisfaction of supervisory authorities that these inherent transformations of market data, performed for the purpose of risk modelling, do not distort the representation/dynamic of the PL function and provide documentation thereof as required in Article 16(2).
Of course, allowing data alignment on these transformations would not mean that the institution would be able to disregard the quality of its shock model. If for instance it is chosen in the risk model to use historical data of the 10Y data point (#24) to calibrate the shocks for the longer segment of the curve (data points #25 to #32), then these proxies should inevitably be captured in the PLA test, that will assess the materiality of the approximation.