Response to consultation Paper on draft RTS on liquidity horizons for the IMA
Go back
Nevertheless, some point should be considered by the EBA:
Cap on liquidity horizons as defined in Article 325bd(4) of Regulation (EU) 2019/876 (i.e. the definition of the Effective LH in art. 325bd(4)):
o We regret that this point is not part of the EBA mandate and as such of the consultation and we think that it is important to stress this concern in the context of the consultation.
o We believe capping the LH for a risk factor at the maturity of the related instrument should be an option, and not a requirement. Indeed, from a theoretical standpoint, this maturity cap traduces some sort of constant position assumption whereby no position shall be renewed beyond its expiry. It seems inconsistent with the general constant risk assumption prevailing in the ES formula itself, where the intensity of risk factors (from 0d to 10d, from 10d to 20d, from 20d to 40d, etc.) is deemed constant throughout the life of the positions in spite of the portfolio ageing. Added to that, it contradicts the allowance to increase the LH of a broad risk factor category at the desk level to avoid breaking of hedges. Finally, this capping requirement will create added operational/computational burdens to implement as well
o Recommendations: Banks should be allowed to assign all instruments to the regulatory LH; capping at instrument maturity should be an option.
This treatment could generate unintended cliff effects and be overly conservative for banks that opt for a non-look-through approach for index in Expected Shortfall (ES). Indeed, small changes in a component’s rating can result in significant overall change for the entire index: Where even one constituent of an investment grade (IG) index being downgraded will result in the weighted average LH to become slightly greater than 20 days, and at that point, the LH of the entire index will have to be switched to the next highest LH, i.e. 40 days, although the index is economically more liquid than its components.
Additionally, this approach could impose further operational and computational burdens:
o if the liquidity horizons of risk factors modeling homogeneous index must be updated on a too regular basis,
o and, for risk factors that differ from spot/credit spread risk of homogeneous index (eg: volatility, dividend or repo) with components’ weights that may not be directly observable (eg: for volatility index, the weights depend of the implied correlations between the volatilities at component level).
Recommendations: We recommend that, for homogeneous indices:
In order to keep as much comparability as possible between banks that opt for a look-through approach for index in ES and banks that opt for a non-look-through approach:
o When the composition of the index is standardised, the liquidity horizon should be determined consistently with the nature of the entire index. For example, a corporate investment grade (IG) index would be mapped to a liquidity horizon of 40 days and a corporate high yield (HY) index would be mapped to a liquidity horizon of 60 days;
Otherwise, the weighted average liquidity horizon may be used and assigned to the closest corresponding liquidity horizon (out of 10, 20, 40, 60 and 120 days). For example if the weighted average liquidity horizon is 26 days, the liquidity horizon of the index would be 20 days and if the weighted average liquidity horizon is 33 days, it would be 40 days. This approach would accurately reflect that indices are economically more liquid than their components.
For the sake of simplicity:
o the update frequency of liquidity horizons should be aligned with the frequency of the other processes that impact directly the scope of the risk factors in the internal model (RFET and eligibility tests) i.e. quarterly basis;
o the weights used to determine the liquidity horizons of spot/credit spread risk may also be used to determine the liquidity horizons of other subcategories of broad risk factors related to homogeneous index as volatility, dividend and repo.
Added to that, to ensure a consistent treatment of index and multi-underlying instruments, the above proposed approach should also be allowed for multi-underlying instruments where all underlyings have delta risk sensitivities of the same sign.
Finally, this proposal can be slightly adapted to determine the liquidity horizon of non-homogeneous index (see answer to Question 4 below).
the broad risk factor category based on the general mapping methodology described in Article 1;
the liquidity horizon of the risk factor based on the approach proposed in the industry response to question 3 of this consultative paper.
One area where there has not been sufficient focus is the assignment of correlation parameters into liquidity horizons. Since the correlation is derived from variance and covariance, we propose that correlation (A, B) should be allocated to the longest liquidity horizon prescribed for volatility of A and B;
For example, a correlation between two large cap would be assigned to EQ-20 (instead of EQ-60), or correlation between XAU/USD and EUR/USD would be assigned to CO-60 (instead of CO-120).
We appreciate the EBA proposal to consider constituents of main indices as large market capitalisation. This would align the liquidity horizon of a non-decomposed index with that of its constituents in a look-through approach and, consequently, IMA comparable capital charges.
Besides, we proposed for the mapping of homogenous index instruments (cf. response to Question 3) that the index liquidity horizon be set to the closest supervisory LH of the constituents average liquidity horizon rather than the next higher supervisory LH. However, if the EBA ultimately stick to its approach of homogenous index mapping, Option B would have the additional benefit that a homogenous index mapped liquidity horizon would not be “polluted” by a small fraction of its constituents being map to a higher supervisory LH than the majority of its constituents.
Finally, given the merit, highlighted above, of Option B, we do not understand its restriction to European components only indices. The same considerations should lead to the same conclusions and hence all Commission Implementing Regulation (EU) 2016/1646 listed indices should be eligible for defining large market capitalisation.
Q1. Do you agree with the general methodology? If not, please explain why.
The French Banking Federation broadly agrees with the general mapping methodology developed in Article 1 of the draft Regulatory Technical Standard on liquidity horizons.Nevertheless, some point should be considered by the EBA:
Cap on liquidity horizons as defined in Article 325bd(4) of Regulation (EU) 2019/876 (i.e. the definition of the Effective LH in art. 325bd(4)):
o We regret that this point is not part of the EBA mandate and as such of the consultation and we think that it is important to stress this concern in the context of the consultation.
o We believe capping the LH for a risk factor at the maturity of the related instrument should be an option, and not a requirement. Indeed, from a theoretical standpoint, this maturity cap traduces some sort of constant position assumption whereby no position shall be renewed beyond its expiry. It seems inconsistent with the general constant risk assumption prevailing in the ES formula itself, where the intensity of risk factors (from 0d to 10d, from 10d to 20d, from 20d to 40d, etc.) is deemed constant throughout the life of the positions in spite of the portfolio ageing. Added to that, it contradicts the allowance to increase the LH of a broad risk factor category at the desk level to avoid breaking of hedges. Finally, this capping requirement will create added operational/computational burdens to implement as well
o Recommendations: Banks should be allowed to assign all instruments to the regulatory LH; capping at instrument maturity should be an option.
Q2. Besides systemic risk factors (i.e. risk factors capturing the market/systemic component of the modelled risk), are there other risk factors/parameters that would reflect risks embedded in more than one subcategories or more than one categories?
No answer.Q3. Do you agree with the treatment reserved for homogenous indices?
While we appreciate EBA’s effort, we do not agree with the treatment reserved for homogenous indices in Article 2 of the draft Regulatory Technical Standard on liquidity horizons: This treatment could generate unintended cliff effects and be overly conservative for banks that opt for a non-look-through approach for index in Expected Shortfall (ES). Indeed, small changes in a component’s rating can result in significant overall change for the entire index: Where even one constituent of an investment grade (IG) index being downgraded will result in the weighted average LH to become slightly greater than 20 days, and at that point, the LH of the entire index will have to be switched to the next highest LH, i.e. 40 days, although the index is economically more liquid than its components.
Additionally, this approach could impose further operational and computational burdens:
o if the liquidity horizons of risk factors modeling homogeneous index must be updated on a too regular basis,
o and, for risk factors that differ from spot/credit spread risk of homogeneous index (eg: volatility, dividend or repo) with components’ weights that may not be directly observable (eg: for volatility index, the weights depend of the implied correlations between the volatilities at component level).
Recommendations: We recommend that, for homogeneous indices:
In order to keep as much comparability as possible between banks that opt for a look-through approach for index in ES and banks that opt for a non-look-through approach:
o When the composition of the index is standardised, the liquidity horizon should be determined consistently with the nature of the entire index. For example, a corporate investment grade (IG) index would be mapped to a liquidity horizon of 40 days and a corporate high yield (HY) index would be mapped to a liquidity horizon of 60 days;
Otherwise, the weighted average liquidity horizon may be used and assigned to the closest corresponding liquidity horizon (out of 10, 20, 40, 60 and 120 days). For example if the weighted average liquidity horizon is 26 days, the liquidity horizon of the index would be 20 days and if the weighted average liquidity horizon is 33 days, it would be 40 days. This approach would accurately reflect that indices are economically more liquid than their components.
For the sake of simplicity:
o the update frequency of liquidity horizons should be aligned with the frequency of the other processes that impact directly the scope of the risk factors in the internal model (RFET and eligibility tests) i.e. quarterly basis;
o the weights used to determine the liquidity horizons of spot/credit spread risk may also be used to determine the liquidity horizons of other subcategories of broad risk factors related to homogeneous index as volatility, dividend and repo.
Added to that, to ensure a consistent treatment of index and multi-underlying instruments, the above proposed approach should also be allowed for multi-underlying instruments where all underlyings have delta risk sensitivities of the same sign.
Finally, this proposal can be slightly adapted to determine the liquidity horizon of non-homogeneous index (see answer to Question 4 below).
Q4. Do you have any example of other risk factors that should be subject to the treatment specified for indices?
In our view, other risk factors should be subject to the treatment specified for indices, such as indices or baskets on multi-asset classes. In such cases, we recommend determining: the broad risk factor category based on the general mapping methodology described in Article 1;
the liquidity horizon of the risk factor based on the approach proposed in the industry response to question 3 of this consultative paper.
Q5. Are there any other risk factors for which an ad-hoc treatment should be specified?
The ad-hoc treatment should be specified for other risk factors such as correlation parameters: One area where there has not been sufficient focus is the assignment of correlation parameters into liquidity horizons. Since the correlation is derived from variance and covariance, we propose that correlation (A, B) should be allocated to the longest liquidity horizon prescribed for volatility of A and B;
For example, a correlation between two large cap would be assigned to EQ-20 (instead of EQ-60), or correlation between XAU/USD and EUR/USD would be assigned to CO-60 (instead of CO-120).
Q6. What is your preferred option? Please explain why.
French banks are supporting consistent definition across regulations. For that purpose, we see fit to consider market capitalisation of 1 B€ as a relevant threshold for the determination of large market capitalisation, i.e. the same threshold used for the determination of main indices [i.e. Commission Implementing Regulation (EU) 2016/1646, Recital 1].We appreciate the EBA proposal to consider constituents of main indices as large market capitalisation. This would align the liquidity horizon of a non-decomposed index with that of its constituents in a look-through approach and, consequently, IMA comparable capital charges.
Besides, we proposed for the mapping of homogenous index instruments (cf. response to Question 3) that the index liquidity horizon be set to the closest supervisory LH of the constituents average liquidity horizon rather than the next higher supervisory LH. However, if the EBA ultimately stick to its approach of homogenous index mapping, Option B would have the additional benefit that a homogenous index mapped liquidity horizon would not be “polluted” by a small fraction of its constituents being map to a higher supervisory LH than the majority of its constituents.
Finally, given the merit, highlighted above, of Option B, we do not understand its restriction to European components only indices. The same considerations should lead to the same conclusions and hence all Commission Implementing Regulation (EU) 2016/1646 listed indices should be eligible for defining large market capitalisation.