EBA Guidelines on Stressed Value At Risk (Stressed VaR) EBA/GL ...

<strong>EBA</strong> <strong>Guidelines</strong>onStressed Value At Risk (Stressed VaR)<strong>EBA</strong>/GL/2012/2London, 16.05.2012

ContentsI. Executive Summary .......................................................................................................... 3II. Background and Rationale ........................................................................................... 4III. <strong>EBA</strong> <strong>Guidelines</strong> on Stressed VaR ............................................................................. 6Title I -Subject matter, Scope and Definitions ..................................................................... 8Title II – Requirements regarding institutions’ Stressed VaR modelling.................... 9Title III- Final Provisions and Implementation ................................................................... 17IV. Accompanying documents ..................................................................................... 18a. Feedback on the public consultation and on the opinion of the BSG ................ 18V. Confirmation of compliance with guidelines and recommendations ........ 312

I. Executive SummaryThe amendments to the Capital Requirements Directive 1 by Directive 2010/76/EU(CRD III) 2 relate, among others, to Stressed Value-at-Risk (Stressed VaR) in thetrading book. According to these amendments, the predecessor of the <strong>EBA</strong>, theCommittee of European Banking Supervisors (CEBS) 3 is tasked with monitoring therange of practices in this area and drawing up guidelines in order to ensureconvergence of supervisory practices.The amendments to the Capital Requirements Directive by Directive 2010/76/EU(CRD III) entered into force on 31 December 2011.Providing guidance on Stressed VaR modelling by credit institutions using theInternal Model Approach (‘IMA’) for the calculation of the required capital for marketrisk in the trading book, is seen as an important means of addressing weaknesses inthe regulatory capital framework and in the risk management of financial institutionsthat contributed to the turmoil in global financial markets. It is also expected toreduce reliance on cyclical VaR-based capital estimates as well as to contribute tothe development of a more robust financial system.The first chapter, on ‘Identification and validation of the stressed period’, elaborateson the value-at-risk model inputs calibrated to historical data from a continuous 12-month period of significant financial stress relevant to an institution’s portfolio anddeals with i) the length of the stressed period, ii) the number of stressed periods touse for calibration, iii) the approach to identify the appropriate historical period andiv) the required documentation to support the approach used to identify the stressedperiod. The second chapter, on ‘Review of the stressed period’ provides guidance onthe frequency and monitoring of a stressed period. The third chapter on ‘StressedVaR methodology’ deals with i) consistency issues between the VaR and StressedVaR methodologies and ii) the use and validation of proxies in Stressed VaRmodelling. The fourth and final chapter, entitled ‘Use tests’ specifies use testrequirements.The <strong>Guidelines</strong> on Stressed VaR are expected to contribute to a level playing fieldamong institutions and to enhance convergence of supervisory practices among thecompetent authorities across the EU. It is expected that the national competentauthorities around the EU will implement the <strong>Guidelines</strong> by incorporating them withintheir supervisory procedures within six months after publication of the finalguidelines. After that date, the competent authorities must ensure that institutionscomply with the <strong>Guidelines</strong> effectively.1 Capital Requirements Directive (CRD) is a colloquial reference to Directives 2006/48/EC and Directive 2006/49/EC.In this document, references to ‘Directive 2006/48/EC’ and ‘Directive 2006/49/EC’ or the ‘CRD’ are to the amendedversions of the Directives; references to a particular Article of the CRD refer to the Directives as amended and inforce.2Directive 2010/76/EU was published on 24 November 2010 and can be found under: http://eurlex.europa.eu/LexUriServ/LexUriServ.do?uri=OJ:L:2010:329:0003:0035:EN:PDF.3 The European Banking Authority (<strong>EBA</strong>) was established by Regulation (EC) No. 1093/2010 of the EuropeanParliament and of the Council of 24 November 2010. The <strong>EBA</strong> came into being on 1 January 2011. It has taken overall existing and ongoing tasks and responsibilities from the Committee of European Banking Supervisors (CEBS).3

II. Background and RationaleThe CRD III trading book amendments, including the requirement of Stressed Valueat Risk (VaR) modelling for the calculation of the regulatory capital for market risk inthe trading book, are the result of widespread international (G20, Basel, FSF)recognition in 2008 that further regulatory reform was needed to addressweaknesses in the current regulatory capital framework and in the risk managementof financial institutions that contributed to the turmoil in global financial markets.In January 2009, the Basel Committee on Banking Supervision (BCBS) proposedsupplementing the current VaR-based trading book framework with, among othermeasures, an incremental risk capital charge (IRC), which includes default risk aswell as migration risk for unsecuritised credit products and a stressed value-at-riskrequirement 4 .As observed losses in banks' trading books during the financial crisis have beensignificantly higher than the minimum capital requirements under the Pillar 1 marketrisk rules, the BCBS proposed to enhance the framework through requiring banks tocalculate, in addition to the current VaR, a stressed VaR taking into account a oneyearobservation period relating to significant losses. The additional stressed VaRrequirement is expected to help reduce the pro-cyclicality of the minimum capitalrequirements for market risk.In the process of refining capital requirements for market risk, the BCBS conducteda quantitative impact study 5 . In the summer of 2009, the Trading Book Group (TBG)investigated the impact of the provisions of the ‘Revisions to the Basel II market riskframework’ and of the ‘<strong>Guidelines</strong> for computing capital for incremental risk in thetrading book’ consultation papers published in January 2009, focusing (generally) onbig internationally-active banks with extensive trading activities.The amendments to the Capital Requirements Directive by Directive 2010/76/EU(CRD III) relating to Stressed VaR in the trading book are a direct transposition ofthe proposals from the BCBS in the EU context.The European Banking Authority is requested to monitor the range of practices inthis area and to provide guidelines on Stressed VaR models.The objectives of these <strong>Guidelines</strong> on Stressed VaR are:I. To achieve a common understanding among the competent authorities across theEU on Stressed VaR modelling in order to enhance convergence of supervisorypractices;II.III.To create more transparency for institutions when implementing Stressed VaRinto the calculation of the required capital for market risk in the trading book andinto their risk management practices; andTo create a level playing field among institutions in this area.4 Revisions to the Basel 2 market risk framework - final version (July 2009), <strong>Guidelines</strong> for computing capital forincremental risk in the trading book - final version (July 2009), Enhancements to the Basel II framework (July2009).5 Analysis of the trading book quantitative impact study (October 2009).4

The guidelines presented in this paper do not aim to be a comprehensive set ofrules, but rather to complement the new CRD provisions relating to Stressed VaRwhere additional guidance by the <strong>EBA</strong> was deemed necessary or appropriate.Given that the <strong>Guidelines</strong> discussed in this paper do not go beyond the provisions ofthe CRD but rather clarify how the rules are to be applied in practice a detailedassessment of the costs and benefits associated with them is not required. Thesecosts and benefits are unlikely to be incremental to those identified in the EUCommission’s Impact Assessment accompanying its CRDIII proposal.5

III. <strong>EBA</strong> <strong>Guidelines</strong> on Stressed VaRStatus of these <strong>Guidelines</strong>1. This document contains guidelines issued pursuant to Article 16 of Regulation(EU) No 1093/2010 of the European Parliament and of the Council of 24November 2010 establishing a European Supervisory Authority (EuropeanBanking Authority), amending Decision No 716/2009/EC and repealingCommission Decision 2009/78/EC (‘the <strong>EBA</strong> Regulation’). In accordance withArticle 16(3) of the <strong>EBA</strong> Regulation, competent authorities and financial marketparticipants must make every effort to comply with the guidelines.2. <strong>Guidelines</strong> set out the <strong>EBA</strong>’s view of appropriate supervisory practices withinthe European System of Financial Supervision or of how Union law should beapplied in a particular area. The <strong>EBA</strong> therefore expects all competentauthorities and financial market participants to whom guidelines are addressedto comply. Competent authorities to whom guidelines apply should comply byincorporating them into their supervisory practices as appropriate (e.g. byamending their legal framework or their supervisory rules and/or guidance orsupervisory processes), including where particular guidelines are directedprimarily at institutions.Notification Requirements3. According to Article 16(3) of the <strong>EBA</strong> Regulation, competent authorities mustnotify the <strong>EBA</strong> as to whether they comply or intend to comply with theseguidelines, or otherwise with reasons for non-compliance, by 16.07.2012. Inthe absence of any notification by this deadline, competent authorities will beconsidered by the <strong>EBA</strong> to be non-compliant. Notifications should be sent bysubmitting the form provided at Section V to compliance@eba.europa.eu withthe reference ‘<strong>EBA</strong>/GL/2012/2’. Notifications should be submitted by personswith appropriate authority to report compliance on behalf of their competentauthorities.4. The notification of competent authorities mentioned in the previousparagraph shall be published on the <strong>EBA</strong> website, as per article 16 of <strong>EBA</strong>Regulation.6

ContentsIII. <strong>EBA</strong> <strong>Guidelines</strong> on Stressed VaR ............................................................................. 6Title I – Subject matter, Scope and Definitions ................................................................... 81. Subject matter ............................................................................................ 82. Scope and level of application ....................................................................... 83. Definitions .................................................................................................. 8Title II – Requirements regarding institutions’ Stressed VaR modelling.................... 9A. Identification and validation of the stressed period .................................... 94. Length of the stressed period ........................................................................ 95. Number of stressed periods to use for calibration ............................................ 96. The approach for identifying the appropriate historical period ......................... 107. Documentation to support the approach used to identify the stressed period .... 11B. Review of the stressed period ................................................................... 128. Frequency ................................................................................................ 129. Monitoring the stressed period .................................................................... 12C. Stressed VaR methodology ........................................................................ 1310. Consistency with VaR methodology ............................................................ 1311. Estimation of proxies for Stressed VaR ....................................................... 1512. Validation of proxies ................................................................................ 1613. Validation of model inputs/outputs ............................................................. 16D. Use test ..................................................................................................... 1714. Use test.................................................................................................. 17Title III – Final Provisions and Implementation ................................................................. 1715. Date of application ................................................................................... 177

Title I – Subject matter, Scope and Definitions1. Subject matterThese guidelines aim at achieving a common understanding among the competentauthorities across the EU on Stressed Value at Risk (VaR) models in order toenhance convergence of supervisory practices in line with Annex V of Directive2006/49/EC, as amended by Directive 2010/76/EU.2. Scope and level of application1. Competent authorities should require institutions to comply with the provisionslaid down in these <strong>Guidelines</strong> on Stressed VaR.2. These guidelines should apply to institutions using an Internal Model Approach(IMA) for the purpose of calculating the capital requirement for market risk in thetrading book.3. The guidelines apply to institutions at the level (solo and/or consolidated) onwhich the model is authorised to be used by the relevant competent authority,unless stated otherwise in these <strong>Guidelines</strong>.3. DefinitionsIn these guidelines the following definitions should apply:a. The term institutions should mean credit institutions and investment firms asset out in Directives 2006/48/EC and 2006/49/EC.b. The term antithetic data under point 6 of these <strong>Guidelines</strong> should mean pricemovements which are considered relevant irrespective of their direction.c. The term de-meaning under point 10 of these <strong>Guidelines</strong> should mean aquantitative process to remove a trend from historical data. Depending on thepositions and the size of the trend, not removing the drift from the historicaldata to simulate the price variations could generate mainly profitablescenarios and very few and limited losses.d. The term proxy under point 11 of these <strong>Guidelines</strong> should mean an observablevariable or price taken from a liquid market that is used to substitute avariable that cannot be observed (or whose hypothetical price does not reflectreal transactions from a deep two-way market) and thus cannot be accuratelymeasured. Institutions use proxies both for valuation and risk measurementpurposes. From a theoretical perspective three types of proxies can beidentified: those applied in the valuation of instruments (which would affectthe adequacy of VaR and Stressed VaR as capital measures); those used for8

VaR calculations (which would also be present in Stressed VaR metrics); andthose affecting solely the Stressed VaR calculation.Title II – Requirements regarding institutions’ Stressed VaRmodellingA. Identification and validation of the stressed period4. Length of the stressed period1. The requirement set out in the CRD that the historical data used to calibrate theStressed VaR measure have to cover a continuous 12-month period, applies alsowhere institutions identify a period which is shorter than 12 months but which isconsidered to be a significant stress event relevant to an institution’s portfolio.2. The approach to be applied to identify the stressed period in order to meet therequirement of Paragraph 10a of Annex 5 of Directive 2006/49/EC as amended byDirective 2010/76/EU, to calculate a Stressed VaR measure calibrated to acontinuous 12-month period of financial stress relevant to an institution’s portfolio, isthe most material element determining the output of the model and is thereforesubject to approval by the competent authorities.5. Number of stressed periods to use for calibration1. For the purposes of approval of the choice of the stressed period by institutions, acompetent authority is the competent authority responsible for the exercise ofsupervision on a consolidated basis of this EU institution and, in the case of aninternal model also recognised at a subsidiary’s level, the competent authorityresponsible for the exercise of supervision of this EU institution’s subsidiary.2. When the competent authorities approve the stressed period defined at grouplevel according to Article 37(2) of Directive 2006/49/EC referring to Article 129 ofDirective 2006/48/EC, a single stressed period should only be required to be definedat group level.3. As an exception from the above, the competent authorities should require an EUinstitution to determine a different stressed period at a subsidiary’s level if thestressed period defined for the group is not considered relevant to the subsidiary’sportfolio. Where a single group-wide stressed period is used in an institution that hasa subsidiary with a locally approved VaR model, institutions should provide proofthat this group-wide stressed period is relevant to the subsidiary’s portfolio.9

6. The approach for identifying the appropriate historical period1. In order to choose a historical period for calibration purposes, institutions shouldformulate a methodology for identifying a stressed period relevant to their currentportfolios, based on one of the following two ways:i. judgement-based approaches; orii.formulaic approaches.2. A judgement-based approach is one that does not use a detailed quantitativeanalysis to identify the precise period to use for calibration, but rather relies on ahigh-level analysis of the risks inherent in an institution’s current portfolio and pastperiods of stress related to those risk factors. Where this judgement-based approachis used by institutions, it should include quantitative elements of analysis.3. A formulaic approach instead is one that applies, in addition to expert judgement,a more systematic quantitative analysis to identify the historical period representinga significant stress for an institution’s current portfolio. This more systematicapproach could be employed in a number of ways:i. A risk-factor based approach: an institution identifies a restricted numberof risk factors which are considered to be a relevant proxy for themovement in value of its portfolio. The historical data for these risk factorscan then be fully analysed to identify the most stressed period (forexample, through identification of the period of highest volatility of the riskfactors), in the historical data window.ii.A VaR based approach: the historical period is identified by running eitherthe full VaR model or an approximation over a historical period to identifythe 12-month period which produces the highest resulting measure for thecurrent portfolio.4. This approach should be employed to determine a historical period that wouldprovide a conservative capital outcome rather than just selecting the period ofhighest volatility.5. While either approach may be used by institutions, the use of the formulaicapproach, where possible, should be preferred for the identification of the historicalperiod.6. Institutions may also combine the above two approaches to limit thecomputational burden of the formulaic approach. This can be done by using thejudgement-based approach to restrict the historical data periods to be considered inthe formulaic approach.7. Irrespective of the approach used, institutions should provide evidence that thestressed period is relevant for their current portfolio and that they have considered a10

ange of potential historical periods in their analyses. The institutions should alsohave to prove that the portfolio on which the identification of the stressed periods isbased is representative of the institutions’ current portfolio, e.g. by applying theapproach to identify the stressed period to other typical or previous portfolios. As anexample, for many portfolios, a 12-month period relating to significant losses in2007/2008 would adequately reflect a period of such stress, but, in addition to that,other periods relevant to the current portfolio should also be considered byinstitutions.8. In all cases no weighting of historical data should be applied when determiningthe relevant historical period or when calibrating the Stressed VaR model, as theweighting of data in a stressed period would not result in a true reflection of thepotential stressed losses that could occur for an institution’s portfolio.9. Finally, competent authorities may require institutions to use antithetic data whencalibrating the Stressed VaR model, especially where an institution’s portfolio ischaracterized by frequent position changes.7. Documentation to support the approach used to identify the stressedperiod1. Irrespective of the approach applied, institutions must produce robustdocumentation justifying the choice of approach made. This should in all casesinclude quantitative assessments to support the current choice of the historicalperiod and its relevance for the current portfolio. This should also includedocumentation of the modelling of risk factors’ returns.2. Where institutions apply a formulaic approach to identify the stressed period thefollowing issues should, as a minimum, be addressed in the related documentation:i. Justification for the choice of risk factors used if a risk-factor basedapproach is applied and where fewer than the modelled risk factors areselected.ii.Justification of any simplifications where a simplified VaR model is used toidentify the historical period.3. Where a formulaic approach is applied, which is based on a simplified VaR model,an institution should also provide adequate evidence that the simplified measuregives directionally the same VaR results as the full VaR model (and therefore isaccurate in determining the most stressed period). This evidence should includeempirical analysis.4. Where a formulaic approach is applied, which aims at identifying the most volatileperiod for a set of risk factors, an institution should provide adequate evidence thata period of high volatility is a suitable proxy for a period in which the VaR measure11

would be high and that the lack of inclusion of correlations or other factors thatwould be reflected in the VaR measure does not result in rendering this proxyunsuitable.B. Review of the stressed period8. Frequency1. The requirement of the CRD, for the review of the identified 12-month period ofsignificant stress to be performed at least yearly by institutions, means that differentcircumstances, including a very high turnover in the trading book or specific tradingstrategies, may require a review of the stressed period with a higher frequency.2. Any changes to the choice of the historical period following the outcome of thereview of the stressed period should be communicated to the competent authoritybefore the intended implementation date of the proposed changes.9. Monitoring the stressed period1. In addition to the above-mentioned regular review, an institution should have inplace procedures which ensure, on an on-going basis, that the specified stressedperiod remains representative, including when ,market conditions or portfoliocompositions have been subject to significant change.2. In order to put in place sound procedures for the ongoing monitoring of therelevance of a stressed period, an institution should document the soundness of theimplemented approach. Monitoring may be based on a variety of factors which maydiffer among institutions. Factors to be considered include changes in marketconditions, in trading strategies or also in portfolio composition. These factors maybe analysed by comparing them to changes in the allocation of market values ornotionals, in risk factor loadings, in the level of VaR or sensitivities, in the repartitionof VaR or sensitivities over portfolios and risk categories, in the P&L and back-testingresults or also by the impact of newly approved products on the risk profile.3. In addition to the above-mentioned procedures, monitoring of Stressed VaRrelative to VaR should be performed on an on-going basis, because, while in theory,due to differences in parameterisation, Stressed VaR can exceptionally be smallerthan VaR, also at inception, this should not structurally be the case. The ratiobetween Stressed VaR and VaR at the moment of identification of the relevantstressed period should be used as a reference value for ongoing monitoring.Significant decreases in the ratio should be considered as indications for a potentialneed for review of a stressed period. A ratio between Stressed VaR and VaR belowone should be considered as a warning signal triggering a review of the stressedperiod.12

C. Stressed VaR methodology10. Consistency with VaR methodology1. The Stressed VaR methodology should be based on the current VaR methodology,with specific techniques required, where applicable, in order to adjust the currentVaR model into one that delivers a Stressed VaR measure. Any risk factor occurringin the VaR model should therefore be reflected in the Stressed VaR model.2. With respect to standards used in both measures, and further to the onesprescribed by the Directive (e.g. the 99% confidence level), institutions mayconsider the use of ‘square root of time’ scaling to calculate a 10-day Stressed VaRmeasure. Nevertheless, given some known limitations of the scaling factor, ananalysis to demonstrate that the assumptions underlying the use of the ‘square rootof time’ rule are appropriate, should form part of the internal model validationprocess.3. While the Stressed VaR model should share some of the current VaR standards,others may diverge due to explicit Directive requirements or to methodologicalincompatibilities related to the Stressed VaR concept. In particular, this is the case inthe following areas:(i) Length of the stressed PeriodGiven the length of the stressed period must be 12 months, any action toreduce or increase the stated stressed period based on the need forconsistency between VaR and Stressed VaR should not be permitted.(ii) Back-testing requirementThe multiplication factor m s used for capital requirements should be at least 3and be increased by an addend between 0 and 1 depending on the VaRbacktesting results. Nevertheless, backtesting is not a requirement in itself fordetermining the Stressed VaR measure.(iii)Periodicity of the Stressed VaR calculationAs the CRD provides that the calculation of the Stressed VaR should be atleast weekly, institutions may choose to compute the measure morefrequently, for instance, daily, to coincide with the VaR periodicity.If, for example, institutions decide on a weekly Stressed VaR computation,and assuming a one-day Stressed VaR scaled up to 10 days, for the dailycalculation of capital requirements based on internal models the followingwould apply:13

a) The same Stressed VaR number would be used for 5 subsequentbusiness days following the running of the Stressed VaR model;b) With respect to the calculation of the average Stressed VaR numbersduring the preceding sixty business days, institutions should use theprevious 12 Stressed VaR numbers to compute that average;c) An institution should be able to prove that, on the day of the weekchosen for Stressed VaR calculation, its portfolio is representative of theportfolio held during the week and that the chosen portfolio does notlead to a systematical underestimation of the Stressed VaR numberswhen computed weekly. For example, proof that the VaR is notsystematically lower on the day of the week chosen for Stressed VaRcalculation could be considered sufficient.4. Stressed VaR standards may diverge from VaR standards in other circumstanceswhere there could be methodological incompatibilities between the current VaR andthe Stressed VaR model. One example includes changes in the current VaRmethodology that cannot be translated into the Stressed VaR measure and the useof local valuation (sensitivity analysis/proxies) as opposed to full revaluation, whichis the preferred approach for Stressed VaR.5. As a general rule, changes in an institution’s VaR model or VaR methodologyshould be reflected in changes to the model/methodology used to calculate theStressed VaR charge.6. Under exceptional circumstances, if an institution can demonstrate that it cannotincorporate enhancements to the current VaR methodology in the Stressed VaR,such situations should be documented and the institution should be able todemonstrate that the impact (for example, in terms of VaR or capital requirements)resulting from the current VaR developments which are not implemented in theStressed VaR measure is limited.7. Where sensitivities rather than full revaluation are used within a VaR model, theinstitution concerned should demonstrate that this approach is still appropriate forStressed VaR where larger shocks are applied. A sensitivity-based approach forStressed VaR may require that higher order derivatives/convexity are factored in.8. Any revaluation ladders or spot/volatility matrices employed should be reviewedand extended to include the wider shocks in risk factors that occur in stressfulscenarios. It is preferable that full revaluation be used for Stressed VaR with shocksapplied simultaneously to all risk factors.14

9. In terms of calibration to market data, the process of ‘de-meaning’ is notconsidered necessary for Stressed VaR. If there is a significant drift in market data,the use of antithetic data is preferable to ‘de-meaning’.10. The table below summarises the main issues described above concerning thelevel of consistency between the methodological aspects of the current VaR andStressed VaR measure.Is consistency between VaR and Stressed VaRrequired?Yes for… No for… Subject to verificationConfidence WeightinglevelschemeChanges to modelsHolding period Back-testingUse of Taylor seriesapproximationsLength ofhistoricalobservationperiodFrequency ofcomputationScaling method11. Estimation of proxies for Stressed VaR1. Given that the data constraints that make necessary the use of proxies for VaR,become even more relevant for Stressed VaR and that it is expected that any proxiesused in VaR will also be necessary for Stressed VaR, while additional ones may alsobe needed, whereas any new risk factor not present in the historical data shouldnaturally require the use of a proxy for VaR calculation, but only on a ‘temporary’basis (e.g. after one year there would be enough real information to complete a 12-month data series),the same proxy should be more ‘permanent’ for Stressed VaRpurposes (due to the more constant nature of the historical time series).2. If a risk factor is missing in the stressed period because it was not observableduring that period (for example for a newly listed equity) the institution may useanother risk factor (in this example, another equity from the same sector and with asimilar risk and business profile) for which there is information available and forwhich a highly correlated behaviour with the factor that the institution is trying tocapture can be demonstrated. Where these proxies are used, institutions shouldconsider whether an assumption of 100% correlation between the risk factor and itsproxy is appropriate.15

3. Institutions may alternatively map the missing factor to another one similar interms of volatility (though not necessarily correlated). If this approach is used,institutions should demonstrate that it is conservative and appropriate.4. If a VaR model is enhanced by incorporating a risk factor, an institution shouldalso incorporate it into its Stressed VaR calculations. In certain cases, this may meanreviewing the historical data series for the risk factors and introducing anappropriate proxy. For example where a new risk factor used for valuation purposesis incorporated into the VaR model as required under Annex V point 12 firstParagraph of Directive 2006/49/EC as amended by Directive 2010/76/EU.5. In all cases, the use of these proxies, including simplifications and any omissionsmade, will only be acceptable provided they are well documented and theirlimitations are taken into account and addressed in the institution’s capitalassessment.12. Validation of proxies1. Whereas validation of a proxy should be broadly performed in the same way forVaR and Stressed VaR, any proxy validated for the day-to-day VaR is notautomatically acceptable for Stressed VaR. Proxies in use should be reviewedperiodically to assess their adequacy and ensure that they provide a conservativeoutcome.2. Regarding those proxies which might be used for Stressed VaR purposes only (forinstance, due to lack of data in the selected period), an institution should ensurethat the risk factor used as proxy is conservative.13. Validation of model inputs/outputs1. All qualitative standards defined for the control of consistency, accuracy andreliability of data sources of VaR also apply to Stressed VaR.2. Underlyings for which institutions do not have a history of data complete enoughto cover the reference period, should be shocked by approximation, using closelyrelated underlyings (same market, similar structure and characteristics). Followingthe same process that has been approved for the institutions’ internal models, inorder to ensure the quality of historical data used for the reference period,institutions should document the methodology followed for identifying and forproxying missing data. Institutions should also perform tests of the potential impactof the use of these proxies.3. With a view to preserving arbitrage inequalities, institutions may need to applydata cleaning for Stressed VaR. Where this is the case, the removal of outliers from16

historical data series should be appropriately justified and documented, as it shouldnot end up decreasing the magnitude of extreme events.4. As Stressed VaR entails, by definition, the application of highly stressed scenariosto current market parameters, which may lead to incoherent market conditions (e.g.negative forward rates) more frequently than within a VaR computation, institutionsshould monitor the calibration failures that may materialise. Institutions using fullrevaluation when estimating their Stressed VaR may be more frequently confrontedwith those calibration failures than institutions not using full revaluation, not becausefailures will not happen, but because their methodology will not enable them to spotthese calibration failures when they occur.D. Use test14. Use test1. The Stressed VaR model should be subject to a use test through use of StressedVaR output in risk management decisions. Stressed VaR output should be in place asa supplement to the risk management analysis based on the day-to-day output of aVaR model. The results of Stressed VaR should be monitored and reviewedperiodically by senior management.2. Where Stressed VaR outputs reveal particular vulnerability to a given set ofcircumstances, prompt steps should be taken to manage those risks appropriately.15. Date of applicationTitle III – Final Provisions and ImplementationCompetent authorities should implement these <strong>Guidelines</strong> by incorporating themwithin their supervisory procedures within six months after publication of the final<strong>Guidelines</strong>. Thereafter, competent authorities should ensure that institutions complywith them effectively.17

IV.Accompanying documentsa. Feedback on the public consultation and on the opinion of theBSG1. The European Banking Authority (<strong>EBA</strong>) officially came into being on 1 January2011 and has taken over all existing and ongoing tasks and responsibilities fromthe Committee of European Banking Supervisors (CEBS).2. On 16 November 2011, the draft <strong>Guidelines</strong> on Stressed VaR were presented tothe <strong>EBA</strong>’s Banking Stakeholder Group (BSG). The BSG provided broad commentsand suggestions, to be considered by the <strong>EBA</strong>, when finalizing the <strong>Guidelines</strong> 6 .3. On 30 November 2011, the <strong>EBA</strong> submitted the draft <strong>Guidelines</strong> on Stressed Valueat Risk (Stressed VaR) for public consultation. The consultation period ended on15 January 2012. Ten responses were received 7 . In addition, a public hearing washeld on 13 December 2011 at the <strong>EBA</strong>’s premises in London, to allow interestedparties to share their views with the <strong>EBA</strong>.4. The responses to the consultation paper were generally positive and supportive of<strong>EBA</strong>’s work and required only some clarification; however, on some paragraphs inthe consultation paper, the majority of the respondents disagreed or requestedsignificant clarification.5. A detailed account of the comments received and the <strong>EBA</strong>´s responses to them isprovided in the feedback table below. The feedback table is divided betweengeneral remarks and specific comments received from respondents and includes asection with <strong>EBA</strong>’s point of view on them and the changes made in the finalguidelines to address them.6. In some cases, several respondents made similar comments. In such cases, thecomments, and <strong>EBA</strong>’s analysis of them are included in the section of the detailedpart of this paper where <strong>EBA</strong> considers them most appropriate.6 A summary of the discussion with the BSG has been published on the <strong>EBA</strong> website in the BSG meetings andminutes section (November 2011 meeting).7 The public responses to CP 48 have been published on the <strong>EBA</strong> website together with the consultation paper.18

Feedback table on CP 48: analysis of the responses and suggested amendmentsThe first column of the feedback table makes reference to the terminology and paragraph numbering used in the CP on Draft<strong>Guidelines</strong> on Stressed Value at Risk. The last column refers to the terminology and numbering in the final <strong>EBA</strong> guidelines.CP 48 Summary of comments received The <strong>EBA</strong>’s response Amendmentsto theproposals<strong>Guidelines</strong> on Stressed Value at RiskGeneral CommentsStatus GLThe status of the guidelines is not clear. Whilethe paper referred to as ‘guidance’ thewording in their scope and level of applicationis more consistent with rules-based regulation.Furthermore there are examples in the paperwhere it would be difficult for a firm to complywith both the national rules and guidance andthe <strong>EBA</strong>'s proposed guidelines.Competent authorities have a legal requirement tonotify the <strong>EBA</strong> if they comply or intend to comply withthe guidelines within two months after publication. Inorder to maximize harmonization between competentauthorities and Member States, the <strong>EBA</strong> has chosen towrite the guidelines in a more rules-based manner.No change.Timing GLThe timing of <strong>EBA</strong>'s guidelines areunfortunate.Most banks have already submitted detailedmodel documentation for approval and oftenreceived approval under existing national rulesand guidance. Where the final <strong>EBA</strong> guidelinesdiffer from the BCBS guidance and/or nationalrules and guidanceand banks are obliged toAs the competent authorities have been involved whiledrafting the guidelines well before the publication of theconsultation paper and institutions’ model approvals, the<strong>EBA</strong> expects that the competent authorities havefollowed the <strong>EBA</strong> guidelines as much as possible whenapproving models and does not expect that institutionswill have to substantially change models to comply withNo change.19

comply with the <strong>EBA</strong> guidelines this will bedifficult and will lead to significant additionalcosts and resources which are already tied upwith the implementation of Basel III/CRDIVrequirements.the guidelines or that it will lead to significant additionalcosts.ImplementationGLThe timing for making the model changesremains unclear: we understand that the <strong>EBA</strong>will issue final guidance late in the first quarterof 2012 and that there will then be a sixmonth period in which the guidance will betransposed into national requirements atwhich point banks would have to be compliant.If so the timetable is too short to implementany potential model changes.Indeed, as mentioned in the guidelines, the <strong>EBA</strong> expectscompetent authorities to implement the guidelineswithin six months after publication of the finalguidelines. Indirectly this means that <strong>EBA</strong> expectsinstitutions to comply with the guidelines as soon asthey are implemented by the competent authorities intheir supervisory practices, but in any case within sixmonths after publication of the final guidelines.No change.The <strong>EBA</strong> believes sufficient time is given to competentauthorities and institutions to implement the guidelines.Level PlayingField with USThe US continues to be stalled in theimplementation of the Basel 2.5 Trading Bookamendments because of the Dodd-Frank Actrequirement to remove all references toratings from regulation. However, the Noticeof Proposed Rulemaking that was issued bythe US authorities in December 2010 alreadydiverges in a number of areas from the CRD 3requirements. The <strong>EBA</strong> draft guidelines furthertie EU firms to an inflexible model before theUS has implemented Basel 2.5. Flexibilityshould be retained so that a level playing fieldAlthough the comment raised is important, it is notdirectly relevant to the guidelines on Stressed VaR.The <strong>EBA</strong> will communicate the concerns raised to theEuropean Commission.No change.20

can be achieved if the US moves ahead.BaselFundamentalreview ofTrading bookThe Basel Committee’s Fundamental Review ofthe Trading Book will most likely lead to acomprehensive change in the treatment ofmarket risk. It is expected that the BaselCommittee will issue new proposals before theend of 2012. The adjustments to the modelsset out in these <strong>EBA</strong> consultation papers maybe redundant following the FundamentalReview.The guidelines on Stressed VaR relate to the CRD IIIand have also been included in the current CRR/CRDIVproposals.In any case, the <strong>EBA</strong> is following the developments ofthe fundamental review of the trading book that iscurrently conducted in Basel and it does not expect anyfinal conclusion soon nor any transpositions to theCRR/CRD following the fundamental review in the shortor medium term.No change.A. Identification and validation of the stressed periodPara 5.2Most respondents argued that the wording ofparagraph 5.2 is too vague.The provisionappearsto demand the selection of a specific12-month historic stress period for each legalentity that reports VaR. This would beextremely problematic from a processperspective. Banks would have to run theperiod selection process for multiple legalentities. Globally active banks should bepermitted to apply universally the stressperiod they select on group level.Clarification is also required about whichcompetent authority should approve theapproach to identifying the stressed periodadopted by globally active banks, given thatThe choice of the stressed period is subject to approvalby the competent authorities.The competent authorities for an EU institution are thecompetent authorities responsible for the exercise ofsupervision on a consolidated basis of this EU institutionand, in the case of an internal model also recognised ata subsidiary’s level, the competent authoritiesresponsible for the exercise of supervision of this EUinstitution’s subsidiary.The competent authorities approve the stressed perioddefined at group level according to Article 37(2) ofDirective 2006/49/EC referring to Article 129 ofDirective 2006/48/EC. They may alsorequire an EUinstitution to determine a different stressed period at asubsidiary’s level if the stressed period defined for theSection 5 hasbeen re-drafted.21

the internal models of its different legalentities are authorised by different supervisoryauthorities.group is not considered relevant to thesubsidiary’sportfolio.Para 6.4One respondent argued that the formulaicapproach implies that the period of stress tobe selected should be the period of highestVaR for the given portfolio. This is notconsistent with the stated objective ofcalculating VaR over a period of stress and(para 6.4) ‘a conservative capital outcomerather than just selecting the period of highestvolatility’. It was recommended that ajudgment override be specifically introducedfor the formulaic approach.The <strong>EBA</strong> has a preference for the formulaic approach.However it recognizes that competent authorities shouldalso take into consideration the size and sophisticationlevel of the institution when an approach is chosen.No change.On the other hand, one respondent questionedwhy the <strong>EBA</strong> would accept judgement-ledapproaches; VaR approximations and a rangeof less onerous stress period monitoringtechniques when a national regulator has ruledthese out, and urge a re-think beforecompleting the final guidelines.Para 6.7The second part of the first sentence prohibitsweighting of historical data during the stressedperiod. In our view such exclusion is pointless.In practice, weightings could be calibrated in away that ensures that a predetermined weightshall simultaneously correspond to apredetermined period. To this end, morerecent time series could be given a higherweighting than periods dating back to thePara 6.7 follows the requirement from Basel on thispoint.No change.22

Para 6.8more distant past. Since the general exclusionof weightings imposes an unnecessaryrestriction upon the methodology, we wouldlike to suggest deleting the respective part ofthe sentence.When it comes to the calibration of thestressed VaR model, this section stipulatesthat it shall be permissible to use the so-called‘antithetic data’. Yet, the method fordetermining such data is left unclear. Thesame applies to the potential compatibility ofthis provision with the other requirementsconcerning the stressed VaR.Please provide a clear definition of the conceptof ‘dynamic portfolio’ as the historical periodmust apply to the whole (IMM-approved)trading book.Paragraph 6.8 has been re-drafted. Paragraph 6.8has been redrafted.Chapter 7A VaR-based selection process creates acircular reference and is thus not ideal as itamplifies model risk.A VaR-based process is considered an acceptableapproach for the identification of the stressed period.No change.Para 7.1 Under the provisions of Paragraph 6, Section 3that historic period shall be selected whichproduces the highest VaR measure. Pursuantto Paragraph 7, Section 1, ‘statisticalassessments’ have to be produced as part ofthe justification for the respective stressedperiod chosen. In this context, once the periodwith the highest VaR measure has beenselected, it remains unclear which ‘statisticalClarify wording to ‘quantitative assessments’ if aninstitution uses a judgment-based approach.Change wording.23

assessments’ would still be necessary.Para 7.2Complete historical data may not be availablefor those periods, as acknowledged in thedraft <strong>Guidelines</strong>, and some risk factors mightnot have been known at all. Hence, at aminimum, consideration of those periodswould not be meaningful and, additionally, theapplication of quantitative analysis (alsorequired for judgement-based approaches)might lead to misleading results.Paragraph 7.2should be deleted or amended accordingly.<strong>EBA</strong> recognises comments made by the profession andhas deleted this provision in the final guidelines.Delete para 7.2.B. Review of the stressed periodChapter 8Regarding the frequency firms are required toreview the stressed period, the CRD clearlystates that this is required at a minimum onan annual basis, which is echoed in the CP,with a preference stated for more frequentmonitoring, whereas some regulators haveinsisted upon a minimum quarterly review.A competent authority may require an institution toreview the stressed period on a quarterly basis, afrequency that is in compliance with the guidelines.No change.Para 8.3A two weeks’ notice to the competentauthorities prior to applying the new stressedperiod creates an unnecessary additionalworkload for sides, i. e. banks and competentauthorities. In our view, any changes to thechoice of the stressed period – most notablywhen this involves only minor deviations interms of the result – should merely engender<strong>EBA</strong> agrees with the comment and has deleted the ‘atleast two weeks’ requirement from para 8.3.Suppression ofthe two weeksrequirement.24

the need for an ex-post notification which is tobe made during the notification of minormodel adjustments that has to take place atleast once a year.Chapter 9This section is overly-conservative regardingregular review and monitoring of theappropriateness of the SVaR time window(e.g. when de-risking trades have been putonto the book), although in practice there maynot be a large impact if the SVaR windowchanges by a few months during the financialcrisis (which is likely to be the relevant periodfor most banks). Furthermore this sectionstipulates the need for a regular review (‘ongoingbasis’) in order to ensure that thespecified stressed period is still representativefor the portfolio. We hold the view that theprovisions on the rotating review or reviewstriggered by certain circumstances (c. f.section 8) will be sufficient.<strong>EBA</strong> believes that having procedures and communicatingdocumentation on the monitoring of the stressed periodwill ensure that the stressed period remainsrepresentative on an on-going basis and providesadditional value above the requirements set out inchapter 8.No change.Para 9.2It is not clear how to implement quantitativelythe procedures described in paragraph 9.2. Itwould be very useful if some clarifyingexamples could be provided.<strong>EBA</strong> believes para 9.2 is sufficiently clear.No change.Para 9.3The mere fact that positions have been takento reduce (hedge) the Stressed VaR shouldautomatically trigger a review of the StressedVaR period. A review of the Stressed VaRperiod should be triggered only if there is a<strong>EBA</strong> has decided to delete paragraph 9.3. Paragraph9.3’s objective was to prevent regulatory arbitrage. <strong>EBA</strong>believes that this objective is addressed by otherparagraphs.Delete para 9.3.25

eason to believe that the risk structure of theportfolio has significantly changed or if a newperiod of greater stress has emerged. Thispoint is in contradiction with 15.2. Weunderstand the <strong>EBA</strong> intention here is more toprevent ‘arbitrage’ actions that are artificiallyreducing the sVaR and should be reformulatedaccordinglyPara 9.3Does this mean that the use of proxies duringhedging transactions has to be accompaniedby an appropriate reflection of the residualbasis risk?<strong>EBA</strong> has decided to delete paragraph 9.3. Paragraph9.3’s objective was to prevent regulatory arbitrage. <strong>EBA</strong>believes that this objective is addressed by otherparagraphs.Delete para 9.3.Para 9.5The decision as to whether or not a positionshould be entered into the trading book oughtto be left to the discretion of traders. Forinstance, due to operational reasons in theevent of several thousands of individual tradesper day in larger organisations, monitoringthat an individual position was merely enteredfor the purposes of reducing Stressed VaRmay become virtually impossible for riskcontrolling.This comment was linked to para 9.3 which has nowbeen deleted.No change.3. Stressed VaR methodologyChapter 10This chapter is further evidence for anincreasing inconsistency between the VaRmethodologies and the Stressed VaRcalculation. This is a source of growing<strong>EBA</strong> believes this chapter is consistent with the Baselrequirements on Stressed VaR methodology and doesnot see any reason to deviate from the Baselrequirements.No change.26

concern for us. We feel that this will only bewarranted under exceptional circumstances.Para 10.2The requirement to review theappropriateness of the square root of time toscaling up a one-day stressed VaR measure isunnecessary. If the methodology required forstressed VaR to be consistent with VaR andthere is an existing CRD requirement to justifythis approach for VaR then there should be noadditional need to review this technique asapplied to stressed VaR.The <strong>EBA</strong> believes it is even more important to reviewthe appropriateness of the square root of time forscaling up the stressed VaR than for the VaR. Thisrequirement is therefore maintained.No change.Para 10.3 (iii)The proposed equation is inconsistent with theguidance provided by our national regulator.The NSA rules require a calculation based on asixty-day average with the weekly values ofsVaR rolled each day.<strong>EBA</strong> and the competent authority referred in thecomment letter do not believe there is an inconsistencybetween the <strong>EBA</strong> guidelines and the competentauthorities’ rules on this pointNo change.Para 10.3 (iii)cWe’d like to clarify that this requirementshould not be construed as a de factorequirement for a daily computation of theStressed VaR. This requirement should besatisfied by other means or indicators. Forexample, proof that the VaR is notsystematically higher on the day of the weekchosen for Stressed VaR calculation should beconsidered sufficient.This paragraph does not require institutions to computea daily Stressed VaR. For example, proof that the VaR isnot systematically lower on the day of the week chosenfor the SVaR calculation could be considered sufficientby competent authorities.Addition of asentence to para10.3 (iii)c.Para 10.7The present proposal stipulates the need toprovide evidence for the fact that in sensitivitybased risk calculations the approach for<strong>EBA</strong> is of the opinion that full revaluation is preferable tothe use of sensitivities and that banks should movetowards full revaluation. This paragraph has beenClarificationadded to para10.727

measuring risk is adequate and that this evenholds true for the extreme values resultingfrom the stressed periods. However, since thisperiod constitutes a historic period duringwhich, more often than not, the VaR modelwill already have been in use, in our view thisquestion is academic. After all, the aboveevidence will already been provided in theform of VaR adequacy.clarified to better reflect this view.Where sensitivities rather than full revaluationare used is where this VaR model has beenapproved. Therefore this approach, on astressed period which already occurred, isalready regarded as appropriate as it isauthorised.Para 10.9We believe the use of antithetic data shouldremain an example among others in the list ofpossible adaptations of the VaR methodologyto the Stressed VaR calibration (consistentwith basel 2.5) and should either be a‘compulsory’ or ‘preferred’ method. Oneobvious reason for that is the fact that usingantithetic date is relevant for historical VaRbut much less so for Monte Carlo where itcould distort the distribution.<strong>EBA</strong> agrees and believes the use of antithetic data is thepreferable approach in case of a significant drift inmarket data as mentioned in para 10.9.No change.Para 10.10Section 7 and 8 spell out a preference for fullrevaluations in the context of Stressed VaR.There are two reasons why this gives us causefor concern: On the one hand this shall informthe approach towards risk measurement. On<strong>EBA</strong> is indeed of the opinion that banks should movetowards full revaluation.No change.28

the other hand it means that VaR and thestressed VaR model will no longer be identicalin terms of their calculation method, resultingin the deployment of two different risk models.Para 12.6This addition of a new vague, unspecifiedcapital add-on, determined by the bank itself,is not appropriate. In fact, the supervisor isalready able to impose capital add-ons, ifdeemed necessary to take account of anyshortcomings in the internal models used.<strong>EBA</strong> has changed the wording in the final guidelines toclarify this requirement.Change wordingof para 12.6.Para 13.1Separate validation of the same proxy for VaRand SVaR would be extremely burdensome. Itwould effectively force the implementation oftwo inconsistent VaR models.<strong>EBA</strong> guidelines do not explicitly require a separatevalidation for VaR proxies used for Stressed VaRpurposes. However, institutions are required to showappropriateness of the use of same proxies for VaR andStressed VaR and to consider if/when a separatevalidation of proxies for Stressed VaR is necessary.No change.D. Use TestsChapter 15The concept of a use test for stressed VaR isdifficult. The more remote in time a stressperiod becomes, the less relevant risk driversand correlations become, and the harder itbecomes to demonstrate model use. Similarly,the use of the stressed VaR to validate theimpact of current VaR modelling choices doesnot appear very relevant.<strong>EBA</strong> believes the Stressed VaR can be used as anadditional risk management tool and can support riskmanagement decisions.No change.29

We recommend removal of references to theUse Test and insertion of a requirement thatregulators observe how firms use stressed VaRand provide guidance.Para15.1In the Stressed VaR, the current reference ofthe risk parameter is replaced by a historicone. As a result, portfolio management andthus the SVaR-based use test become moredifficult. We therefore feel that monitoringstressed VaR at various levels of aggregationwould be unconstructive. Also, it is not clear tous in how far this requirement deviate fromthe CRD III. Recommend to delete thewording ‘at different aggregation levels’ inpara 15.1.<strong>EBA</strong> agrees with the comment and has deleted ‘atdifferent aggregation levels’ in para 15.1.The examples given in para 15.1 form a non-exhaustivelist and indeed there are different ways to satisfy theuse test requirements.Delete ‘atdifferentaggregationlevels’ in para15.1.Please clarify that the examples given in thisparagraph (limit setting, reporting andescalation procedures, etc) are examples onlyand that there are different ways to satisfy theuse test.Para 15.2We believe the use of the Stressed VaR as atool to validate the current VaR modellingchoices is very questionable. This would notbe very objective and cause us to ‘look overour shoulder’ when we should rather belooking forward when improving VaRmodelling.<strong>EBA</strong> agrees with this comment and has deleted the firstsentence of this paragraph.Deletion of thefirst sentence ofpara 15.2.30

V. Confirmation of compliance with guidelines andrecommendationsDate:Member/EFTA State:Competent authority:<strong>Guidelines</strong>/recommendations:Name:Position:Telephone number:E-mail address:I am authorised to confirm compliance with the guidelines/recommendationson behalf of my competent authority:YesThe competent authority complies or intends to comply with the guidelines andrecommendations:Yes No Partial complianceMy competent authority does not, and does not intend to, comply with theguidelines and recommendations for the following reasons 1 :Details of the partial compliance and reasoning:Please send this notification to compliance@eba.europa.eu 2 .1 In cases of partial compliance, please include the extent of compliance and of non-compliance and provide thereasons for non-compliance for the respective subject matter areas.2 Please note that other methods of communication of this confirmation of compliance, such as communicationto a different e-mail address from the above, or by e-mail that does not contain the required form, shall not beaccepted as valid.31