Ernst & Young Global Limited is a company limited by guarantee registered in England and Wales No. 4328808.

Ernst & Young Global LimitedBecket House1 Lambeth Palace RoadLondonSE1 7EU

Tel: +44 [0]20 7980 0000Fax: +44 [0]20 7980 0275ey.com

Tel: 023 8038 2000

Basel Committee on Banking SupervisionBank for International SettlementsCH-4002 BaselSwitzerlandSubmitted electronically through the BIS website (www.bis.org)

30 April 2015

Dear Basel Committee members,

Invitation to comment – Consultative Document: Guidance on accounting for expectedcredit losses

Ernst & Young Global Limited, the central coordinating entity of the global EY organisation,welcomes the opportunity to offer its views on the above Consultative Document.

We agree with the Committee that there needs to be a high quality implementation of the newexpected credit loss (ECL) approach to measure the impairment of financial instruments. Itsintroduction will be a major challenge for the banks, auditors, users of the accounts andsupervisors. The challenge for the banks for all ECL approaches includes sourcing historicalloss data, making forecasts of future economic conditions and adjusting historical experienceto reflect current products and economic circumstances and expectations. For those banksreporting under IFRS 9, there is the additional requirement to determine when there hasbeen a significant increase in credit risk, which will be especially challenging when there islittle forward-looking information available at the individual obligor level and thisdetermination has to be made on a collective basis.

It is worth stressing that the US ECL model has not yet been finalised and there are a numberof IFRS 9 interpretation and implementation issues that will need to be discussed at furthermeetings of the IASB’s Impairment Transition Resource Group (ITG). Some of thesediscussions may lead to clarification of IFRS 9 by the IASB or the IFRS InterpretationsCommittee. Meanwhile, the Enhanced Disclosure Task Force will develop common ECLdisclosure practices, while the International Auditing Standards Board is writing a standard onhow to audit ECLs. We are pleased that the Basel Committee is involved in the ITG and,ideally, there should be no need for separate guidance on the ECL model to be provided bybanking supervisors. However, if such guidance is to be issued, we strongly agree that anyguidance should be issued by the Basel Committee, rather than by individual countrysupervisors. We agree with the Committee that implementation of the ECL model byinternationally active banks should be as consistent as possible around the world and a singlesource of guidance will help promote this. We hope that supervisors will continue to worktogether to apply this guidance in a consistent manner. We are also pleased that theCommittee has sought to provide the guidance quickly, given that it will have a major impacton the manner in which banks implement the new standard and major investment decisionsneed to be made in the next few months if a high quality implementation is to be achieved inthe available time.

2

In this letter we have sought to provide constructive comments on the ConsultativeDocument, based on our understanding of ECL accounting requirements and, particularly,IFRS 9 and the challenges that it will pose. This is set out under three headings:

1) Comments on the structure of the draft guidance

2) Whether the guidance can be read to conflict with the requirements of IFRS 9 FinancialInstruments

3) Other comments on the content of the draft guidance

We also attach an appendix that sets out wording of the draft Guidance that is unclear, orcould be improved, together with our suggested solutions.

1) Comments on the structure of the draft guidance

a) Scope of the document

The title of the document is “Guidance on Accounting for Expected Credit Losses”;paragraph 1 limits the scope to include credit risk practices affecting the assessment andmeasurement of credit loss allowances. This is narrower than the scope of the 2006guidance, as described in paragraph 2, that the new Guidance would replace. However,the document actually contains a mixture of guidance on the accounting for expectedlosses and, in addition, guidance on wider aspects of credit risk management; the title ofthe main section is “Credit practices that interact with expected credit lossmeasurement” (emphasis added).

Guidance that does not affect the assessment and measurement of credit loss allowancesincludes:

i) Paragraph 19 (a), which refers to compliance with applicable laws and regulations

ii) Paragraph 19 (b), which includes reflecting allowances appropriately in supervisoryreports

iii) Paragraph 24 (n), which refers to policies and procedures on write offs and recoveries,more broadly than just the accounting for these events

iv) Paragraph 24 (j), which requires banks to have a realistic view’ of their lendingactivities

v) Paragraph 31, which refers to sound underwriting practices, a topic that goes beyondthe initial assessment of the probability of default and into the bank’s tolerance of risk.Also, a subsequent reduction in risk tolerance does not mean that the risk of anexposure has significantly increased.

3

vi) Principle 7 and paragraphs 31 and 66, which refer to the pricing of credit risk and thatit should reflect inherent risks. Pricing reflects many factors, such as required profitmargins, costs of funding and capital, the value of ongoing banking relationships andcompetitive pressures and aspirations. It also depends on the level and quality ofcollateral, which is not necessarily related to the probability of default. This is a muchwider subject than the measurement of ECLs.

vii) Much of the guidance in paragraph A8 on lending to high risk exposures is not aimedat the measurement of expected credit losses

We understand that supervisors are keen to emphasise the importance of embedding ECLaccounting in credit risk management practices, and wish to use the introduction ofIFRS 9 to upgrade credit risk management more broadly. However, the mixing of guidanceon both the application of an ECL approach and more general credit risk managementmakes for a complex document, not least because the language that supervisors use todescribe credit risk management differs from the language normally used by accountantsand which appears in accounting standards such as IFRS 9. Meanwhile, the languageintroduced by the Committee includes a large number of superlatives (e.g., ‘highestquality’, ‘best practices’, ‘full spectrum’ ) to describe the Committee’s aspirations, whichdo not sit comfortably alongside accounting standards. The guidance would certainly beclearer if it focused on ECL accounting and how it should build on and link to credit riskmanagement practices. Much of the wider material could usefully be included by crossreference to existing guidance, since changes in the accounting model should notnecessarily change the way that credit risk is managed.

A number of the issues we raise in this response arise because the present drafting mixesthe two separate subjects of risk management and an ECL accounting model. If theCommittee intends to continue with the inclusion of the wider credit risk guidance thenthere will need to be a number of adjustments to the text (not least to its title and thescope as described in paragraph 1) as we recommend in this letter and the appendix.

b) Banks in scope

We understand why the Committee has chosen to provide guidance to those banks thatare internationally active or sophisticated in the business of lending and we agree that, asset out in Principle 1, the credit practices and internal controls to determine credit riskpractices should be commensurate with the size, nature and complexity of lendingexposures. However, as drafted, the guidance suggests that a ‘high quality’implementation, without concern for cost or effort should be applied to all of the activitiesof internationally active banks. We believe that the application should be ‘proportionate’,to those activities, so that a less costly approach may be suitable for smaller portfolios forwhich the accounting benefit is small and where the effect is not material. As theguidance will be read by risk managers and others who may not be familiar with normalaccounting concepts of materiality, it would be helpful to make it clear thatimplementation should not lose sight of materiality.

4

Further, the quality of information available to implement the ECL model will often vary bylocation within a bank, so that, in some countries, it may be necessary for the approach tobe less sophisticated than in others. A more flexible, proportionate approach would beconsistent with that used currently for supervisory purposes, as not all parts of a bank’sbusiness are regarded as suitable for A-IRB capital models. Any approach must beoperationally robust, with regard to the local environment and data constraints.

It is also worth stressing that the determination of which banks are deemed to beinternationally active banks and those banks more sophisticated in the business of lendingwill need to be made by banking supervisors. Also, to the extent that the supervisorsadopt a proportionate approach for other, less complex banks, as mentioned in paragraph12 of the draft Guidelines, the supervisors will need to set out their expectations of this‘proportionate approach’.

c) Individual versus portfolio assessment and measurement

The document would be clearer if it were organised so as to differentiate better theaspects relevant for situations where forward-looking information is available at theindividual exposure level, and for those where it is not and so banks must ‘overlay’obligor-specific information with a collective assessment or calculation.

► At one extreme, for larger or more specialised corporate exposures, any assessmentof significant increase in credit risk or estimation of ECLs is likely to be bespoke forthe customer. For such exposures, the requirements of paragraphs 29 (a) (requiringbanks to demonstrate how ECL estimates will fluctuate with changes in forwardlooking assumptions), and (d) (on backtesting), will be less applicable. Also, whileforward-looking information including macroeconomic forecasts will inform the riskassessment and estimation of ECLs, it will often not be possible to attribute theallowance to individual macroeconomic ‘drivers’ and, hence, it will not be possible toprovide the information required by paragraphs 34, 76, 79, and A21.

► At the other extreme, for many retail exposures, significant use will need to be madeof collective assessments and calculations, both where there is a significant increasein credit risk, and for measuring the ECLs. For this group, paragraphs such as parts of27 (borrower assessment) and 37 and 40 (credit rating systems) cannot often beapplied. Also, the procedures required by paragraph 27 may be carried out on initialassessment but will not all occur in subsequent periods, e.g., it would be rare that thebank will fully update its records of a mortgage customer’s income and financialposition. Subsequent assessment tends, instead, to be carried out at a portfolio level.Meanwhile, if a bank uses a credit rating system for retail customers, this will notusually make use of forward-looking information, but will be driven by delinquency,credit scores and other historical data, which is why some form of overlay will usuallybe needed to assess significant increases in credit risk and to measure ECLs.

► There will also be a third category of exposures that fall between the two extremesand for which some, or parts, of these paragraphs will be relevant but not others.

2) Whether the guidance can be read to conflict with the requirements of IFRS 9

5

While we understand that it was not the Committee’s intention that the guidance shouldconflict with IFRS 9 and that representatives of the IASB have not identified any suchconflicts, the wording, in places, is not very clear and can be read to conflict. The maininstances include the following:

a) Paragraphs 24 (f) includes as a factor that might require a qualitative adjustment, creditconcentrations. While we understand the importance of this topic for risk managementand supervision and that it features in the way that loss allowances are required to bemeasured in the US, we do not understand how this would be relevant for application ofthe IFRS 9 ECL model or how, in practice, such an adjustment would be made.

b) Paragraph A5 states, on the subject of what is a ‘default’, “the list of elements provided inthe Basel framework as indications of unlikeliness to pay should be supplemented withother factors … either on an individual or collective basis and adjusted to incorporatecurrent conditions and forward-looking information.” This seems to combine ‘default’ withthe assessment of significant increases in credit risk, and implies that banks shouldextend the notion of default in the 12-month ECL estimation to situations when an actualdefault will have not yet occurred. This would not comply with IFRS 9 and should beremoved. Also, the list of indicators of unlikeliness to pay in the Basel framework includesmaking an ‘account-specific provision’. Given that all exposures will have an account-specific provision under an ECL model (even if this is sometimes calculated collectively),this would logically mean that all exposures are in default from day one.

c) We do not understand why paragraph 24(f) says that a qualitative adjustment to theallowance may be required for a change in underwriting standards or lending policies –such changes do not necessarily affect the credit risk of existing financial assets. Forinstance, a change in underwriting standards or lending policies might reflect a change inrisk tolerance or strategic direction. (Banks will have a different attitude to risk and itspricing if they are seeking to break into a market or to reduce their involvement, whichwould not necessarily affect the risks of existing exposures). Similarly, paragraphs 24(q)and 28(c) refer to ‘changes in the bank’s business model’, even though such a changewould not necessarily affect the credit risk of existing exposures. We agree that a changein collection methods may affect the level of expected losses, but this would be a muchnarrower type of change than those described in the draft Guidance.

d) Paragraph 47 states that when the credit risk increases on a group basis, the entiregroup should migrate to a higher credit risk rating. Combined with the assertionsmade in paragraph A27, this implies that the whole group should subsequently bemeasured on a lifetime ECL basis. This is supported by paragraph A35, which states“the Committee expects that, in instances where it is apparent that one or moreexposures in a group have experienced a significant increase in credit risk, therelevant group or subgroup will transfer to LEL measurement of ECL”. This wouldseem to conflict with IFRS 9, as illustrated by the top-down approach shown inExample 5 of the Implementation Guidance to IFRS 9, in which only 20% of floatingrate loans are deemed to be measured on a lifetime ECL basis. We understand that

6

this reading was not intended and we recommend the wording should be amended soas to make this clear.

e) In paragraph 63, the reference to ‘determine either a single amount …’ can be read toconflict with the requirement in IFRS 9 B5.5.42 to reflect ‘at least two outcomes’.

f) Paragraph A7 states that “a bank must be able to demonstrate that these exposures havenot experienced a significant increase in credit risk since initial recognition” (emphasisadded). This can be read to imply that exposures should be recorded using lifetime ECLsby default unless the bank can justify using a 12 month basis. This would subtly amendthe requirement of IFRS 9 paragraph 5.5.4 that a lifetime basis should be used wherethere is a significant increase in credit risk considering all reasonable and supportableinformation. (Moreover, in practice it is very difficult, if not impossible, to demonstrate anegative).

g) In paragraph 80, the statement that ‘recoveries of amounts previously written off’ shouldbe an item in the allowance reconciliation is incorrect under IFRS. Under IFRS, a write-offis a derecognition event and the accounting for a subsequent recovery would not makeuse of an allowance account.

3) Other comments on the content of the draft guidance

While we list a number of concerns about the drafting of the document in the Appendices, thefollowing are the most problematic, in addition to those already listed:

a) The document uses may terms that are different from those used in IFRS 9, are undefinedand are often unclear. These include words such as ‘sound’, ‘robust’, ‘timely’, ‘notch’,‘suitable’, and ‘appropriate’. The guidance would be more useful if the words of the IASBare used to the maximum extent feasible and that any new terminology introduced to helpprovide guidance is defined and used consistently so that the Committee’s meaning isclear without the need for further guidance.

b) There are several references to ‘forward-looking information and macroeconomic factors’,whereas macroeconomic factors are an example of forward looking information. Given thenumber of times that this phrase is used in the document, it would be useful to listexamples of what is considered “forward-looking information” in more detail. For example,does it include borrower-generated financial forecasts, or industry published growthprojections, or analyses of predicted technological change? These examples, whilecertainly forward-looking information, could potentially double-count macroeconomicfactors like GDP growth, interest rate changes, or measures of inflation. The two items,forward-looking information and macroeconomic factors are not as separable as thedocument seems to suggest. Also, the provision of examples of what types of forward-looking information might be used to assess ECLs in an individual exposure or at portfoliolevel, and how to avoid double counting of their effect, would provide useful guidance tobanks and help clarify supervisory expectations.

7

c) The Guidance is inconsistent in its emphasis on the level of judgement that will berequired in applying an ECL model. For instance, paragraph 61, rightly, points out that itmay not always be possible to demonstrate a strong link between the informationavailable and the credit risk of some exposures or portfolios. Yet paragraph A29 statesthat ‘reliable’ calibration of the linkages between drivers and increase in credit risk will be‘critical’. In practice, calibrations will always be estimates, hypotheses based on pastexperience and judgements about the future. And, as history rarely repeats itself, theseestimates will often differ from how situations actually unfold. It is important that banksstrive to find calibrations that are as reliable as is practically achievable, but to describethem as ‘reliable’ implies that the exercise is much simpler and more mechanical than itwill actually be. There is a risk that the guidance in A29 could lead to solutions that areunduly complex, are too focused on past data and imply a spurious level of accuracy.

The application of the ECL model will certainly require the application of experiencedjudgement, with the consequence that no two banks will come to the same assessment ormeasurement of two identical exposures. But the last line of paragraph 61 probablyplaces too much emphasis on the use of judgement (since it implies that ‘experiencedcredit judgement’ can substitute for forward-looking information and is not necessarily‘supportable’ evidence as required by the standard. It would be helpful to provideconsistent guidance that seeks a middle way between the approaches described inparagraphs 61 and A29.

d) Paragraph 75 indicates that the Committee expects disclosure of the sensitivity of ECLestimates to change in assumptions. This goes beyond what is currently required byIFRS 7 or IFRS 9 and, without further guidance, could result in disclosures with limitedcomparability between entities. It would probably be better to remove disclosure as atopic from the Guidance, perhaps removing Principle 8, and to feed these concerns intothe planned project of the Enhanced Disclosure Task Force to devise suitable disclosuresfor the ECL model.

Should you wish to discuss the contents of this letter with us, please contact Tony Clifford on+44 (0)207 951 2250.

Yours faithfully

8

Appendix: Wording of the draft Guidelines that is unclear, or could be improvedCommentnumber

Paragraphreference

in thedraft

Suggested solution

1 Para 1 The paragraph states that “thescope of credit risk practices islimited to those practicesaffecting the assessment andmeasurement of allowances …”yet the title of the secondsection of the document refersto “practices that interact withexpected credit lossmeasurement” and containsguidance that does not relate toassessment and measurement ofallowances.

We recommend that thescope of the document isreassessed. If the objective isconfirmed to be consistentwith paragraph 1, thencertain material should bedeleted, as set out in ourcover letter. If the Committeedecides to keep the widerscope, the title of theGuidance and paragraph 1should be amended.

2 Para 7 In this paragraph the term‘robust’ is used for the first ofmany times. We understand thatthe word is already used inregulatory guidance, but what isit supposed to mean in thecontext of assessing andmeasuring expected creditlosses? It would seem that itsmeaning is intended to bebroader than its common usageas ‘resilient’.

We recommend that theCommittee does notintroduce new terminologyover and above that alreadycontained in the standard or,where to do so would helpprovide useful guidance, toprovide a set of definitions.‘Robust’ should ideally bereplaced with a word thatdescribes more preciselywhat the Committee meansin each context.

3 Para 11 While we agree thatinternationally active banks andthose banks which are moresophisticated in the business oflending should implement anECL model well, most will havesmaller operations for which theamount of investment andcomplexity of approach shouldbe commensurate with theirscale and risks, as required by

We recommend adding“commensurate with the size,nature and complexity oftheir lending exposures” tothe end of the paragraph.

9

Principle 1.

4 Para 13 The scope of the paper isunclear as to which activities ofbanks are covered by theGuidance and which are not. Forinstance, are deposits with otherbanks and central banks deemedto be ‘lending exposures’ andwhen would reverse repos orleasing activity fall within thescope of the Guidance?

The reference to ‘debt securities’(paragraph 13) does not workwell for a document withinternational relevance. NeitherIAS 39 nor IFRS 9 differentiatesbetween loans and securitiesand some items held by banksthat arise from lending activitiesmay be securities in legal form.

It would be helpful to clarifywhat is meant by ‘lendingexposures’, possibly byadding to the list of what it isnot deemed to cover, such as(potentially) deposits, reverserepos and finance leases, allof which are (normally) in thescope of the IFRS 9 ECLmodel.

While we support the ideathat the low risksimplification should beavailable for exposures thatdo not arise from lendingarrangements, we suggestamending the text to ‘such asdebt securities and otherfinancial assets traded inactive markets’.

5 Para 21 In the third sentence, thereference to ‘updated eachreporting period’ is notconsistent with the manyreferences in the document suchas ‘regular’ and ‘timely’ but withlittle guidance on what theymight mean. For instance, whenparagraph 84(a) uses the word‘timely’ does this mean banksmust continuously assess allcollateral? Paragraph 42requires credit ratings to bereviewed whenever newinformation is received, yet aformal review need only beannual, while paragraph 49 uses“as early as possible” andparagraph 55 refers toperiodically “consistent with thebank’s financial and regulatory

It would be helpful if theguidance were clearer on thisissue.

10

reporting requirements”.

At the 22 April ITG meeting, theIASB members appeared toconfirm that IFRS 9 onlyrequires impairment allowancesto be assessed and measured athe period end, except whenassets are derecognised.

6 Para 21(secondsentence)

The reference to ‘interpretation’is erroneous in this context,since interpretation of theaccounting framework is notrelated to the use of informationand assumptions across thebank.

Remove the reference to‘interpretation’.

7 Para 24(a) A method cannot include aprocess that ‘equips’. A methodcan only be designed to achievean outcome and cannot, on itsown, guarantee that outcome,as that will be dependent onexecution.

The wording should perhapsrefer to a ‘process designedto equip’.

8 Para 24(b) The statement thatmacroeconomic factors ‘may beat the local level’ is acontradiction and even thereference to ‘regional’ isquestionable, given that‘macroeconomic’ is normallydefined as relating to a nationaleconomy.

Replace ‘macroeconomic’with ‘economic’.

9 Para 24(f) This sub-paragraph indicatesthat a qualitative adjustmentmight be necessary for “changesin expectations ofmacroeconomic trends andconditions”.

Also, it is unclear how riskconcentrations would affect

Wouldn’t any adjustmenthave to be quantitative, evenif arrived at using judgementrather than an algorithm?

Delete the reference to‘credit concentrations’, or

11

expected loss allowances underIFRS 9.

else explain how they mightaffect an IFRS ECLapplication.

10 Para 24(h) Any approach that describes thesituations in which methods,inputs or assumptions wouldneed to change would beuninformative, since it wouldmerely be when they are nolonger appropriate.

More helpful might be torefer to the triggers thatmight prompt a reassessmentof the methods, inputs orassumptions.

11 Para 24(q)and 80

Paragraph 24 (q) refers to‘collective allowances’, whileparagraph 80 states that theCommittee expects thereconciliation of the allowanceaccount to be providedseparately for ’collective andindividual allowances’, implyingthose that are measuredcollectively or individually.However, the example given inparagraph IG 20B of theImplementation Guidance ofIFRS 9 refers to lifetime lossesthat are assessed collectivelyand individually, implying thosethat are determined to requirelifetime ECLs on a collective orindividual basis. Did theCommittee intend to add a newrequirement?

Note, also, that if ameasurement overlay is used toinclude forward-lookinginformation, then the allowancefor an individual asset would bemeasured partly individually andpartly collectively. Splitting outthese two components wouldnot seem to be very informative.The concept of a ‘collectiveallowance’ makes more sense in

In paragraph 80, replace‘that this will be providedseparately for collective andindividual allowances’ with‘that lifetime loss allowanceswill be provided separatelyfor those exposures assessedindividually and collectively’.

The term ‘collectiveallowances’ should beavoided.

12

the context of IAS 39 thanIFRS 9.

12 Para 28(e) Under IFRS, ’charge-off’ is amere accounting concept,reflecting the timing of when aloan and any associatedallowance are removed from thefinancial statements.Consequently, it is not a factorthat would affect the bank’sability to recover amounts due.

Remove the reference to’charge-off’, or else make itclear how the term might berelevant for credit riskassessment

13 Para 29(b) This paragraph talks aboutdocumenting the process fordetermining the time horizon,including ’how ECL is estimatedin the period following thatwhich is reasonably estimable’and then in a footnote, indicatesthat the term ’reasonablyestimable’ is not included in IFRS9. The Committee shouldconsider removing thisreference to ‘reasonablyestimable’ as it implies thatanything beyond that horizonmust not be reasonablyestimable, and will encourageentities not to try to developbetter methods and systems andcollect better data sets toforecast losses for longerperiods. While we understandthe Committee’s aim in thisparagraph, the wording may bemisleading without the contextthat has been explored in the USdiscussions, including howprojections should revert tohistorical experience.

Consider removing thereference to ‘reasonablyestimable’, or else expandthe discussion to providesome useful guidance on thistopic for IFRS-reportingbanks.

14 Para 29(d) The reference to backtestingcould be inferred to includebacktesting the macroeconomic

We recommend that thiswording is amended to avoid

13

assumptions themselves, whichwould not be a valuable exercise.

this risk.

15 Para 30 The term ‘full spectrum ofinformation that is relevant’implies that the bank shouldhave access to and use allinformation that might ideallyexist, whether or not it actuallyexists and whether or not it isaccessible for the bank. Thismakes the requirementunachievable or unclear.

Further, it is unclear whetherany assessment of futuremacroeconomic conditionsshould be probability weighted,such that the uncertain outcomeof a high impact future event(e.g., an independencereferendum) should be broughtinto the ECL calculation basedon the relative probabilities ofthe possible outcomes. Theproblem with such an approach,particularly if the sameassumptions are required to beused to underpin budgets,strategic and capital plans, isthat a probability-weightedforecast of the future is one thatwill certainly never occur. Also,the quantitative results of suchhigh magnitude events will oftenbe impossible to model with anyconfidence. It makes more senseto use something like aconsensus view when it comesto macroeconomic assumptions,which will be informed by anunderstanding of futurepossibilities. The potential effectof such uncertainties wouldmore usefully be qualitatively

The term should be replacedwith something that is clearerand more achievable. Thewords in paragraph A6, ‘allreasonably availableinformation’ are probablycloser to what the Committeeintends.

It is possible that discussionon this subject may occur atfuture ITG meetings. If not,then it would be useful forthe Committee to make itsexpectations clear. In anyevent, the wording aboutusing the ‘full spectrum’ isunhelpful.

14

disclosed for financial reportingand quantitatively captured onlythrough stress testing forprudential purposes.

16 Principle 4 Principle 4 is strangely worded.How could ECLs calculatedunder IFRS 9 and US GAAP bothbe said to be ‘consistent’ withwhat the Basel Core Principlesdescribe as ‘adequate’? We notethat the Basel Core Principlesrefer to “If […] provisions aredeemed to be inadequate forprudential purposes (e.g. if […]the provisions do not fullyreflect losses expected to beincurred), the supervisor has thepower to require the bank toadjust its classifications ofindividual assets, increase itslevels of provisioning, reservesor capital”. The ‘losses expectedto be incurred’ originallyenvisaged by the Core Principleswere presumably regulatory 12-month ECLs, and the text wasdrafted to reflect an incurredloss accounting model. If,alternatively, these expectedlosses are taken to refer tolifetime ECLs, then the wordingwould seem incompatible withthe phased recognition model inIFRS 9 that starts with 12-month ECLs and also conflictswith the IFRS 9 accounting forpurchased or originated credit-impaired financial assets.Instead, this sounds more likethe FASB lifetime ECL model.

We suggest that referencesto the Basel Core Principlesare deleted, unless thePrinciples are to be amended.

17 Para 33 It would be helpful to explainwhat is meant by ’components of

Clarify what is meant by’components of credit risk’.

15

credit risk’.

18 Paras 33to 42

Paragraphs 33 to 42 (relating tocredit assessment and creditrating systems) do not relate toPrinciple 3, which is about‘appropriate grouping’ of lendingexposures on the basis of sharedcredit risk characteristics underan ECL accounting approach.

There seems to be a principlemissing, to which theseparagraphs would betterrelate, or else the principleshould be reworded to deletethis guidance and, instead,just explain how ratingsystems should be used oradjusted for ECL accountingpurposes.

19 Para 36 While significant increases incredit risk may be assessed on aportfolio basis, it would beunusual to assess credit ratingson a portfolio basis unless thechange in conditions affects thewhole portfolio. For instance,when a top down approach isused to assess deterioration, asset out in IE 39 of theImplementation Guidance to thestandard, it would be unlikelythat all floating rate borrowersor an arbitrary 20% of suchborrowers would bedowngraded.

Amend the wording to makeit clear that the use of‘portfolio’ in this context iscloser to the ‘bottom up’assessment in IE 38.

20 Para 41 What is meant by ‘the full rangeof credit risk’? Does this refer toall sources of risk or all possibleoutcomes?

Clarify what is meant by ’fullrange of credit risk’.

21 Para 42,48 andA18

Para 42: indicates that ‘creditrisk ratings assigned shouldreceive a periodic formal review(e.g. at least annually or morefrequently if required in ajurisdiction) to reasonablyensure that those ratings areaccurate and up to date.’

Does the wording “if required in

If the Committee meant thatthe bank should carry out aformal review as often as itreports to shareholders,paragraph 42 could berevised to: ‘e.g. at leastannually or more frequentlyif the bank issues interimfinancial statements’. Inpractice, it may be possible to

16

a jurisdiction’ refer to where theentity reports to shareholdersmore often than once a year? Ifso, in practice, entities may havesome discretion as to thefrequency with which theyprepare interim financialstatements, so ‘if required’ is notquite right.

Or does the Committee expect aformal review once a year unlessbanking regulatory requirementswould require a formal reviewmore often?

Similarly, paragraph 48 indicatesthat ‘the group of exposuresassigned should receive aperiodic formal review (e.g. atleast annually or morefrequently if required in ajurisdiction) to reasonablyensure that those groupings areaccurate and up to date’.

Paragraph A18 discusses theneed for timely transfer to LELas soon as credit risk hasincreased significantly. How isthis consistent with paragraph42 and 48 that seem to indicatean annual review might besufficient?

comply with the standardwithout a full formal riskreview every quarter.

Meanwhile, paragraph 48could be perhaps reworded tosay the groupings should bereviewed when significantcircumstances change.

It would be helpful if theseexpectations wereharmonised using consistentlanguage.

22 Para 45 Paragraph 45 in unclear.Presumably it means that thegroups should be sufficientlygranular that increases in creditrisk in some sections of thegroup should not be obscured,but that point is already made inthe next paragraph.

Clarify or possibly deleteparagraph 45.

17

23 Para 46 The second sentence shouldrefer to ’information obtainedthat would indicate’ changes incredit risk after initialrecognition (see paragraph 48)and logically would follow thatparagraph.

Amend accordingly.

24 Para 47and A 33

Paragraph 47 states that whenthe credit risk increases on agroup basis, the entire groupshould migrate to a higher creditrisk rating. Combined with theassertions made in paragraphA27, this implies that the wholegroup should subsequently bemeasured on a lifetime ECLbasis. This is supported byparagraph A35, which states‘the Committee expects that, ininstances where it is apparentthat one or more exposures in agroup have experienced asignificant increase in creditrisk, the relevant group orsubgroup will transfer to LELmeasurement of ECL’. “Thiswould seem to conflict withIFRS 9, as illustrated by the top-down approach shown inExample 5 of theImplementation Guidance toIFRS 9, in which only 20% offloating rate loans are deemedto be measured on a lifetimeECL basis.

This paragraph says that, ‘ininstances where it is apparentthat one or more exposures in agroup have experienced asignificant increase in creditrisk, the relevant group orsubgroup will transfer to LELmeasurement of ECL even

Presumably these paragraphsshould say that ’a portion ofthe group or a sub-groupshould migrate or willtransfer…’

18

though it is not possible toidentify this on an individualexposure basis’. As worded, thismeans that it only takes oneexposure within a group to showevidence of a significantincrease in risk for the allowancefor the entire group to berequired to be measured on alifetime ECL basis. This wouldclearly be wrong and wouldconflict with the standard.

We understand that this readingwas not intended.

25 Para 52 The reference to ‘simpleaverages’ is hard to understand.We would expect that anyaverage would at least beweighted by the size of the loansor defaults.

Was this intended to read,instead, ‘simple techniquessuch as averages …’?

26 Para58(c)(ii)

Why is there reference to’plausible stresses’? Does thisjust mean that the model canaccommodate relatively stressedassumptions, or is there asuggestion that stressedassumptions form part of thecalculation of ECLs (which wouldconflict with the standard)?

Explain what is meant by’plausible stresses’ or elsedelete the reference.

27 Principle 6and paraA49

Principle 6 refers to ‘reasonablyavailable’ information. This is intension with the statement inparagraph A49, that banks’should develop systems andprocesses to use all reasonableand supportable information’(emphasis added). To use all theinformation would be undulyexpensive, operationallycomplex and, correspondingly,less robust. There will always be

If the Committee had notintended for banks to look for‘all reasonable information’,but rather to ensure a‘robust’ implementation ofthe standard, then thewording of paragraph A49should be amendedaccordingly.

19

the need to apply judgement indeciding what information touse, to ensure that process ismanageable. Also, as with allaccounting estimates, there is atrade-off between reliability ofevidence on the one hand andrelevance on the other. Someinformation will be more reliablethan others and some will bemore relevant.

28 Para 63 This says that ‘in estimatingECL, banks may determineeither a single amount or arange of possible amounts’ Thiscan be read to conflict with therequirement of paragraphB5.5.42 of IFRS 9, whichrequires a probability-weightedcalculation that reflects at leasttwo possible outcomes.

It is suggested that thiswording is deleted or elseconformed to the wording inthe standard.

29 Para 75 The bank’s basis for groupinglending exposures into portfolioswith similar creditcharacteristics is not a ‘policy’ or‘definition’ – it is a judgementmade in making an estimate.

It is also unclear what is meantby ‘management and users havedifferent objectives’.

We suggest the wording beamended to ‘to highlightpolicies, definitions andjudgements made that areintegral to the estimation …’

We suggest deleting thesewords.

30 Para 77 Paragraph 77 includesdocumentation of process andduplicates (or perhaps belongsunder) principle 3.

We recommend that processdocumentation is deal withunder Principle 3.

31 Para 79 In paragraph 79, does theCommittee require quantitativedisclosure? If so, it would behelpful to make this clear.

Clarify whether quantitativedisclosures are required.

As we write in our coverletter, we recommend thatthe guidance should not

20

address disclosures, sincethis will duplicate the EDTFinitiative.

32 Para 80 The statement that ’recoveriesof amounts previously writtenoff’ should be an item in theallowance reconciliation isincorrect under IFRS. A write-offis a derecognition event and theaccounting for a recovery willnot involve the allowanceaccount.

Remove the reference to‘recoveries of amountspreviously written off’ andmake this a separatedisclosure recommendation.However, as we write in ourcover letter, we recommendthat the guidance should notaddress disclosures, sincethis will duplicate the EDTFinitiative.

33 Para 81 Should product concentrations,etc., have been referred tobefore paragraph 81 in principle8? It seems odd to ask for itemsto be regularly reviewed withoutasking them to be disclosed inthe first place. Also, productconcentrations seem to be aseparate issue from ECLs.

We suggest that thereference to ‘productconcentrations’ is deleted oramended accordingly.

34 Para 82(c) It is unclear why the timelyidentification of changes incredit risk is emphasised for’riskier exposures’ when it isalready required by the draftGuidance for all exposures (seeparagraphs 42, 48 and 87).

Clarify the emphasis for’riskier exposures’.

35 Para 84(c) The reference to the allowancebeing appropriate in accordancewith the relevant accountingrequirements and in relation tothe total credit risk in the bank’sportfolio, implies that theallowances could be appropriatein accordance with relevantaccounting requirements andyet not be appropriate inrelation to the total credit risk

The meaning of this isunclear and it should bedeleted or reworded.

21

exposure. This cannot be right.

36 Para A1 IFRS 9 is not ‘the impairmentstandard’.

Also, contrary to what is writtenin A1, nil allowances will not berare where the exposures areadequately collateralised, suchas for much mortgage lending.This should be addressed in themain text, rather than explainedin the footnote, for two reasons:

a) As many exposures will befully collateralised, a nilallowance will be common.

b) The message in the secondhalf of the footnote: ’asvaluation of collateral maychange over the life of theloan’ seems too importantto be relegated to afootnote.

Delete the word ‘impairment’in the first line.

In line 5, we recommendrevision to ‘exposures, andthat (unless the exposure isfully secured by collateral) anil allowance will be rare’ andthat footnote 26 is deleted.

The warning in footnote 26that valuation of collateralmay change over the life ofthe loan, would warrant anentire paragraph in the maintext, that an ECL modelrequires continuedmonitoring of collateralvalues.

37 Para A7 The last sentence places aburden on banks to demonstratethat there has been nosignificant increase in credit risksince initial recognition. Inpractice, it is very difficult, if notimpossible, to demonstrate anegative and banks can onlywork with the evidence theyhave. It can also be read toconflict with the requirement asset out in IFRS 9.

We suggest deleting thesentence.

38 Para A8 While high-risk exposures arelikely to more rapidly decline incredit quality because of greatervolatility, the IFRS 9 ECL model

Clarify that lifetime ECLallowances are only requiredwhen there has been asignificant increase in credit

22

only requires allowances to bemeasured on a lifetime ECLbasis if there has been asignificant increase in credit risksince initial recognition. It isclear in the standard that, if theexposure is initially high risk,then a significant deteriorationwould involve a greater increasein the absolute PD than if theexposure was initially lower risk.

In addition, there is guidance inthis paragraph that goes beyondthe scope of the appendix,including the requirement todocument the rationale for high-risk lending along with thegovernance processes, and tohave sound commensurateunderwriting practices.

risk and realign the wordingin the paragraph with IFRS 9.

We suggest that theCommittee should delete thewording in this paragraphthat does not relate to theassessment andmeasurement of expectedcredit losses under IFRS 9.

39 Para A9 The wording ‘taking account ofthe migration of credit risk’ isconfusing since it implies thatthe 12-month ECL requirementshould anticipate ‘migrations’ toa lifetime ECL measure, whichwould not be consistent with thestandard. Further, onlyexposures can ‘migrate’, creditrisk cannot.

Presumably what thisparagraph is trying to say isthat if the credit risk isconsidered to have increased,then this should be reflectedin the 12-month ECLmeasurement before theincrease is sufficientlysignificant that the allowanceis measured on a lifetimebasis. This could be achievedby deleting ‘taking intoaccount the migration ofcredit risk’.

40 Para A15 The rationale cited as a ‘belief’of the Committee is a fact, evenif, in practice, pricing will takeinto account other factors aswell. In particular, there may bea significant increase in theprobability of default on a fullysecured loan with no effect on

We suggest deletion of ‘TheCommittee believes that …’.

23

pricing, or a change in pricingdue to competitive pressureswithout any significant changein credit risk.

Also, because loan pricing takesaccount of many factors otherthan credit risk, it is likely thatthere may be reasonably widebands of credit risk that wouldbe priced similarly. Hence, it isprobably not true that ‘any post-origination increase in credit riskis unlikely to be fullycompensated by the interestrate charged’ (emphasis added).This is an important point sincethe notion of ‘significantincrease’ in the standard doesnot imply any increase, and thisissue is reflected in a number ofparagraphs in the appendix.

Possible wording might be toturn this around, to say, ‘Ifthe post-origination creditrisk has increased to such anextent that it is unlikely thatit is compensated by theinterest rate charged, it islikely that there has been asignificant increase in creditrisk’. We also recommenddeleting the footnotereference to the Snapshot,first, because it is not neededand second, because thisdocument has noauthoritative status.

41 Para A17 This paragraph states that’Ensuring that the approach isconsistent across entities withina group is important’. Yet, thecomplexity of lending activityand the range of reasonable andsupportable information that isavailable may differ from oneentity to another, so that theremay well need to be differencesin approach. Similarly, theeconomic assumptions shouldbe appropriate to the differentjurisdictions in which a bankoperates.

We believe that the intentionhere is that there needs to beconsistent interpretation ofthe requirements of thestandard, a consistent overallmethodology, consistentapproaches to forecastingeconomic conditions andconsistent governanceprocesses.

42 Para A22 It is not true that ‘delinquencydata [is] seldom appropriate inthe implementation of an ECLapproach by banks’. It is justthat, at least for the moresophisticated lenders, it will only

We recommend that thissentence is amended to add‘on their own’ between‘seldom’ and ‘be appropriate’.

24

ever form part of theinformation that is used andwould be inappropriate on itsown.

43 Para A23 The word ‘objective’ in ‘objectivelevel of credit risk’ is notcorrect– it is the nature of creditrisk that it is never objective.

It is not clear what the use ofthis word is supposed tomean in his context. Could‘objective’ be deleted?

44 Para A27 This paragraph is unclear. Whyhas the Committee decided torequire that ‘particularattention’ be paid to only a fewof the indicators set out inparagraph B5.5.17 of thestandard? Is it the Committee’sintent that these should be givenhigher priority than the ‘may berelevant’ wording used in thestandard? In which case, is anydowngrade considered to be asignificant increase in creditrisk? And does ‘deterioration ofrelevant factors’ insubparagraph e) effectivelymean ‘anything else’, in whichcase, the words ‘particularattention’ have no practicalmeaning?

Also, footnote 33 does notrecognise that elements of loanpricing include profit margin,risk appetite and strategy. Forinstance, a bank may price aparticular type of loan morehighly not because the risk hasincreased, but because the bankis no longer seeking to expandits business in that direction. Tomake it a rebuttablepresumption that any increase inspread is due to an increase incredit risk would appear unduly

We would recommenddeleting this paragraph, orelse amending it so as tomake the Committee’sintentions clear.

We recommend that thefootnote is deleted.

25

conservative. The presumptionwould also be difficult to rebut.

Further, the rebuttablepresumption refers to ‘anyincrease in credit spread’, whichwould be impossible to apply ifthe bank cannot distinguishincreases in credit spread fromchanges in gross margin due toother factors. Hence, as worded,the test is unworkable.

45 Para A29 The statement that ‘smallchanges in credit quality can beassociated with a large increasein the probability of default’ iscontradictory – we question howcredit quality would be defined ifthis statement were true.

The wording later in theparagraph that ‘it is necessary tolook beyond how many“notches” a rating downgradeentails because the change inPD for a one notch-movement isnot linear’… reads oddly, since(as already stated in thisparagraph) the IFRS 9 ECLmodel is based on relative ratherthan absolute increases in creditrisk, and so a non-linergradation system would seem toprovide what is needed. Therewould only be a problem if a onenotch movement was based on alinear and absolute change inPD.

Also in paragraph A29, a ‘notch’can be read to refer to thesmallest gradation used by thecredit rating agencies, in whichcase it is unclear what would be

What is possibly intendedhere is that some grades orrating categories can span alarge range of PDs so thatthere can be a large increasein PD within a single grade.Possible wording might be, ‘asmall change in the gradingcan be associated with alarge increase in theprobability of default.’

If this was the Committee’sintended meaning, werecommend that the secondhalf of the paragraph startingwith ‘It is also necessary’ isreplaced by ‘Depending onhow finely graded is a bank’sexposure grading system, it ispossible that a significantincrease in credit risk mightoccur before an exposure isdowngraded.’ Alternatively, ifthe Committee wasconcerned that loan gradingsmay not reflect all theinformation required to makean individual assessment, thewording would have to becompletely different. As wehave already noted in the

26

meant by, ‘it is possible that asignificant increase in credit riskcould occur before lendingexposures experience even aone-notch downgrade’. In whatcircumstances would a declinesmaller than one notch besignificant?

Or is the Committee trying tomake the point that a loangrading system may not provideall the information necessary toassess whether there has been asignificant increase in creditrisk, i.e., where this will need tobe assessed on a top downportfolio basis?

cover letter, the guidancewould be much clearer ifthose sections appropriate toloans assessed individuallywere separated from thosethat relate to a collectiveassessment.

In any event, the term ‘anotch’ should be avoided, orelse defined.

46 Para A33 Wouldn’t the ‘differences in theseniority of individual exposures’be relevant for the loss givendefault rather than for theprobability of default? Ingeneral, an obligor is likely todefault on all of its obligations atthe same time, particularlygiven the way that defaultevents are often defined.

We suggest deletion of ‘in theseniority of individualexposures’

47 Para 44and A34

Generally, the word‘homogenous’ is too strongwhen considering exposuresmanaged on a portfolio basis.The best that can probably beachieved is that a portfolio willbe ‘reasonably homogenous’.

Shouldn’t the wording justrefer to ’shared riskcharacteristics’ as used in thestandard? Hence, remainhomogenous’ in paragraph A34 would be written as’continue to share riskcharacteristics’?

48 Para A35 See item 23 concerningparagraph 47.

49 Para A36 Strictly, IE39 should not bedescribed as a paragraph ofIFRS 9 – the IEs are not part of

We suggest this is rewrittenas ’Consistent with paragraphIE 39 of the Implementation

27

the standard and, thus, do nothave the same authoritativestatus.

Guidance to IFRS 9’.

50 Paras A 31and A37

The wording of paragraph A37can be improved. It is obviousthat the ‘significant’ notion inIFRS 9 has nothing to do withthe notion of significance instatistics. But the next clause,‘meaning that the assessmentshould not be based solely onquantitative analysis’ is a non-sequitur. The point is not aboutqualitative versus quantitativeanalysis, but that ‘significant’has nothing to do with statisticalsignificance.

Also paragraph A31 againseems to mix statisticalsignificance and the IFRS 9significant increase criterion,and the point it makes would beclearer without this potentialsource of confusion.

We recommend thatparagraph A37 is simplifiedto say, ‘Significant’ shouldnot be equated withstatistical significance’.

Meanwhile, in paragraph A31we suggest the deletion of‘the deterioration in aparticular factor isstatistically very small(judged in relation to pastdata on it), or’

51 Para A49 The final sentence states that‘This will potentially requirecostly upfront investments innew systems and processes butthe Committee considers thatthe long-term benefit of a high-quality implementation faroutweighs the associated costs,which should therefore not beconsidered undue.’

The Committee mightconsider amending the finalsentence to ‘This willpotentially require costlyupfront investments in newsystems and processes andacquisition or calculation ofhistorical loss information,but the Committee ….’

52 Para A57 Paragraphs A53 to A58 seemunnecessary if the target banksare not expected to use the lowrisk simplification. Presumablythis guidance is not intended tobe applied to securities, as theyare outside the scope of this

Consider removingparagraphs A53 to A58.

28

Guidance.