mixture of the bottom up and top down approaches, as described in Examples 47.16 and
47.17 above. Macroeconomic indicators are assessed, as in the top down approach, but the
effect is determined by assessing the effect on particular exposures. One possible method is
to determine the expected migration of loans through a bank’s risk classification system, by
recalibrating the probabilities of default based on forward-looking data. This could be used
to forecast how many additional loans will get downgraded as well as the associated ECLs.
Another is to focus on more vulnerable categories of lending, such as interest-only
mortgages, secured loans with high loan-to-value ratios, or property development loans,
and assess how these might respond to the economic outlook, The more information about
customers that a lender possesses, the more this might look like the illustrated bottom up
approach. It is likely that banks will use different approaches for different portfolios,
depending on how they are managed and what data is available.
All the examples in the illustrative examples simplify the fact pattern to focus on just one
driver of credit losses, whereas in reality there will be many, and it may not be possible to
find a historical precedent for the combination of economic indicators that may now be
present. Further, to delve into the past to predict the future requires a level of data that
banks may lack. The example in the standard bases the percentage on historical
experience, but it is more than 20 years since most developed countries last saw a 200 basis
points rise in interest rates, and products and lending practices were then very different, as
was the level of interest rates before they began to rise and the extent of the increase.
Hence, the past may not be a reliable guide to the future. In practice, banks will need to
determine the main macroeconomic variables that correlate with credit losses and focus
on modelling these key drivers of loss. The banks can make use of work that has already
been carried out for stress testing. Also, it should be stressed that banks will generally use
3816 Chapter 47
one single model to estimate forward-looking PDs for both for the assessment of significant
increases in credit risk and the measurement of ECLs (see 5.9.3 above).
The example of an anticipated increase in interest rates is very topical, given that rates
in many countries are expected to rise in future from the all-time low levels that have
been experienced since the financial crisis. This gives rise to an observation that is
relevant to any ECL model: banks and (hopefully) borrowers have presumably known
that new variable loans made since the crisis would likely increase in rate as the
economy improves. If the increase was anticipated at the time of origination,
expectation of a rise in interest rate should not be viewed as a significant deterioration
in credit risk. Yet, there is a concern that rising rates will bring difficulty for many
borrowers who have over stretched themselves, implying that the inevitable rise was
not fully factored into lending decisions. With any forward-looking approach it is
necessary to understand what risks were already taken into account when loans are first
made, to assess whether there has been a significant increase in risk.
6.6
Determining the credit risk at initial recognition of an identical
group of financial assets
In practice, entities may hold a portfolio of debt securities that are identical and cannot
be distinguished individually (e.g. all securities have the same international securities
identification number (ISIN)) and over the lifetime of the portfolio, entities may acquire
additional securities or sell some of those previously acquired. In such instances, entities
have to determine the credit risk at initial recognition of those securities that remain in
the homogeneous portfolio at the reporting date.
IFRS 9 contains no specific guidance on how to calculate the cost of financial assets for
derecognition purposes when they are part of a homogenous portfolio. Under IAS 39,
which is also silent on this topic, entities choose between the following cost allocation
methods for available-for-sale securities: the average cost method, the first-in-first-out
(FIFO) method or the specific identification method. Specific identification can be applied
if the entity is able to identify the specific items sold and their costs. For example, a
specific security may be identified as sold by linking the date, amount and cost of
securities bought with the sale transaction, provided that there is no other evidence
suggesting that the actual security sold was not the one identified under this method.
For IFRS 9, the question arises whether entities can continue to apply one of the above
methods for debt instruments, not only for determining the cost of the security at
derecognition but also for determining their initial credit risk. We believe that:
• the method used for recognising and measuring impairment losses should normally
be the same as that used for determining the cost allocation method on derecognition;
• a FIFO approach or a specific identification method as described above constitute
acceptable accounting policy choices to be applied consistently; however
• it would not normally be appropriate to use the weighted-average method to
determine the credit risk at initial recognition, as averaging the different levels of
initial credit risk of debt securities purchased at different dates would result in an
identical initial credit risk for each item. It therefore, would create bias when
assessing whether the credit risk of debt securities has increased significantly.
Financial instruments: Impairment 3817
6.7
Multiple scenarios for the assessment of significant increases in
credit risk
At its December 2015 meeting the Impairment Transition resource Group (ITG)
discussed not only the need to consider multiple scenarios for measurement of ECLs
(see 5.6 above), but also for the purposes of assessing whether exposures should be
measured on a lifetime loss basis.
Similar to the measurement of ECLs, the ITG members noted that where there is a non-
linear relationship between the different forward looking scenarios and the associated risks
of default, using a single scenario would not meet the objectives of the standard.
Consequently, in such cases, an entity would need to consider more than one forward
looking scenario. Further, there should be consistency, to the extent relevant, between the
information used to measure ECLs and that used to assess significant increases in credit
risk. An example of when the information might not be relevant is the value of collateral.
It may be necessary to calculate the effect of multiple scenarios to value collateral to
measure ECLs, but this information may not be relevant to assessing significant changes in
credit risk unless the value has an effect on the probability of default occurring.20
As with the measurement of ECLs, the ITG members noted that IFRS 9 does not
prescribe particular methods of assessing for significant increases in credit risk.
Consequently, various methods could be applied, depending on facts and circumstances
and these may includ
e both quantitative and qualitative approaches. An entity should
not restrict itself by considering only quantitative approaches when considering how to
incorporate multiple forward-looking scenarios. Whichever approach is taken, it should
be consistent with IFRS 9, considering reasonable and supportable information that is
available without undue cost and effort. Once again, this is an area of judgement and so
appropriate disclosures would need to be provided to comply with the requirements of
IFRS 7 (see 15 below and Chapter 50 at 5.3).
A further issue was raised at the ITG meeting, which was not referred to in the minutes
but was addressed in the 25 July 2016 IASB webcast. If a number of scenarios are applied
to an individual asset, in some of which there is no significant increase in credit risk and
in others there is, is it possible that it could be measured partly based on 12 month losses
and partly on lifetime losses? It was not the intention of the IASB that an asset should
be regarded as being in more than one stage at the same time. For staging as well as for
measurement, IFRS 9 applies to the unit of account which is the individual financial
instrument. The financial asset cannot be considered to have partly significantly
deteriorated and partly not. Hence, for instance, if the staging assessment is based on a
mechanistic approach which considers the change in the lifetime probability of default,
the entity should use the multiple scenario probability-weighted lifetime probability of
default to assess whether there has been a significant increase in credit risk. The asset
should then be measured using the weighted 12-month probability of default if it is
considered to be in stage 1, or the weighted lifetime probability of default if it is
considered to be in stage 2.
However, as described in 6.5.3 above, the webcast also noted that, for a collectively
assessed portfolio of assets, a proportion of the portfolio only may be deemed to have
significantly deteriorated while the rest of the portfolio has not, due to differences in
sensitivities of credit risk to a change in a particular parameter.
3818 Chapter 47
The IASB also illustrated how multiple scenarios can be reflected in a non-PD-based
approach, using the example of a scorecard system. If the entity determines that there
is non-linearity in the effect of the scenarios on the credit risk of the customers, one
possibility is to look at the scorecard inputs and to determine which of these inputs have
a non-linear relationship with the macroeconomic parameters. The entity then adjusts
the scorecard, for example, using a scaling factor to reflect the impact of non-linearity,
assesses whether there has been a significant increase in credit risk and measures ECL
on the basis of the adjusted scorecard.
The approach set out in this discussion is broadly the same as ‘the top down’ approach
to collective assessments illustrated by Example 47.17 in 6.5 above.
It is important to note that the ITG did not state that it is always necessary to use
multiple scenarios and probability-weighted lifetime probabilities of default to assess
significant increases in credit risk.
What it did state is that:
(a) it is necessary to consider more than one scenario if there is non-linearity in the
possible distribution of losses;
(b) qualitative approaches may be included as well as quantitative ones, so that, for
instance, it might be possible to take account of non-linearities by scaling the
output from score cards; and
(c) the assessment should be based on reasonable and supportable information that is
available without undue cost or effort (see 5.9.1 above).
Nevertheless, the ITG did state that there should be consistency, to the extent relevant,
between the forward-looking information used for measurement and for the assessment
of significant increases in credit risk. There would not always be a direct mapping of the
relevant information, because in some cases information might have an impact on the
measurement of ECLs but not on the assessment of significant increases in credit risk
(and vice versa). Also, various methods of assessing for significant increases in credit risk
could be applied, depending on the particular facts and circumstances, and an entity
should not restrict itself by considering only quantitative approaches when considering
how to incorporate multiple forward-looking scenarios.
In the July 2016 webcast, the IASB also stressed the importance of adequate disclosures.
Because there is no one right approach and because this area involves a high level of
judgement, disclosures are very important to enable users of financial statements to
understand how entities’ credit risk is affected by forward-looking scenarios and how
they have affected the application of the ECL model. It would also be useful to disclose
if relevant forward looking information has not been reflected in the assessment of
significant deterioration on the basis that it is not reasonable and supportable.
In practice, many banks that use multiple scenarios of lifetime probabilities of default to
measure assets in stage 2, intend to use them also for assessing if there has been a
significant increase in credit risk. Moreover, as with measurement, banks will need to
consider regulators’ expectations (see 7.1 below).
Financial instruments: Impairment 3819
7
OTHER MATTERS AND ISSUES IN RELATION TO THE
EXPECTED CREDIT LOSS CALCULATIONS
This section discusses other matters and issues that are relevant to applying the IFRS 9
impairment requirements.
7.1
Basel guidance on accounting for expected credit losses
The Basel Committee published the final version of its Guidance on Credit Risk and
Accounting for Expected Credit Losses (sometimes referred to as ‘G-CRAECL’, but in
this publication, as ‘the Basel guidance’ or just ‘the guidance’) in December 2015 (see 1.6
above). The guidance deals with lending exposures, and not debt securities, and does
not address the consequent capital requirements.
The guidance was originally drafted for internationally active banks and more
sophisticated banks in the business of lending. The final version does not limit its scope
but allows less complex banks to apply, ‘a proportionate approach’ that is
commensurate with the size, nature and complexity of their lending exposures. It also
extends this notion to individual portfolios of more complex banks. It follows that
determining what is proportionate will be a key judgement to be made, which is likely
to be guided in some jurisdictions by banking regulators. The guidance issued in
June 2016 by the GPPC (see 7.2 below) will also be relevant in making this
determination. The final version of the guidance acknowledges that due consideration
may also be given to materiality.
The main section of the Basel Committee’s guidance is intended to be applicable in all
jurisdictions (i.e. for banks reporting under US GAAP as well as for banks reporting
under IFRS) and contains 11 supervisory principles. The guidance is supplemented by
an appendix that outlines additional supervisory requirements specific to jurisdictions
applying the IFRS 9 ECL mod
el.
It is important to stress that the guidance is not intended to conflict with IFRS 9 (and,
indeed, this has been confirmed by the IASB), but it goes further than IFRS 9 and, in
particular, removes some of the simplifications that are available in the standard. It also
insists that any approximation to what would be regarded as an ‘ideal’ implementation
of ECL accounting should be designed and implemented so as to avoid ‘bias’. The term
‘avoidance of bias’ is used several times in the guidance and we understand it to have its
normal accounting meaning of neutrality. Hence, for instance, if a bank were ever
dependent on past-due information to assess whether an exposure should be measured
on a lifetime ECL basis, it is guided to ‘pay particular attention to their measurement of
the 12-month allowance to ensure that ECLs are appropriately captured in accordance
with the measurement objective of IFRS 9.’21
Perhaps one of the most significant pieces of guidance provided by the Basel Committee
relates to the important requirement in IFRS 9 that ECLs should be measured using
‘reasonable and supportable information’. The Committee accepts that in certain
circumstances, information relevant to the assessment and measurement of credit risk may
not be reasonable and supportable and should therefore be excluded from the ECL
assessment and measurement process. But, given that credit risk management is a core
competence of banks, ‘these circumstances would be exceptional in nature’.22 This attitude
3820 Chapter 47
pervades the guidance. It also states that management is expected ‘to apply its credit
judgement to consider future scenarios’ and ‘[t]he Committee does not view the unbiased
consideration of forward looking information as speculative’.23 The guidance, therefore,
establishes a high hurdle for when it is not possible for an internationally active bank to
estimate the effects of forward looking information. It is possible that banking regulators
would expect banks to make an estimate of the effects of events with an uncertain binary
outcome that is highly significant, such as the result of a referendum as discussed by the
ITG in September 2015 (see 5.9.5 above).
A connected piece of the guidance relates to another important principle in IFRS 9, that
International GAAP® 2019: Generally Accepted Accounting Practice under International Financial Reporting Standards Page 755
