with those scenarios that are more benign. To use statistical terminology, the
distribution is skewed. Depending on how it is calculated, a single scenario gives the
mode of this distribution (i.e. the most likely outcome) or the median (the central
forecast). In contrast, the standard requires the use of the mean (i.e. a probability-
Financial instruments: Impairment 3763
weighted estimation). A possible distribution of the losses in the portfolio consistent
with the above example is shown in Figure 47.4 below.
Figure 47.4 Distribution of losses
Likelihood of occurrence
Mode (£70)
Mean (£92)
£0
£100
£200
£300
£400
At the ITG meeting, it was noted that there are a number of possible approaches that
might be used to incorporate multiple economic approaches. IFRS 9 does not prescribe
any particular method of measuring ECLs and the measurement should reflect an entity’s
own view. What the standard does require is that the expected losses must reflect:
(a) an unbiased and probability-weighted amount that is determined using a range of
possible outcomes; and
(b) reasonable and supportable information that is available without undue cost or
effort at the reporting date.
With respect to reasonable and supportable information, ITG members made the
following observations:
(a) although IFRS 9 does not specifically require an entity to consider external
information, an entity should consider information from a variety of sources in
order to ensure that the information used is reasonable and supportable;
(b) the information considered could vary depending on the facts and circumstances
including the level of sophistication of the entity, geographical region and the
particular features of the portfolio; and
(c) while entities are not expected to consider every possible scenario, the scenarios
considered should reflect a representative sample of possible outcomes.16
ITG members recognised that materiality considerations would need to be taken
into account.
In an IASB webcast on 25 July 2016, it was noted that, having considered:
(a) whether the effect of non-linearity is material;
(b) whether the entity has a reasonable and supportable basis for this multiple scenario
analysis; and
(c) whether the application is possible without undue cost or effort;
3764 Chapter 47
A conclusion may sometimes be reached that it is not necessary to actually use multiple
scenarios to apply the impairment requirements in IFRS 9. However, multiple scenarios
must always be considered.
At the December 2015 ITG meeting, the ITG also noted that consideration should be
given to the consistency of forward-looking information used for the measurement of
ECLs and for other purposes within the organisation, such as budgeting and forecasting.
ITG members acknowledged that there might be differences, but observed that these
should be understood and explainable.
ITG members also observed that the incorporation of forward-looking scenarios will
require judgement. Consequently, they emphasised the importance of the IFRS 7
disclosure requirements relating to how forward-looking information has been
incorporated into the determination of ECLs (see Chapter 50 at 5.3).17
Since December 2015, banks have given significant attention to how multiple economic
scenarios can be incorporated into ECL calculations. We have seen three main
approaches being explored, as follows:
(a) Probability weighted scenarios – this is similar to the method discussed at the
ITG meeting in December 2015 and illustrated in Example 47.6 above. It
involves establishing a number of scenarios (typically three scenarios but we
have seen varying numbers, generally between two and four), estimating the
losses that would arise in those scenarios and allocating a weighting to each
scenario. Unlike Example 47.6 above, these do not normally model economic
variables such as unemployment rates in isolation – to do so, would also require
complex modelling of the correlations between those variables. Instead, each
scenario is normally a coherent combination of economic variables. For
example, a scenario relevant to mortgage loans might include assumptions about
unemployment, interest rates and house prices. This approach is transparent,
but it may be difficult to assign the weightings to each scenario, requiring
judgement as well as experience of the past. While selecting scenarios and
respective weights, we expect banks to take into consideration the entire
distribution of macroeconomic scenarios and select points (i.e. scenarios) from
that distribution, with their respective weights representing the portion of the
distribution represented by the scenario. We would also expect that the mean
of the selected scenarios and weights is similar to that of the entire distribution.
(b) The second approach is to calculate ECLs based on a central forward-looking
scenario and to adjust the outcome where necessary by a factor to reflect the non-
linearity of the loss distribution. In practice it may be that a method similar to (a)
above will need to be used in order to calculate this factor – so that it is not a very
different approach. However, some banks view the merits of this approach as
being less mechanistic and allowing more room for judgement.
(c) Monte Carlo simulation – this method seeks to calculate the expected losses
associated with the entire distribution of possible scenarios around the bank’s
central economic forecast. It has the advantage that it does not require the bank to
formulate specific scenarios or assign weightings to them, but the simulation is
dependent on assumptions that may not be transparent to either users or preparers,
Financial instruments: Impairment 3765
so that this solution can seem a ‘black box’. It is also very demanding as to the
volume of data that has to be manipulated and it is not how most banks manage
credit risk today. This method is quite rarely applied in practice.
The effect of multiple scenarios will affect not just the probability of default but also the
losses given default. For instance, for property-based lending it will be necessary to
forecast the value of collateral associated with each economic scenario that is modelled.
A consequence of this is that there may be a need to record an ECL allowance for an
asset that, based on the central forecast of future collateral values, is fully collateralised.
(Also, as a result, the loss allowance for a stage 3 asset may be higher than for an impaired
asset under IAS 39).
The use of multiple scenarios may also have an effect on the estimated EAD.
A number of other observations can be made about the use of multiple scenarios:
(a) Whichever approach is used to calculate the effect of non-linearity, it will be
necessary for banks to communicate the result of the calculation in a manner
which can be understood by readers of the financial statements. One possible
approach would be for banks to report the losses associated with the central
forecast and then, separately, the effec
t of the consideration of other scenarios.
This would allow banks to communicate the amounts they expect to lose and
would permit comparison between banks of the effect of the adjustment for non-
linearity, even if the banks use different methods to make the calculation.
(b) It would seem that the effects of non-linearity depend on the countries in which
banks operate and the economic characteristics of those countries. For instance,
the effect of alternative scenarios of interest rates and unemployment may be
greater in countries where there is more of a ‘boom and bust’ economic cycle. The
size of the effect is also dependent on origination practices and the particular
lending products – variable rate loans being more sensitive to interest rates than
fixed-rate ones, while defaults on credit cards are more affected by unemployment
rates. In some cases the issue is seen as most relevant for exposures to a particular
economic variable, a topical example being lending to companies involved in the
oil industry. In this example, banks might model a number of scenarios as to how
oil prices could evolve. A similar approach may be relevant for non-banks with
similar exposures through long term construction contracts or leasing activities.
There is also more likely to be non-linearity in the calculation of ECLs when
exposures are collateralised by assets whose values also change in response to the
economic conditions that drive the probability of default. An example is residential
mortgage loans.
(c) It should be stressed that the ITG discussion highlighted the importance of
calculating the effect of non-linearity using only reasonable and supportable
information, implying that if the information is not available then there is a limit to
what can be done. However, banks will also need to take into account their
regulators’ expectations (see 1.6 above and 7.1 below for Basel Committee
guidance).
The process of forecasting future economic conditions is discussed further in 5.9.3 below.
3766 Chapter 47
5.7
Time value of money
An entity needs to consider the time value of money when measuring ECLs, by
discounting the estimated losses to the reporting date using a rate that approximates the
EIR of the asset. [IFRS 9.5.5.17, B5.5.44]. This has two components:
• discounting recoveries to the date of default, hence ‘a credit loss arises even if the
entity expects to be paid in full but later than when contractually due’; [IFRS 9.B5.5.28]
and
• discounting losses from the date of default to the reporting date. This is needed as
the gross amortised cost of the asset is based on the contractual cash flows
discounted at the EIR, and therefore not discounting cash flows that are now not
expected to be received would overstate the loss.
It is rare that customers just fail to pay amounts when due. In most cases, default also
involves payments being paid late, while default can lead to the acceleration of payment
of amounts that are not contractually due until a later date. Therefore, modelling losses
involves modelling the timing of payments when default occurs and different patterns
of timing of recoverable cash flows, such as the time it takes to foreclose on and sell
collateral and complete bankruptcy proceedings, before the ECLs can be discounted
back to the reporting date.
Of these two components, the first is typically included by banks in their calculation of
the LGD (although not necessarily using the EIR). However the second will also need to
be calculated to comply with the standard.
The standard and its illustrative examples are silent on how the calculation should be
made. In Illustrative Example 9 the present value of the observed loss is assumed and in
Illustrative Example 8, a footnote states that, ‘because the LGD represents a percentage
of the present value of the gross carrying amount, this example does not illustrate the
time value of money’.
One approach would be to model various scenarios as to how cash is collected once the loan
has defaulted, and probability-weight the discounted cash flows of these various scenarios.
The discount rate is calculated as follows:
• for a fixed-rate financial asset, entities are required to determine or approximate
the EIR on the initial recognition of the financial asset, while for a floating-rate
financial asset, entities are required to use the current EIR; [IFRS 9.B5.5.44]
• for a purchased or originated credit-impaired financial asset (see 3.3 above),
entities are required to discount ECLs using the credit-adjusted EIR determined
on the initial recognition of the financial asset; [IFRS 9.B5.5.45]
• for a loan commitment (see 11 below), entities are required to use the EIR of the
asset that will result once the commitment is drawn down. This would give rise to
a consistent rate for a credit facility that includes both a loan (i.e. a financial asset)
and an undrawn commitment (i.e. a loan commitment). If the EIR of the resulting
asset is not determinable, then entities are required to use the current risk-free rate
(i.e. the discount rate that reflects the current market assessment of the time value
of money). This should be adjusted for risks specific to the cash flows, but only if
Financial instruments: Impairment 3767
the cash flows have not already been adjusted for these risks, in order to avoid
double counting; [IFRS 9.B5.5.47, B5.5.48]
• for financial guarantee contracts (see 11 below) entities are required to use the
current risk-free rate adjusted for risks specific to the cash flows, again to the
extent that those cash flows have not already been adjusted for the risks;
[IFRS 9.B5.5.48] and
• for lease receivables (see 10.2 below), entities are required to discount the ECLs
using the same discount rate used in the measurement of the lease receivables in
accordance with IAS 17 or IFRS 16 (when applied). [IFRS 9.B5.5.46].
LGD data available from Basel models should include a discounting factor and
sometimes this may be different from the rate required by IFRS 9. Furthermore, the
discount rate used in Basel models only covers the period between default and
subsequent recoveries. Therefore, entities will have to find ways to adjust their LGDs
to reflect the discounting effect required by the standard (i.e. based on a rate that
approximates the EIR and over the entire period from recoveries back to the reporting
date). Given the requirement to use an approximation to the EIR, entities will need to
work out how to determine a rate that is sufficiently accurate. One of the challenges is
to interpret how much flexibility is afforded by the term ‘approximation’.
At its meeting in December 2015, the ITG also discussed what was meant by the current
EIR when an entity recognises interest revenue in each period based on the actual
floating-rate applicable to that period. The ITG first noted that the definition of the EIR
in IFRS 9 was carried forward essentially unchanged from the definition within IAS 39.
Consequently, similarly to IAS 39, IFRS 9 does not specify whether an entity should use
the current interest rate at the reporting date or the projected interest rates derived from
the current yield curve as at the reporting date. There should be consistency between the
rate used to recognise interest revenue, the rate used to project future cash flows
(including cash shortfalls) and the rate used to discount those cash flows (see 5.7 above).
In relation to the guidance in paragraphs B5.5.47 and 48 on loan commitments when the
EIR on the resulting asset is not determinable and for financial guarantee contracts, we
make the following observations:
• Although it is not clear in the standard, any adjustment for the risks specific to the
cash flows would be a reduction of the risk free rate, not an increase. This would
be consistent with the approach applied to provisions in IAS 37 and as was made
clear in the staff paper presented to the Board when this treatment was discussed
in December 2013. For financial guarantee contracts, the reduction in the risk-free
discount rate will increase the present value of the obligation to pay claims to the
guarantee holder. This reflects the additional compensation that would be
demanded to take on this risky obligation, in particular to bear the risk that claims
payments will be higher than the probability-weighted expected amount.
• For loan commitments when the EIR on the resulting asset is not determinable,
this approach provides a prudent calculation of ECLs, given that it is likely that the
entity which enters into the commitment will receive a credit spread on the loan if
it is drawn down. It is in a much better position than the issuer of a financial
3768 Chapter 47
guarantee contract, who will receive no credit spread should it be required to pay
out on the guarantee.
• The idea that the rate should be adjusted only if the cash flows have not already
been adjusted for the risks may not be easy to apply in practice. This is because
the cash flows should have already been estimated with regard to any non-
linearities in the distribution of losses (see 5.6 above) and so will already have been
partly adjusted for risk. It may not be easy to calculate the necessary adjustment to
reflect a market assessment of the remaining risks.
5.8
Losses expected in the event of default
This section discusses the measurement of ECLs taking into account credit
International GAAP® 2019: Generally Accepted Accounting Practice under International Financial Reporting Standards Page 744
