Академический Документы
Профессиональный Документы
Культура Документы
maknowledgeservices.com
October 2016
Contents
1.
Background ............................................................................................................................................. 3
2.
3.
Scope ........................................................................................................................................................ 5
4.
4.2.
5.
6.
4.1.2.
4.1.3.
Challenges .......................................................................................................................... 7
4.1.4.
4.2.2.
4.2.3.
Challenges ........................................................................................................................ 12
4.2.4.
5.2.
5.3.
6.2.
Author
Veer Bahadur Singh
vee.singh@moodys.com
2
1. Background
The International Accounting Standards Board (IASB) introduced International Financial Reporting Standards 9 (IFRS 9),
Financial Instruments in July 2014, replacing International Accounting Standard 39 (IAS 39) Financial Instruments:
Recognition and Measurement. In 2005, the IASB and the Financial Accounting Standards Board (FASB), which sets US
national standards, had begun working toward a long-term objective of improving and simplifying the reporting for
financial instruments, upon requests from users of financial statements and other interested parties. These interested
parties raised concerns that the requirements in the IAS 39 were difficult to apply and interpret. They requested the IASB
to develop a new Standard for the reporting of financial instruments that was
principle-based and less complex.
We believe the
implementation of the
new ECL model will pose
certain practical challenges
in terms of system/
information requirements
and cost of implementation
The final version of the standard combines the classification and measurement,
impairment, and hedge accounting phases of the IASBs project to replace
IAS 39. IAS 39 had many different classification categories and associated
impairment models based on the incurred loss approach.
On the basis of the asset classifications, IFRS 9 introduced a single forwardlooking expected credit loss (ECL) model, replacing the multiple impairment
models in IAS 39, which would be applied to all financial instruments that
are subject to impairment accounting, thereby removing a major source of
complexity in IAS 39.
This ECL model will likely result in the timely recognition of loan losses, addressing the issue of delayed recognition of
credit losses in the existing accounting standard, which was identified during the recent financial crisis.
We believe the implementation of the ECL model will likely pose certain practical challenges in terms of system/information
requirements and implementation cost(s). In this paper, we cover the new impairment model in detail, focusing on
the practical challenges financial institutions would face and how existing available information and resources can be
efficiently utilized. A few solutions to address the challenges are discussed.
Timeline for Implementation
The effective date for the implementation of IFRS 9 (published in July 2014) for firms is January 1, 2018 or earlier.
Impairment Recognition
12-month expected credit losses
Interest Revenue
Effective Interest on gross
carrying amount
Stage 2:
Under-Performing (Assets with
significant increase in credit risk since
initial recognition)
---------------------------------------------
---------------------------------------------
Stage 1:
Stage 3:
Non-Performing (Credit-impaired
financial assets)
*There is a separate guidance for purchased or originated credit-impaired financial assets (See IFRS 9, Paragraph 5.5.13)
As soon as a financial instrument is purchased or originated, 12-month expected credit losses are recognized in profit or
loss. Moreover, a loss allowance is established. This serves as a proxy for initial expectations of credit losses. For all other
financial instruments, e.g. already on books, the rules below are followed.
1. It is not applicable for purchased or originated credit-impaired financial assets (see IFRS 9, Paragraph 5.5.13)
2. See IFRS 9, Appendix A, Defined terms
Note:
The ECL model takes into account the relative assessment of credit risk while moving an instrument from 12-month
ECL to lifetime ECL. Therefore, two loans with the same credit risk on the reporting date may be allocated to different
stages, depending on the credit risk at the time of initial recognition. On a similar note, different loans from the same
counterparty could be categorized into different stages, depending on the credit risk each loan had at the time of
origination. For financial institutions to be able to address these intricacies, increased level of sophistication would be
required in data storage and reporting.
3. Scope
The IFRS 9 new impairment model is applied to:
Debt instruments (such as bank deposits, loans, debt securities, and trade receivables) measured at amortized cost
or FVTOCI
Loan commitments and financial guarantees not measured at fair value through profit or loss (FVTPL)
Lease receivables
Contract assets
4. Model Implementation
The two major components of the IFRS 9 ECL model are
measurement of expected credit losses and assessment
of a significant increase in credit risk. In this section,
we discuss the requirements, information to be used,
challenges, and solutions to these challenges.
Current
Conditions
Historical
Information
Forward looking
Information
ECL
A Probability
Weighted
Outcome + Time
Value of Money
4.1.3. Challenges
PIT PD
TTC PD
2.5%
2.0%
1.5%
1.0%
0.5%
0.0%
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
2011
2012
2013
2014
2015
LGD and EAD models from the Basel framework can be used directly, but post-adjustments, e.g. in the case of LGD,
downturn adjustments need to eliminated and forward-looking information needs to be incorporated
to the extent possible.
0.40
0.2
0.1
0.30
0.25
Firm
remains
Baa
0.20
0.15
Aa
0.10
Aaa
-2
Default Area
0.05
0.0
Default Area
-4
Bad Days
(Z- -2)
0.35
Probability
0.3
Ba
----------------------------------------------
0.45
0.00
-5
-4
-3 -2 -1
0
1
2
Credit-change indicator
Good Days
(Z-2)
The Z-index is calculated such that once applied to each state in a periods transition matrix, it would minimize the
difference between the probabilities of historical PIT transitions and TTC transitions (see Figure 5).
Figure 5: Calculating Z-Index (Illustration)
Corresponding TTC matrix using inverse
probability function for a standard normal
distribution
TTC matrix
X1
X2
X1
X2
X1
X2
X1 91% 8% 1%
X1
(inf,-1.35)
(-1.35,-2.38) (-2.38,-inf)
X1
PIT11
PIT12
PIT13
X2
X2
(inf,-1.56)
(1.56,-1.63)
X2
PIT21
PIT22
PIT23
7% 87% 6%
(-1.63,-inf)
Z index
Using the above approach, Z-index for each year is calculated, and a regression model is used to estimate the
relationship between Z-index and macroeconomic variables. By using macroeconomic forecasts, Z-index is
estimated and applied to the TTC transition matrix to obtain the PIT transition matrix, which in turn gives PIT PDs for
each grade.
Figure 6: Estimate and Forecast of Z-Index (Illustration)
1.5
1
0.5
25
20
23
24
20
22
20
21
20
20
20
19
20
18
20
17
20
16
20
20
14
15
20
13
20
12
20
11
20
10
20
09
20
08
20
07
20
06
20
05
20
04
20
03
20
02
20
01
20
20
00
01
20
99
20
19
97
98
19
96
19
19
94
95
19
93
19
92
19
19
19
91
0
-0.5
-1
-1.5
=
=1
In this method, rating actions are modeled, e.g. B1 to B2 and B1 to B3 are grouped into one rating action B1
downgraded. Once the probabilities of rating actions are estimated, they are further distributed into more granular
rating transitions based on a static conditional empirical probability distribution. For example, the estimated probability
of B1 downgraded will be distributed into B1 to B2 and B1 to B3 using the historical average transition probability
distribution of B1 to B2 and B1 to B3. Figure 7 shows illustrative rating actions to be modeled.
Figure 7: Rating Actions
Starng
Rang
B1
B2
B3
B4
C
WR
DEF
B1
5.B2_UP
10.B3_UP
15.B4_UP
19.W_UP
10
B2
B3
1.B1_DN
1.B1_DN
6.B2_DN
10.B3_UP
15.B4_UP
19.W_UP
15.B4_UP
19.W_UP
Ending Rang
B4
1.B1_DN
6.B2_DN
11.B3_DN
19.W_UP
WR
DEF
2.B1_C
7.B2_C
12.B3_C
16.B4_C
3.B1_WR
8.B2_WR
13.B3_WR
17.B4_WR
20.W_WR
4.B1_DEF
9.B2_DEF
14.B3_DEF
18.B4_DEF
21.W_DEF
A multinomial logistic regression is chosen for modeling. Such a regression has many advantages compared with
other methods. The estimates of standard errors and p values are straightforward. Moreover, a multinomial logistic
regression requires fewer assumptions than other models, as it does not assume normality, linearity, or homogeneity of
variance for the independent variables.
Figure 8 shows rating paths for an obligor in rating A. For multi-horizon forecasting, we would need to compute all
possible rating paths. The final transition matrix would be computed by using the average of the probabilities of all
rating paths.
Figure 8: Rating Paths
A_Up (10%)
Aa
A 100%
A (81%)
A_Dn (8%)
Baa
D
Def (1%)
11
12-Month Point-in-Time PD
PDs NOT comparable
7.0%
1.4%
6.0%
1.2%
5.0%
1.0%
4.0%
0.8%
3.0%
Comparable Points
0.6%
2.0%
0.4%
As of Origination
0.2%
0.0%
t=0
Lifetime PD
8.0%
As of Reporting Date
6
10
As of Origination
1.0%
0.0%
t=0
As of Reporting Date
6
10
For some loans (e.g., bullet loans or CRE loans with a maturity of say 10 years and amortization term of say 30 years or
interest-only loans, which can be refinanced at maturity), it is quite clear that 12-month PD is not sufficient to assess a
significant increase in default risk. But for usual C&I term loans, it may appear that 12-month PD is a good alternative;
however, even in these cases, sometimes 12-month PD may not be a good method to ascertain risk. For example, in the
left chart in Figure 9, 12-month PDs appear to be almost the same, i.e. default risk has not increased significantly, but the
forecast as of the reporting date suggests a downturn after a year or so. Hence, when we compare lifetime PD, we observe
a significant increase in default risk since origination (as shown in the diagram on the right). This suggests 12-month
PD may not be sufficient in all conditions even though financial instrument characteristics suggest otherwise.
For the same counterparty, multiple loans may be treated differently, depending on the credit risk at the time of
origination of the loans. This, again, requires enhanced risk management systems.
12
13
Key Requirements
Basel
IFRS 9
Default definition
Default window
12 month
No such requirement
PD Type
PIT (Forward-looking)
LGD Type
No downturn adjustment
EAD
12 month EAD
EL
PD*LGD*EAD
Key Requirements
IRB Banks
Non-IRB Banks
12-Months PD Model
Lifetime PD Models
Need to develop
Need to develop
LGD Models
Need to develop
EAD Models
Need to develop
Models forportfoliosnot
covered under IRB
Need to develop
Need to develop
Modelling Challenges
14
15
maknowledgeservices.com
2016 Moodys Corporation, Moodys Investors Service, Inc., Moodys Analytics, Inc. and/or their licensors and affiliates (collectively, MOODYS). All rights reserved.
SP23703/101215/IND-103