Expected Credit

Loss (ECL)
Turning Theory into Action

January 2025

kpmg.com/in

KPMG. Make the Difference.

2 | Expected Credit Loss (ECL)

Table of contents
01 Introduction 03

02 Expected Credit Loss (ECL) framework 04

03 Development of different ECL components 05

04 Validation of the models 13

05 Key challenges 14

06 Key impacted areas 16

07 Conclusion 17

Glossary 18
 Expected Credit Loss (ECL) | 3

1. Introduction
Expected Credit Loss (ECL) was implemented in The paper is organised as follows. This section
different countries under IFRS9 standard in 2018. provides the introduction to the paper. The
In U.S.A. also, the standard came in effect as part of subsequent section provides an overview of the ECL
ASC 326 – Current Expected Credit Loss (CECL) in framework and key requirements for robust
2022. In India, Reserve Bank of India (RBI) has framework implementation. Section three provides
deferred the implementation of the standard for key components of ECL framework. Section four
banks, but any Non-Banking Financial Company provides an overview of key considerations in
(NBFC) which has transitioned to IndAS, the validation of different models and framework as per
requirement to compute ECL is applicable as per regulatory guidelines and industry best practices.
IndAS 109. Section five provides an overview of key challenges
that any FI faces while implementing ECL framework
Mostly all major economies are now transitioned to
and how those can be overcome. Section six
IFRS9 standard as issued by the International
provides an overview of key areas impacted due to
Accounting Standards Board (IASB). To transition to
implementation of ECL framework and how FIs can
ECL from current incurred loss method RBI has
better prepare for such impacts. Section seven
issued discussion paper in January 20231. In
provides conclusion on how Indian FIs can
continuation, RBI, on 4 October 2023, had
implement a robust
constituted a nine-member committee to
ECL framework.
recommend framework for ECL loan loss
provisioning for Indian Financial Institutions (FIs).
Final guidelines basis the recommendations of the
committee are expected to be released by RBI and
thus it is imperative for all Indian FIs to work towards
implementing ECL computation policies and
framework.

1. “Introduction of Expected Credit Loss Framework for Provisioning by

Banks” issued by the Reserve Bank of India in January 2023.
4 | Expected Credit Loss (ECL)

2. Expected Credit Loss (ECL) framework

As ECL will impact Financial Statements of the FI, Below are the key considerations for effective
it is imperative that the framework is robust which implementation of ECL framework:
adheres with applicable regulatory guidelines and
industry best practices. Once ECL is implemented,
FIs need to compute either 12-month or lifetime ECL
for the facility from the day of loan disbursement.
01 Business model and Solely Payment of
Principal and Interest (SPPI) policy

This will impact the financial statements of the FI

including profitability, capital adequacy and other
financial indicators. 02 Modelling for ECL and its components such
as stage, segmentation, PD, LGD, and EAD

Basis our experience, it is important that different

departments including underwriting, treasury,
finance, risk, IT, and operations should have good
03 Model document, policies, and SOPs

understanding of the framework and work closely

to ensure that the impact of adoption of such a 04 Validation of models
framework is not large for the bank. Also, as part of
Target Operating Model (TOM), there should be
clearly defined roles and responsibilities of all these
05 Financial reporting and disclosures

departments to ensure minimum volatility on the

statement of profit and loss due to
movement of provisions.
06 Governance and controls.

As per IFRS9, FIs can compute ECL basis different methods. Below are three approaches as provided in
standard to compute impairment losses basis underlying asset:

General Measurement Simplified approach Purchased or Originated

Model (GMM) Under this approach, stage
Credit-Impaired (POCI)
Under this approach, FIs have assessment is not required and This approach is applicable for
to assess the stage of the lifetime ECL is computed. assets which are credit
facility and accordingly Generally, this approach is impaired either at origination
compute ECL for next used for assets such as trade or purchased as credit
12-month or lifetime. ECL for receivable, and lease receivable impaired. Generally, lifetime
most of the assets such as where there is no significant ECL is computed for such
retail loans, corporate loans, financing component assets and credit adjusted
and bonds will be computed effective interest rate (CEIR) is
under this approach used to discount the ECL.

In the next section of the paper, we have elaborated on key components for the most common approach as
used for impairment calculation i.e., PD, LGD approach, which forms a part of GMM.
 Expected Credit Loss (ECL) | 5

3. Development of different ECL components

In general, different ECL frameworks (IFRS9, IndAS 109, ASC 326) are principal based and do not prescribe a
single method that is to be used for ECL computation. As per ASC 326-20, “an entity may use discounted
cash flow methods, loss-rate methods, roll-rate methods, probability-of-default methods, or methods that
utilise an aging schedule to compute allowance for credit loss.2

In our experience, most common approach to compute ECL is based on PD, LGD and EAD estimation.
Under this approach, FIs can compute ECL basis below formula:

T
ECL =  PD
t =1
t  LGDt  EADt  Dt
Where;
PD is probability of default which defines the likelihood that the borrower will default on its obligations
when they come due within specific time.
LGD is loss given default which defines the amount that the FI will lose in case a borrower defaults on its
obligations.
EAD is exposure at default which defines the exposure that will be at risk in case a borrower defaults on
its obligations. This will include principle outstanding, accrued interest and future interest that the FI is
expecting to receive during the lifetime of the contract.
D is discount rate which can be computed as Effective Interest Rate (EIR) under GMM and Simplified
approach or Credit-Adjusted Effective Interest Rate (CEIR) under POCI approach.

2. "Financial Instruments—Credit Losses (Topic 326): Measurement of Credit Losses on Financial Instruments” issued by the Financial Accounting
Standards Board (FASB) in June 2016.
6 | Expected Credit Loss (ECL) Expected Credit Loss (ECL) | 6

Below we will discuss on different key components to be considered by FIs while computing ECL using this
method:

3.1 Definition of default

FIs should have clear and consistent default In our view, FIs should assess different portfolios and
definition. We have noted that many FIs do not either finalise definition of default. Once the definition is
have defined definition or have different default finalised, FIs can move to next steps of data
definition for different purposes such as internal collection basis good and bad customers. Also, any
credit risk management, regulatory reporting, change in default definition should be applied
financial reporting, ECL computation, capital prospectively for computation purposes.
adequacy and any other internal monitoring
purposes. An FI can use guidelines as issued by RBI
as well as other regulators such as European
Banking Authority (EBA) for this purpose.
Some of the parameters that can be considered as
part of default definition are listed below:
1. Counterparty classified as NPA
2. Exposure to counterparty is restructured or
modified due to financial difficulties
3. Counterparty facing financial difficulty and is
unlikely to pay which can be assessed basis
different parameters.

3.2 Segmentation

Segmentation or pooling refers to dividing the portfolio basis homogenous risk characteristics. All facilities
should have homogenous risks within a segment/pool and heterogenous risks between any two
segments/pools. Some of the parameters that can be considered for creating segments are:

Geography such as base country or continent (India, Rest Asia, Europe), region (north,
1. south, east, west), city (tier one, tier two, metro)

2. Portfolio type such as corporate, NBFC, SME, retail, home loan, trade receivables

3. Borrower characteristics such as salaried, business, age, gender, income, risk (rating, score)

Portfolio characteristics such as secured, unsecured, vintage basis month of

4. book, maturity of loan

Portfolio management such as internal policy for credit screening/underwriting/monitoring,

5. credit policy, business projections

Quantitative methods such as K-means clustering, Principal Component Analysis (PCA),

6. Classification and Regression Tree (CART).

In general, we have seen FIs perform segmentation basis portfolio type only without any further analysis
which in our view will not suffice the regulatory requirement. FIs should perform further quantitative and
qualitative analysis basis different parameters to segment/pool the portfolio.
 Expected Credit Loss (ECL) | 7

3.3 Significant Increase in Credit Risk (SICR)

Under ECL requirements, FIs will need to compute ECL for either next 12-month or for remaining lifetime
depending on the change in credit risk since initial recognition of the facility. FIs need to classify the facilities
in either of the three stages as highlighted below:

Stage 1 – facilities which do not have significant increase in credit risk and thus 12-month ECL to be
computed for such facilities

Stage 2 – facilities which have significant increase in credit risk and thus lifetime ECL to be computed
for such facilities

Stage 3 – facilities which are in default basis default definition or serving cool-off period.

FIs can use different quantitative and qualitative should be decided. Additionally, FIs should identify
parameters to assess change in credit risk. Some of other relevant parameters to assess SICR and
the quantitative and qualitative parameters that can accordingly classify facilities in different stages.
be used to assess increase in credit risk are:
FIs should also conduct regular back testing as part
• PD comparison (lifetime or 12-month PD of the validation of SICR criteria and add/remove any
comparison at inception and reporting date) parameter basis the same. Some of the methods that
FIs can use for back testing are:
• Rating/PD movement beyond certain
threshold/notch 1. Type I and Type II error to identify false positive
and false negative in assessment of credit risk
• Regulatory backstop of 30 and 90 DPD
2. Roll forward/backward analysis of facilities
• Restructured/modified due to financial difficulties.
moving in different stages should be relatively
• Watchlist or Special Mentioned Accounts (SMA) stable
• Breach of any covenant 3. SICR in facilities should be aligned with relevant
macroeconomic variable outlook.
• Relevant early warning indicators (macro and
micro). Since stage assessment have material impact on
provision requirement, in our view, FIs should have
Currently, many FIs only use backstop and SMA flag
comprehensive staging policy to ensure accuracy of
for SICR assessment. For corporate customers, rating
credit risk assessment in timely manner. This policy
notch movement is also used but FIs do not conduct
should also be validated as part of the model
internal analysis to identify PD threshold basis which
validation exercise performed by the FI.
notch movement will be decided.
In our view, FIs should conduct an analysis to
identify thresholds basis which notch movement
8 | Expected Credit Loss (ECL)

3.4 Probability of Default (PD)

In general, there are two types of PD:

Through The Cycle (TTC) PD which estimates the probability of default over a longer time
horizon, typically across the entire economic cycle. It smoothens out the short-term
fluctuations in the economic environment by considering a full business cycle, which includes
periods of both economic growth and recession

Point in Time (PiT) PD which estimates the probability of default based on the current
economic conditions and the borrower’s present financial health. It is more sensitive to the
short-term changes in the economy and the borrower’s situation.

As per regulatory requirement, FIs need to use PiT issues that we have noticed in the PD computation
PD for ECL computation. FIs can compute either TTC which we think will not be in line with the industry
PD and convert that to PiT PD with macroeconomic practice:
overlay or directly compute PiT term structure with
• Not using data for an entire business cycle (eight
macroeconomic overlay basis remaining lifetime of
to ten years) or for at least five years, even when
the facility.
such data is available
ECL guidelines are generally principal based and
• Poor data quality with issues such as missing
provide leeway to FIs in selecting the methodology it
values, outliers, and incorrect data are noted
wants to adopt for computation purposes. Some of
during model development phase
the methodologies that FIs can use to compute PD
are: • Some FIs with multiple systems to record data
does not have unique identifier to integrate the
• Gross flow rate method
data for a borrower
• New flow rate method
• FIs using legacy rating/scorecard models as
• Application/behaviour scorecard developed and implemented 10-15 years ago
without conducting comprehensive model
• Internal/external rating-based approach
validation to ensure model output is appropriate
• Pluto Tasche approach
• FIs using legacy models despite increase/decrease
• Vasicek single factor model in portfolio which might require further
segmentation/pooling analysis and update of
• Markov chain
models accordingly
• Weibull model
• FIs using vendor model which was developed
• Credit Index using only external data and/or does not have
adequate model development document or
• Super panel hazard model
validation to ensure model output is accurate
• Asset based Merton model
• FIs using methodology which is not appropriate
• Machine learning based models such as XGBoost. for the underlying portfolio/segment.
In our view, FIs should select methodology As per our view, FIs should validate and redevelop or
considering factors such as portfolio, data availability recalibrate, as required, all the legacy
(both historic, current, and future economic data) as rating/scorecard models which were developed
needed for that methodology. Using a single previously for IRB or ECL disclosure purposes.
approach for all portfolios/segments might not be in Additionally, FIs should conduct comprehensive data
line with the industry practice. assessment and ensure quality and completeness of
all critical data elements that are needed to
Additionally, below we highlight some of the key
compute PD.
 Expected Credit Loss (ECL) | 9

3.5 Loss Given Default (LGD)

LGD is an estimate of loss from a facility in case a borrower defaults and can be estimated basis recoveries
that can be made after default. Recoveries can be in the form of cash (unsecured recovery) or by possession
or sale of collateral (secured recoveries). FIs can use different method to compute LGD. Some of the methods
that can be used are:

Workout LGD Market LGD Asset pricing/implied LGD

where FI’s internal data and where market price of where LGD is derived basis
recovery is used for LGD defaulted instruments is used credit spreads on the non-
computation to estimate LGD defaulted risky bonds or credit
derivatives such as CDS

Quantitative models Regulatory LGD

where LGD is derived basis where regulatory prescribed
methods such as linear unsecured LGD and collateral
regressions, fractional logistic, haircuts are used to compute
Jacob Frye, and decision tree secured LGD.
models

As mentioned earlier from PD perspective, even for 4. In our experience, FIs can reduce LGD with
LGD, FIs should select the methodology after effective collateral framework including collateral
considering factors such as portfolio, data availability allocation among different facilities of a
(both historic, current, and future economic data), borrower, and periodic fair valuation of
data accuracy, etc. as needed. Below we highlight underlying collateral
some other areas that FIs should consider while
5. Depending on methodology adopted, FIs should
developing LGD models:
consider all the relevant parameters, cost, and
1. As per current market practice, single unsecured discount rate to compute LGD.
LGD is computed which is applied for all facilities
irrespective of stage, or residual life of the
facility. As generally been noted, LGD should
increase basis credit risk i.e., LGD for accounts in
stage three should be higher than for accounts in
stage two. Similarly, for stage three accounts
also, recovery drops as time elapses and thus
accounts with higher vintage in default should
have higher LGD. For this, FIs can try to compute
progressive LGD or LGD term structure so that
appropriate LGD is used for ECL computation
2. FIs can conduct vintage analysis to determine
time period in which maximum recovery will
occur and accordingly use data for that time
period for LGD estimate. FIs can use methods
like chain ladder to forecast LGD for recent
default cases where recovery has not been
completed
3. FIs can also compute cure rate basis accounts
which went in default but have paid all arrears
and became standard within short period of time
such as within three to six months
10 | Expected Credit Loss (ECL)

3.6 Exposure At Default (EAD)

As per Global Public Policy Committee (GPPC) paper, ECL which will impact P&L statements. By
as published in June 2016, “EAD is an estimate of incorporating prepayments in EAD, FIs can factor in
the exposure at a future default date, taking into principal repayment that borrower is expected to
account expected changes in the exposure after the make over and above the contractual amount and
reporting date, including repayments of principal and thus reduce ECL charge for stage two accounts with
interest, and expected drawdowns on committed long residual lifetime. FIs can use methods such as:
facilities.”3 Thus, EAD should consider both principal
1. Static models basis historical data and average of
and future interest along with any amortisation, and
the same using Conditional Prepayment Rate
prepayment as expected in future.
(CPR) and Single Monthly Mortality (SMM) is
As FIs need to compute lifetime ECL for stage two used for prepayment modelling
and stage three, FIs need to compute lifetime EAD
2. Dynamic models where other factors such as
basis either contractual or behaviour maturity
relevant risk characteristics and economic
assessment of the facility. Additionally, FIs should
conditions are considered for prepayment
ensure that cash flow modelling for different
modelling
purposes such as ALM, financial reporting, and other
purposes are aligned. 3. Advanced models where advanced quantitative
models such as hazard model or machine
Some of the other modelling aspects that need to be
learning models are used for modelling.
done for EAD modelling are:
In our view, FIs should model EAD considering
Credit Conversion Factor (CCF) – ECL framework
above factors so that they can rationalise ECL and
requires ECL to be computed for non-funded
are not over and under providing in terms of
exposure such as Letter of Credit (LC) and Letter of
provisions.
Guarantee (LG). For the same, FIs can use different
methods to model CCF basis data availability:
1. Basis internal historical data using methods such
as cohort approach or fixed horizon method
2. Regression based model basis relevant factors
3. Monte-Carlo simulation-based models
4. Machine learning based model such as K-Nearest
Neighbors (KNN) and Support Vector Machine
(SVM)
5. Regulatory prescribed factors.

Amortisation- Basis the capability of the ECL system

implemented, FIs can use amortisation schedule as
per contractual cash flows i.e., monthly, quarterly, or
semi-annually or annually as per terms. As of now,
we have seen FIs either do not use amortisation in
EAD computation or assume annual amortisation.
ECL computed under either of the approaches will be
conservative depending upon the stage and residual
maturity of the contract. But FIs will need strategic
system or models on platforms such as
R/SAS/Python which can handle large data set that is
required to use actual cash flow or cash flow
generated basis models.

Prepayment – Incorporation of prepayment in ECL

computation will help FIs rationalise EAD and in turn

3. "The implementation of IFRS 9 impairment requirements by banks” issued by the Global Public Policy Committee of representatives of the six
largest accounting networks in June 2016
 Expected Credit Loss (ECL) | 11

3.7 Effective Interest Rate (EIR)

Effective interest rate is the rate that exactly should compute CEIR and for other assets, it should
discounts estimated future cash payments or compute EIR. The calculation of EIR includes all fees,
receipts through the expected life, or when transaction costs, and all other premiums or
appropriate, a shorter period of the financial asset or discounts which are directly related to the acquisition
financial liability to the gross carrying amount of a of financial assets. Below are some key
financial asset or to the amortised cost of a financial considerations for computation of EIR:
liability. As per the framework, for POCI assets, FIs

How to evaluate whether amortisation of transaction cost/fees will be on EIR/SLM basis

How to evaluate whether volume-based incentives will form part of EIR

How to evaluate whether to use expected life vs contractual life (E.g., housing loan, instruments
with call /put options, prepayment clause)

How to determine whether interest income/expenses will be presented on gross basis

How to compute EIR for floating rate instruments.

12 | Expected Credit Loss (ECL)

3.8 Macroeconomic (MEV) Overlay and Weighted ECL

FIs need to compute ECL considering historical information, current economic conditions, as well as future
macroeconomic outlook. Additionally, the framework requires FIs to compute ECL under different
macroeconomic forecast and compute weighted ECL basis different macroeconomic conditions. Some of the
points that FIs should note in selection and modelling for weighted ECL are listed below:

01 FIs can either forecast MEV basis internal 02 Relevant macroeconomic models for
models or use external forecast published different segments/pools should be selected
by reputable agencies basis quantitative and qualitative analysis

03 FIs should select same variables as far as 04 MEV forecast for the variables should be used
possible for ECL computation and other risk consistently in different risk purposes such as
management and stress testing purposes ALM and stress testing

05 06 Rationale for weights used for different

Assumptions and limitations in selection
macroeconomic scenarios should be
and computation of forward looking ECL
documented and validated as part of model
should be identified and validated
validation.

In our view, not incorporating forecasted MEV in ECL computation will not comply with the regulations. FIs
should select relevant MEV and incorporate these while modelling ECL.

3.9 Post Model Adjustment (PMAs)

and Overlays

Post model adjustments and overlays are used

where the risks and uncertainties cannot be correctly
predicted or quantified basis model. Overlays are the
adjustments made to the existing models’ output,
and these adjustments can be subjective or
judgmental or at times both. These risks are not
captured by the models because they are not
designed to address uncertainties, such as those
seen during the global financial crisis of 2007-08,
COVID-19 pandemic or recent interest rate hikes. The
key considerations for computation of PMA and
overlays are:
• The approach for applying PMAs should be
appropriate and clearly understandable, ensuring
that these adjustments effectively address the
limitations of the model
• The methodology for computing and applying
PMAs must be thoroughly documented
• There should be defined governance structure to
incorporate such PMAs
• FIs should conduct back testing to validate for
relevance and adequacy of PMAs.
 Expected Credit Loss (ECL) | 13

4. Validation of the models

As per RBI guidelines, FIs should have independent first line (model developer) and second line (model
three lines of defence to manage model risk and as validator) have performed due diligence and
per that, all ECL models should be independently complied with all applicable internal policies and
validated4 before deployment as well as periodically procedures as well as regulatory guidelines.
monitored and validated as per MRM policy of the
Model validator to conduct comprehensive check on
FI.5 Additionally, entire process of model
entire model development process and challenge
development and validation has to be reviewed by
model developer. Some of the key areas that
internal audit (third line of defence) to ensure both
validator should check are:

Data: Complete data steps including extraction of data from various sources, data massaging,
imputations, and feature engineering as used to prepare modelling data should be assessed.

Methodology: Model methodology should be assessed basis portfolio/segment as well as

regulatory requirement and industry benchmark.

Assumptions and Limitations: All assumptions (subjective or analytical) and limitations in

data, methodology should be assessed, along with rationale and any controls that are placed
to mitigate risk arising from use of such assumptions and limitations.

Documentation: As mentioned in RBI guidelines, model document should be detailed enough

so that it provides good understanding of model to any independent reader.

Implementation: In case of any system or stand-alone codes, validator should assess

robustness of system implementation or code development to ensure output from these is
accurate.

Controls: Validator should assess controls in ECL computation process to ensure no

inadvertent changes are done in model which might result in inaccurate output.

Overrides: All the management overrides such as those in selection of variable, in binning of
score, and weightage of variable, should be validated along with their impact on model
output.

Sensitivity and back testing: Validator should conduct sensitivity and back testing analysis to
test model robustness and accuracy of output and accordingly raise an issue in case the
results are not in line with the expectations.

PMA: Any overlay/adjustments over and above the model output should be assessed by the
validator to evaluate the need for such overlays and the appropriateness of methodology. The
validator shall also back test the adjustments to ensure that provisions are adequate.

Overall, in our view, FIs should have independent AI/ML based model or advanced statistical models or
model validation team or hire external consultants for simple average based model will be different and
who will perform validation of these models both can be defined in model validation policy. As
pre-deployment as well as going forward. prescribed in RBI draft circular on model risk
Additionally, depending on the methodology used management in credit, such validation rigor will also
for modelling of various components, rigor and tests be applicable for third-party, or consultant developed
will vary. For example, validation requirement for models.

4. Model Risk Management issued by KPMG in India in November 2024.

5. Draft Circular - Regulatory Principles for Management of Model Risks in Credit” issued by the Reserve Bank of India in August 2024.
14 | Expected Credit Loss (ECL)

5. Key challenges
Below are the key challenges, grouped into different categories, that we have noticed while assisting
different FIs globally in their journey of IFRS9/IndAS109/ASC 326 implementation and subsequent validation:

Data

Basis our experience, data is biggest challenge for ECL computation. Some of the data challenges
we noticed are in terms of historical data availability, data quality, data integration across various
systems, data volume, and lack of availability of critical data in digital format in the system. Due
to such issues, many FIs go for regulatory backstop for different components which might result
in conservative ECL estimation

Methodology

Many FIs do not select model methodology basis availability of volume and quality of data, or
considering process, IT, and people capability. In our view, using appropriate methodology is
critical in having robust ECL framework and thus it is selected after due considerations on various
aspects. Also, to avoid undue volatility in provisions, model methodology selected should be
robust and should not require frequent changes

Process

ECL computation requires coordination among different departments and thus having robust
process for entire ECL computations i.e., data extractions, model running, computation and
reporting is critical. Additionally, as the numbers are reported in financial statements, any delay
or break in the process might result in inaccurate or delay in statements which might impact
reputation of the FI. Implementing Target Operating Model (TOM) and Business as Usual (BAU)
process requires a lot of deliberation among different stakeholders

People
Having adequate resources in different departments with relevant knowledge and experience is
critical both during development of framework and as part of BAU. Many a times we find that FIs
have key man risk in critical departments. FIs can take help of consultants as and when required
for this purpose

IT System
Many FIs still rely on spreadsheets for ECL computation which might not be optimum given the
volume of data that needs to be processed for both - computation of different components and
final ECL estimates at each reporting date. In our view, FIs should invest in either strategic or
tactical system depending on volume of data that is expected to be handled
 Expected Credit Loss (ECL) | 15

Governance and controls

Defining appropriate governance with clear roles and responsibilities for all the stakeholders is
critical and requires deliberations with different teams. Also, appropriate Turnaround Time (TAT)
and escalation matrix should also be defined so that there is no delay in financial reporting.
Additionally, as the numbers are reported in the financial statements, designing, and
implementing effective controls for the entire ECL framework process became paramount

MIS
There is no MIS report/dashboard so that senior management have adequate oversight on
movement of ECL provisions and can intervene in timely manner before provisions breach risk
thresholds.

In our view, FIs should assess all such areas to

ensure they have robust ECL framework which
can work efficiently as part of BAU process.
Additionally, there should be clear plan on how
any gaps and challenges will be addressed within
stipulated timelines. To emphasise again, as ECL
provisions need to be reported in the financial
statements along with change in provision at
each reporting date, FIs should strive to have
robust ECL framework so that volatility in
provisions is reduced which will give confidence
to regulators, shareholders, potential investors,
and other stakeholders.
16 | Expected Credit Loss (ECL)

6. Key impacted areas

In our experience, ECL implementation will not only any major failure might have adverse
impact complete credit lifecycle but also other reputational impact. Also, as ECL provisions will
functions and departments and thus it is imperative need to be disclosed on each reporting date, any
that different stakeholders understand the undue increase in provisions might also create
requirements and enhance process accordingly. Key reputational risk event for the FI
areas that will be directly impacted are listed below:
8. Risk appetite and limits: As provisions will
1. Pricing: As ECL impact need to be recorded in the impact financial statements, it is imperative that
books from day one, it is imperative that FIs FIs should update risk appetite and limit
factor in such cost while pricing the loans. In our framework to minimise impact of provisions on
view, FIs should implement risk-based pricing the FI. Also, this framework should be dynamic
with expected provision as one of the factors in and should be updated basis factors such as
determining the rate of interest at which it will macroeconomic conditions, strategic visions, and
offer loans to the borrower business focus
2. Liquidity: As provisions will hit P&L, increase in 9. Stress testing: FIs should add relevant
provisions will result in reduced profitability and scenarios/sensitivities that will impact ECL
retained earnings which will directly impact provisions as part of stress test framework
capital available with the FIs and will impact
10. Other Fair Value regulations: As ECL framework
liquidity ratios such as Liquidity Coverage Ratio
comes into effect, other regulatory requirements
(LCR) and Net Stable Funding Ratio (NSFR). This
such as disclosures (IFRS7/IndAS 107) and
might result in impact on capital and funds
regulation related to FV hierarchy (IFRS13/IndAS
requirement planning of the FIs, and they may
113) may also come into effect. These
require raising capital from alternate methods
regulations will on one hand increase
3. ALM: As ECL framework requires computation of transparency in reporting but on another hand, it
provisions on future cash flows, any increase in also increases compliance requirements.
provisions will reduce assets and might create
In our view, FIs should perform a holistic review of
asset liability mismatch. Additionally, FIs may
ECL computations and its impact on various
face potential mismatch in case different
functions. Accordingly, FIs can plan to address any
assumptions for cash flow modelling (such as
gaps, and high impact areas which might be at risks
prepayment, and amortisation) are used for ECL
once ECL framework is implemented.
computation and ALM
4. Operational: As ECL computation will require
significant modelling and system requirement,
this increases chance of operational risk event
such as system failure, manual error, and
regulatory or auditor finding on framework
5. ICAAP: With high reliance on models to compute
ECL provisions, there will be increase in model
risk faced by FIs which may qualify as material
risk under Pillar 2 of ICAAP exercise. In that
scenario, FIs will need to consider it while
performing ICAAP assessment and might need to
maintain certain capital under the framework
6. Regulatory compliance: Basis our experience,
ECL provisions always qualify as material risk for
both statutory auditor and regulator and will be
the focus area of both the auditor and the
regulator. Any issue identified in the process
might create a compliance issue
7. Reputation: As there are many areas that will be
impacted by implementation of ECL framework,
 Expected Credit Loss (ECL) | 17

7. Conclusion
As highlighted in this paper, ECL provisions
requirements will not only impact credit department
but all other aspects of business and risk
management of the FIs. Thus, it is imperative that FIs
should design TOM in such a way that the
framework is robust as well as optimise provision as
a part of BAU process. Key points that FIs should
note for such frameworks are:

• Every process takes time to settle and thus it is

advisable that FIs to conduct multiple dry runs of
proposed framework and enhance as per gaps
identified

• Identify all gaps and challenges and put in place a

plan to resolve them. The plan should be tracked
at central level with involvement of senior
management

• Allocate adequate resources (both human and

capital) to design and implement the framework.

• Key decisions related to model methodology, and

system (strategic or tactical) should be taken after
considering data as well as medium- and long-
term strategic vision

• Training sessions not only for junior staff but also

senior management should be conducted

• Wherever required, external consultants should

be onboarded to assist in the process

• Holistic view should be taken while developing

the framework after considering the potential
impact on other areas and how it can be
minimised

• Model methodology and assumptions should be

aligned so that there is no mismatch in the same
component used for different purposes.
18 | Expected Credit Loss (ECL)

Glossary
ECL Expected Credit Loss
IFRS International Financial Standards
CECL Current Expected Credit Losses
FI Financial Institution
PMA Post Model Adjustment
RBI Reserve Bank of India
IASB International Accounting Standards Board
LGD Loss Given Default
PD Probability of Default
EAD Exposure at Default
GMM General Measurement Model
POCI Purchased or Originated Credit-Impaired
ICAAP Internal Capital Adequacy Assessment Process
SICR Significant Increase in Credit Risk
CEIR Credit adjusted Effective Interest Rate
SPPI Solely Payment of Principal and Interest
SOP Standard Operating Procedures
EBA European Banking Authority
NPA Non-performing Assets
PCA Principal Component Analysis
CART Classification and Regression Tree
ALM Asset Liability Management
SMA Special Mentioned Accounts
TTC Through the cycle
PiT Point-in Time
IRB models Internal Rating Based models
BAU Business As Usual
TAT Turnaround Time
GPPC Global Public Policy Committee
CCF Credit Conversion Factor
LC/LG Letter of Credit/Letter of Guarantee
KNN K-Nearest Neighbors (KNN)
SVM Support Vector Machine (SVM)
CPR Conditional Prepayment Rate
SMM Single Monthly Mortality
MIS Monthly Information System
EIR Effective Interest Rate
MEV Macroeconomic variable
PMA Post Model Adjustments
LCR Liquidity Coverage Ratio
NSFR Net Stable Funding Ratio
FV Fair Value
IndAS Indian Accounting Standard
NBFC Non-Banking Financial Company
SME Small and Medium Enterprise
 Expected Credit Loss (ECL) | 19

Acknowledgements
We are sincerely grateful to the following people from the ecosystem who have helped in the preparation
of this report.

KPMG in India research team

• Gautam Bhagat (Director – Financial Risk Management)
• Abhijeet Deshmukh (Associate Director – Financial Risk Management)
• Yash Jayawant (Associate Director – Financial Risk Management)
• Sayan Dutta (Manager – Financial Risk Management)
• Akanksha Sharma (Manager – Financial Risk Management)
• Amber Miglani (Assistant Manager – Financial Risk Management)
• Gaurav Bansal (Assistant Manager – Financial Risk Management)

KPMG in India compliance and design team

• Pooja Patel (Assistant Manager)
• Nisha Fernandes (Associate Director)
• Karthika Prabasankar (Executive)
 20

KPMG in India contacts:

Akhilesh Tuteja
Head – Clients & Markets
E: [email protected]

Manoj Kumar Vijai

Office Managing Partner & HoF
E: [email protected]

Rajosik Banerjee
Head - FRM & Deputy Head – RA
E: [email protected]

Amitava Mukherjee
Partner – FRM
E: [email protected]

Somdeb Sengupta
Partner – FRM
E: [email protected]

Rachit Gupta
Director – FRM
E: [email protected]

kpmg.com/in

Access our latest insights

on KPMG Insights Edge

Questions to consider:
Follow us on:
kpmg.com/in/socialmedia
How can the NRC support in designing a
? roadmap for ensuring pay disparities are
lowered in the coming years?
What are considerations for the NRC to
?The information contained
understand theherein
marketis of adrivers
general for
nature and is not intended to address the circumstances of any particular individual or entity. Although
talent
we endeavor to provide accurate and timely information, there can be no guarantee that such information is accurate as of the date it is received or that
attraction?
it will continue to be accurate in the future. No one should act on such information without appropriate professional advice after a thorough
examination of the particular situation.
KPMG Assurance and Consulting Services LLP, Lodha Excelus, Apollo Mills Compound, NM Joshi Marg, Mahalaxmi, Mumbai - 400 011
Phone: +91 22 3989 6000, Fax: +91 22 3983 6000.
© 2025 KPMG Assurance and Consulting Services LLP, an Indian Limited Liability Partnership and a member firm of the KPMG global organisation of
independent member firms affiliated with KPMG International Limited, a private English company limited by guarantee. All rights reserved.
The KPMG name and logo are trademarks used under license by the independent member firms of the KPMG global organisation.
Printed in India. (022_THL1224_KP)