LGD vs PD models
Darshika Srivastava
Associate Project Manager @ HuQuo | MBA,Amity Business School
Rating system validation, LGD model and Risk appetit
Istanbul, 18 December 2014
2?AGENDA
Rating system validation
LGD models development
Early warning for risk appetite
3?Risk Appetite Framework
Regulatory input
The international regulatory capital standards recommended by the Basel Committee (known as Basel standards) in response to the recent financial crisis have been transposed into Turkish law through secondary legislation enacted by the Banking Regulation and Supervision Agency (BRSA)
CREDIT RISK MODELS
Origination
Credit Strategy
Loan Pricing
Early Waring
Risk Appetite Framework
Supervision Agency
Banking Regulation
4?Calculation of capital minimum requirements and risk parameters
PD, LGD and EAD are the fundamental risk parameters for evaluating the credit risk level and consequently defining the minimal capital requirements
PD
Credit risk Parameters
EAD
LGD
Reflects the Probability of Default of the counterparty
Reflects the expected % of loss on a facility after default
Reflects the expected amount of exposure at the time of the default
Internal Rating Based Foundations
Internal rating
Supervisory values (*)
Internal Rating Based Advanced
Standard
Internal LGD
Internal EAD
Supervisory values
External rating
5?Internal Validation Process Risk estimation system
Regulatory Back-ground and Framework
The Basel Committee defines widely the concept of validation..
“…500 : Banks (adopting IRB approach) must have a robust system in place to validate the accuracy and consistency of rating systems, processes, and the estimation of all relevant risk components. A bank must demonstrate to its supervisor that the internal validation process enables it to assess the performance of internal rating and risk estimation systems consistently and meaningfully.”
Accuracy
Consistency
Internal Validation Process
Risk estimation system
Assessment on internal rating system performance (backtesting) but for IRB banks.
Judgmental analysis on methodological approach applied in the rating system in comparison to the best practices.
Constant monitoring on the quality of all the informative tools performing in the rating system.
Application and use of rating in all business areas (use test) and in the firm-wide decisioning governance.
6?Validation Process Validation Process in the Basel Scheme
The scheme below shows key components of Validation Process in Basel Scheme
Internal Validation
Supervisory Examination
Validation of rating system
Validation of rating process
Internal Use by credit officers
Reporting and problem handling
Data quality
Risk Components
Benchmarking
Backtesting
PD
LGD
EAD
Model Design
IT SYSTEM
7?Validation Process Validation Process in the Basel Scheme
The scheme below shows major actors and “theoretical” steps of the validation process
Governance /Reporting
Information Systems
Model design
Use test
Data
Risk components
VALIDATION OF THE RATING SYSTEMS
Supervisor
examination
Internal audit/compliance
Validation/
Authorization dossier
VALIDATION OF THE RATING PROCESS
Individual bank
Regulatory capital
RWA
=?x?8%
EAD
1,06
PD
LGD
;
f (
(
Credit approval
Credit risk management perspective
Regulatory perspective
M
Credit monitoring
……
Authorization
to the use of IRB parameters
Default risk -PD
Loss given default
IRB parameters
Exposure at default
8?Validation of internal credit risk models
The internal credit models system should be validated in order to be compliant to Basel Standards; the validation process takes into account two different areas: quantitative and qualitative
PD
DEVELOPMENT
Basel Standars
IMPLEMENTATION
LGD
EAD
Internal credit models
Qualitative
VALIDATION
Quantitative
PPT fuffa che introduce il tema della validazione: prima c’è lo sviluppo, poi la validazione ed infine l’implementazione.
Importante dire che la fase di sviluppo non è ?one shot? ma dev’essere monitorata tramite la validazione al fine di individuare i punti di miglioramento year by year
9?Validation – Qualitative Area
The qualitative test is meant to certify the relevance and quality of the data used and the correct application of the quantitative methods. A rating process should only be carried out if the internal credit models system receives a positive assessment during the qualitative test
The model’s design is validated on the basis of the rating model’s documentation. In this context, the scope, transparency and completeness of documentation are already essential validation criteria
Delineation criteria for the rating segment
Description of the rating method/model type/model architecture used
Reason for selecting a specific model type
Completeness of the (best practice) criteria used in the model
Data set used in statistical rating development
Quality assurance for the data set
Model development procedure
Quality assurance/validation during model development
Documentation of all model functions
Calibration of model output to default probabilities
Data quality
Model design
Internal Use test
In statistical models, data quality stands out as a goodness-of-fit criterion even during model development. Moreover, a comprehensive data set is an essential prerequisite for quantitative validation
Completeness of data in order to ensure that the rating determined is comprehensible
Volume of available data, especially data histories
Representativity of the samples used for model development and validation
Data sources
Measures taken to ensure quality and cleanse raw data.
Validating the internal use of the rating models (use test) refers to the actual integration of rating procedures and results into the banks in-house risk management and reporting systems. With regard to internal use, the essentialaspects of the requirements imposed on banks using the IRB approach under Basel II
Design of the banks internal processes which interface with the rating procedure
as well as their inclusion in organizational guidelines
Use of the rating in risk management (in credit decision-making, risk-based
pricing, rating-based competence systems, rating-based limit systems, etc.)
Conformity of the rating procedures with the banks credit risk strategy
Functional separation of responsibility for ratings from the front office
Employee qualifications
User acceptance of the procedure
The users ability to exercise freedom of interpretation in the rating procedure
10?Validation – Quantitative Area
Quantitative validation is required for all credit models in use and it should primarily be performed with the data gained through use of the model in the bank. A quantitative test could provide the information concerning the performance of the credit models
The term discriminatory power refers to the fundamental ability of a rating model to differentiate between bad (credit default occurs) and good (credit default not occurs) cases
stability
performance
calibration
The assignment of default probabilities to a rating models output is referred to as calibration. The quality of calibration depends on the degree to which the default probabilities predicted by the rating model match the default rates actually realized
Frequency Distribution of Good and Bad Cases
Transaction matrix
Binomial Test
Chi-Square Test
PD actual vs PD fitted
Accuracy Ratio
ROC Curve
Errors the 1° and 2° type
Cumulative frequency bad /good cases
Denisty function for bad/good cases
Kolmogorov-Smirnov Test
CAP Curve
Changes in the discriminatory power of a rating model given forecasting horizons of varying length and changes in discriminatory power as loans become older
Changes in the general conditions underlying the use of the model and their effects on individual model parameters and on the results the model generates
benchmark
back-testing
11?AGENDA
Rating system validation
LGD models development
Early warning for risk appetite
12?Approaches to LGD estimation
There are broadly 2 ways of measuring LGD
-Sub-Task-
-Description-
-Counterparties-
MARKET
LGD
WORKOUT
It makes use of Market Data
The Loss Value of the market assets after the default event, is the base for the LGD Estimation
It needs a liquid asset market before and after default
Corporate
SME
Retail
It makes use of Internal Data
It needs as first step the analytic calculation of the LGD observed on historical default events, on the basis of expenses, charges and recoveries, observed in a period of time
It needs the collection of the cash flow characterizing the contracts after the default event, and the adoption of appropriate actualization hypothesis.
Large Corporate
Financial Instit.
Sovereign
Focus of the presentation
13?Approaches to LGD estimation
What do we know and we expect about Loss Given default?
Definition
Methodological foundations
The loss is measured according to an “economic logic” (not accounting) considering the effect of time in the loan recovery
The perimeter is represented by the default positions which have completed their process of debt recovery
LGD is usually defined as the ratio of losses to exposure at default, and it considers:
the set of estimated recovery cash flows to be received by the lender resulting from the workout and/or collections process, properly discounted
Workout expenses (collections, legal, etc)
Downturn: LGD must reflect an appraisal of the expected loss on a transaction in case of default and in a downturn scenario (periods where PD are above the average).
Exposure at default (EAD)
A
Expenses
C
Default
B
Closing of the collection actions
Recovery flows
LGD=
EAD – discounted recovery flows + discounted expenses
EAD
A – B + C
A?=
14?Approaches to LGD estimation
Workout estimation can approached differently in terms of Model Design
Full model
Performing
DEFAULT
Performing
LOSS
MODEL FULLY COMPLIANT AND ALIGNED WITH THE RECENT EVOLUTIONS OF THE REGULATOR
WRITE-OFF
Performing
Cure rate (Partial) Model
LOSS
Performing
Default B2
Performing
cure
LOSS
danger
15?High level project outiline – LGD workout estimation
To properly address LGD estimation, CRIF proposes a project framework composed of 6 worksteps
Credit Management Process analysis
Data extractions and risk parameters estimation
Project phases
Preliminary assessment
Credit portfolio analysis
Modeling
framework
Improvement expectations
Analysis which aims to identify main areas of improvement and gap towards local regulatory requirements framework and international best practice
Analysis of the active loan portfolio with a breakdown by:
Product
Collateral
…
Interviews with process owners
Debt collection
Defining the methodological framework for calculating LGD in compliance with the Basel requirements
Data extractions from identified archives
Construction of Development Data mart
Estimation of LGD parameters
Identification of areas for improvement of the LGD estimates
Prioritization and planning of future activities
Activity
Gap analysis
Identifying the relevant dimensions of analysis
Identification of
Key processes characteristics
Any limitations or constraints to the estimation process
Technical meetings on LGD modeling methodology in line with the process and data characteristics
Data request
Presentation of the results (estimates, performance,.)
Proposed action plan
Output
16?Areas of investigations Analysis of the expenses
Project outiline –LGD workout estimation
High level description of Preliminary Assessment phase
Credit Management Process analysis
Data extractions and risk parameters estimation
Project phases
Preliminary assessment
Credit portfolio analysis
Modeling
framework
Improvement expectations
Identifying main areas of improvement and gap towards local regulatory requirements framework and international best practice
Analysis of the active loan portfolio with a breakdown by:
Product
Collateral
…
Interviews with process owners
Debt collection
Defining the methodological framework for calculating LGD in compliance with the BII requirements
Data extractions from identified archives
Construction of Development Data mart
Estimation of LGD parameters
Identification of areas for improvement of the LGD estimates
Prioritization and planning of future activities
Activity
Gap analysis
Identifying the relevant dimensions of analysis
Identification of
Key processes characteristics
Any limitations or constraints to the estimation process
Technical meetings on LGD modeling methodology in line with the process and data characteristics
Data request
Presentation of the results (estimates, performance,.)
Proposed action plan
Output
Areas of investigations
Default definition
Discount rate
Analysis of the expenses
Downturn LGD
17?Approach to LGD estimation
High level project outiline -workout estimation
High level description of Modeling framework phase
Credit Management Process analysis
Data extractions and risk parameters estimation
Project phases
Preliminary assessment
Credit portfolio analysis
Modeling
framework
Improvement expectations
Identifying main areas of improvement and gap towards local regulatory requirements framework and international best practice
Analysis of the active loan portfolio with a breakdown by:
Product
Collateral
…
Interviews with process owners
Debt collection
Defining the methodological framework for calculating LGD in compliance with the BII requirements
Data extractions from identified archives
Construction of Development Data mart
Estimation of LGD parameters
Identification of areas for improvement of the LGD estimates
Prioritization and planning of future activities
Activity
Gap analysis
Identifying the relevant dimensions of analysis
Identification of
Key processes characteristics
Constraints to the estimation process
Technical meetings on LGD modeling methodology
Data request
Presentation of the results (estimates, performance,.)
Proposed action plan
Output
Approach to LGD estimation
Traditional/
empirical approach
Econometric model
It is based on empirical evidences (descriptive statistics)
It does not give evidence of the statistical significance of adopted risk drivers
It is strongly affected by sample size
It allows the use of relevant factors: statistically meaningful model variables
It easily allows the usage of continuous explanatory variables
It provides a measure of the marginal contribution of each variable as well as their interaction
18?Data Requirements definitions Data Aggregation Criteria definition
High level project outiline -workout estimation
High level description of Data extractions and Risk Parameters estimation phase
Credit Management Process analysis
Data extractions and risk parameters estimation
Project phases
Preliminary assessment
Credit portfolio analysis
Modeling
framework
Improvement expectations
Identifying main areas of improvement and gap towards local regulatory requirements framework and international best practice
Analysis of the active loan portfolio with a breakdown by:
Product
Collateral
…
Interviews with process owners
Debt collection
Defining the methodological framework for calculating LGD in compliance with the BII requirements
Data extractions from identified archives
Construction of Development Data mart
Estimation of LGD parameters
Identification of areas for improvement of the LGD estimates
Prioritization and planning of future activities
Activity
Gap analysis
Identifying the relevant dimensions of analysis
Identification of
Key processes characteristics
Any limitations or constraints to the estimation process
Technical meetings on LGD modeling methodology in line with the process and data characteristics
Data request
Presentation of the results (estimates, performance,.)
Proposed action plan
Output
Key milestones
Data Requirements definitions
Data Quality
Criteria definitions
Data Aggregation Criteria definition
Vintage
Correction
DM Design
DM Validation
LGD Estimation
Structural
Specification
Data mart building
Corrective
Actions
LGD Workout
Final LGD Model
19?Segmentation criteria With/without collateral/guarantees
High level project outiline -workout estimation
High level description of Data extractions and Risk Parameters estimation phase
Credit Management Process analysis
Data extractions and risk parameters estimation
Project phases
Preliminary assessment
Credit portfolio analysis
Modeling
framework
Improvement expectations
Identifying main areas of improvement and gap towards local regulatory requirements framework and international best practice
Analysis of the active loan portfolio with a breakdown by:
Product
Collateral
…
Interviews with process owners
Debt collection
Defining the methodological framework for calculating LGD in compliance with the BII requirements
Data extractions from identified archives
Construction of Development Data mart
Estimation of LGD parameters
Identification of areas for improvement of the LGD estimates
Prioritization and planning of future activities
Activity
Gap analysis
Identifying the relevant dimensions of analysis
Identification of
Key processes characteristics
Any limitations or constraints to the estimation process
Technical meetings on LGD modeling methodology in line with the process and data characteristics
Data request
Presentation of the results (estimates, performance,.)
Proposed action plan
Output
Segmentation criteria
With/without collateral/guarantees
The LGD model will be differentiated according to any segmentation criteria in order to capture the peculiarities and specific business practices in debt collection management
Segments
Type of facility
Retail
SME
Corporate
Committed
Uncommitted
…
Unsecured
Personal
Mortgages
…
20?AGENDA
Rating system validation
LGD models development
Early warning for risk appetite
21?Overview
The Risk Appetite Framework (RAF) defines the amount and the type of risks that the financial institution wants to assume according to its ability and to the strategic objectives and business that it has agreed
The recent financial crisis has shown that an effective RAF is an useful and relevant way to obtain a good Governance of a financial institution
For this reason national and international regulators are placing greater attention to this topic in order to ensure that there is consistency between the risks actually undertaken and those perceived by decision-making bodies of the Institute
Overview
The RAF has the objective to support the corporate bodies for the decision making in order to increase the awarness about the risks linked to the business model
The financial institutions should develop an effective RAF that is institution-specific and that reflects its business model and organisation, as well as to enable financial institutions to adapt to the changing economic and regulatory environment in order to manage new types of risk
Objective
Top management
Capital
Risk
Strategy
Shareholders
Rating agencies
Customers
Regulator
22?Key Definitions
The RAF is an useful tool that is able to synthesize the risk profile of an Institution. It expresses the capacity (Risk Capacity), the appetite (Risk Appetite) and the actual amount of risk (Risk Profile) undertook both at organization level and both for each individual business unit / type of risk (size)
Risk Capacity
dimension 1
dimension 3
dimension 4
dimension 2
dimension 5
Risk Appetite
The maximum level of risk that the financial institution can assume given its current level of resources before breaching constraints determined by regulatory requirements or other restrictions imposed by the shareholders
Which is the level of risk that the bank can undertake?
The aggregate level and types of risk a financial institution is willing to assume within its capacity to achieve its strategic objectives and business plan
Which is the level of risk that the bank wants to undertake?
Risk Profile
It is the risk actually undertaken, measured in a certain moment
Which is the level of risk that the bank is undertaking at the moment?
23?Risk Appetite Framework
The development of an effective Risk Appetite is an interactive and circular process in order to achieve a continuous improvement of the methodologies, governance processes and operational tools. The Early Warning system represents the main tool useful for monitoring the correct application of the RAF Framework
Process to define the Risk Appetite(1)
Methodology definition of RA: process to definine the position of the target risk appetite in terms of quality and quantity, based on both internal and external aspects
Approval process of RA: determination of the phases and the key players to approve the placement of the risk appetite defined
Conversion of RA in Risk Limits: application of the position of the RA in the ordinary management of the institute by establishing limits and specific processes
Reporting and monitoring: development and definition, at the overall level, of a system of analysis and communication of trend risks that is specific to business unit and type of risk
Early warning: development of a system of risk thresholds and definition of the means to monitor these thresholds, in addition to the definition of the underlying management process
1
Methodology definition
5?2
Early warning
Actors involved
Board
Commitees
Risk Management
Finance
Business Unit
Approval process
4?3
Reporting and monitoring
Risk Limits
FOCUS OF THE NEXT SLIDES
(1) The process should be linked to the size and the proprety of the financial institution
24?Framework
Early Warning Systems have to be indented within the ongoing monitoring of the performing (non delinquent) credits
Objectives
Anticipating, as much as possible, situations of potential deterioration providing to the managers a solution to prevent the default of the counterparty or limit the damage. (Control of the First Level)
1
Perimeter
Performing credits at high risk, or credit lines not yet expired/exceed but showing a higher degree of risk (i.e: fault signals) or because of a review by the relationship manager on other performing loans.
2
Tools
Warning signals and internal behavioral indicators or external data integrated with internal rating (where available) based on statistics and implemented as a decisional tree.
3
Oggi il loro ruolo sta mutando: da semplice supporto ai gestori per la prioritizzazione delle posizioni da trattare a un sistema fatto di priorità ma anche di regole e responsabilità ben definite e monitorate a loro volta. Il tema di early warning e' oggi oltremodo caldo perche' se prima incorporava solo un significato di "rilevazione" oggi e' strettamente collegato al significato di "azione"
IT
Credit Monitoring solutions and sophisticated systems able to produce a monitoring report connection with a “laboratory” environment; new outsourcing services (EWAAS)
4
Riproduzione vietata
25?Objectives
As within any decision-making system, the Early Warning system requires a definition of a target event to be developed and subsequently measured.
Since the objective is to anticipate the transition to the default, the target event is designed to intercept a 'preliminary‘ stage which is made to match with the "deterioration" of the positions, combining quantitative elements (increased days overrun) with “managerial” elements (worsening of the Customer position).
From the operational point of view the target of the "deterioration" of the positions can be analyzed through the exploration of "roll rate", as shown in the following example:
Within this framework, Early Warning models’ target type deviates significantly from the definition of “default” both in terms of events and time horizon considered.
Il rischio dei sistemi di EW come di qualunque sistema andamentale è che nel tempo l’attività di gestione generata dal processo alteri la percezione del rischio: ad esempio abbiamo visto in qualche caso che posizione gestite in base al numero di giorni di sconfinamento presentavano tassi di rientro in bonis più elevati oltre 180 rispetto a quelli gestiti oltre i 90 semplicemente perché l’azione collegata ai due diversi livelli di delinquency hanno efficacia diversa.
Riproduzione vietata
26?Perimeter
The scope refers to a segmentation of the “Banking book”, well represented (in reverse order of severity)
Handling class
Description
A. Trial, legal handling
Credit collection through legal actions
B. Deteriorated position
Substandard loans, past due, restructured
C. Breaches
Formally performing, but showing significant breaches (i.e. expired or exceeded credit)
D. Performing high risk
Credit lines not yet expired/exceed but showing a higher degree of risk (i.e: fault signals as early warning; classification under observation for other reasons; speculative grade rating; etc.).
E. Performing high risk
Credit lines not yet expired/exceed for which has been given rise to a review of the relationship (increase of overdraft due to a prolonged tension in the use, increased maturity, etc.)
F. Performing
All other performing credits
Early Warning Sysyetms usually take into account cases D, E and F
Riproduzione vietata
27?Approach – Designing a decision tree
1?2?3?4?5
Long list definition
Model development
Backtesting
?Traffic light? design
Implementation
It’s a shared process with credit expert, sales network and external suppliers
Drafting a Long List with different priority levels (High, Medium, Low) for each (Corporate, SME and Retail)
?Performance? definition
Historical indicators data analysis
Implementation of a decision tree
Supplement the decision tree with breaches and Business evaluation
Choosineg an ?out of time? sample
Data Analysis
Performance measurement and evaluation
Grouping breaches in order to prioritize the correction actions to be carried out (?Traffic light?)
Workload measurement and evaluation
Availability analysis of the internal/external data sources
Technical analysis support
Activities
Functional document aimed at supporting the actual implementation
Deliverables
28?Sources
The long list of indicators is formed based on ”groups” which differ in information source used
Le informazioni pubbliche comprendono, infatti, anche le informazioni derivanti da notizie stampa e / o su piazza
Riproduzione vietata
29?Indicators Exceeding the granted limit (from Credit Bureau)
E X A M P L E
Exceeding the granted limit (from Credit Bureau)
When the Customer is in the situation of exceeding a limit granted by a competitor Bank
Trend of usage for overdrafts
When the Customer tends to use the overdraft often close to its limit
Persistent breaches
Exceeding a certain ‘%’ of the overdraft limit for more than a ‘X’number of days:
- [a] number of days
[b] ‘%’ of excess of the overdraft limit
“Light” breach – if [a] < 90 days AND [b] <= 3%
“Medium” breach – if [a] > 90 days OR [b] > 3%
“Heavy” breach – if [a] > 90 days AND [b] > 3%
30?The IT solutions - Design
All operations in the credit monitoring are based on an iterative process, starting from the reports of early warning or recognition of specific anomalies. They are divided into the following (theoretical) phases:
Early Warning report / Breaches detection
E X A M P L E
1?2
Analysis and definition proposal of the handling class, action plan and provisions
3
Decision about of the handling class, action plan and provisions
Actions execution and continuous monitoring
4
31?The IT solution – a functional example
Monitoring external tools
Segment definition
(Retail, SME, Corporate)
Decision process
Other tools
E X A M P L E
Early Warning
Credit Monitoring System
Integration with other internal procedures (Core Banking, CRM, …)
Actions
Regular
Analysis and definition
Monitor
Performing
Classification
Action plan
Provision
Initial past due iniziale
Past due 30
Ordinary handling
Branch, Regular Credits
Branch
Link to other procedures
(Core Banking, CRM, …)
Past due 90 / 180
Anomalies
Special Credits
Branch, Non-regular Credits
Branch
Default
Breaches
Restructured
Special Credits
Non-regular Credits
Non-regular Credits
Charge off
32?The IT solutions – Laboratory
The presence of a laboratory environment, where basic information, indicators, suggested actions, timing and methods of application and quantitative results are historically tracked in order to monitor and correct the Early Warning System, is a pre-condition for the evolution of the system itself.
Actions execution and continuous monitoring
Analysis and definition proposal of the handling class, action plan and provisions
1
Early Warning report / Breaches detection
2?3
Decision about of the handling class, action plan and provisions
4
EW Data Mart
Info out of the process but potentially useful
Performance measurement
Simulation
New system
Efficacia
Rapidità
# di iterazioni
Riduzione dell’esposizione
Prediction capacity
Provisions/
corrections
Effectiveness
Benchmarking
Riproduzione vietatae
Expected Credit Los
Due to the pandemic, it might become too burdensome or difficult for banks to determine the extent and adequacy of collaterals available with them and the subsequent provisioning. There may be additional disclosures required in the financial statements and the computation of capital adequacy for COVID-19.
Given the situation of the lock down in the world, the defaults may increase substantially as many companies would have lost revenue for a long time. An increase in defaults is likely to cause issues in liquidity and capital adequacy.
Banks would therefore be required to maintain robust risk management functions and track their borrowers individually to determine and segregate the permanent impact from the temporary impact and make appropriate provisions & revisit their hedging strategies.
The new impairment provision becomes applicable in times of high NPA levels and stressed asset situation experienced in the banking sector. The new impairment provision would require both financial services entities and the regulator to take a closer look at the impact on capital planning, pricing and alignment to risk management.
The ECL norms are likely to result in enhanced provisions given that they apply to off balance sheet items such as loan commitments and financial guarantees also.
The introduction of the forward-looking ECL model aligns the provision on financial assets consistent with their economic value and is more proactive during an economic downturn. However, the three stage credit loss recognition that requires advanced credit risk modelling skills and high quality data, poses a new challenge to many banks.
Financial assets are classified and measured on the following basis:
? Amortized Cost (AC);
?Fair Value Through Other Comprehensive Income (FVOCI)
?Fair Value Through Profit and Loss account (FVTPL)
Approaches for computation
The objective of impairment requirements is to recognize lifetime ECLs for all financial instruments for which there has been a significant increase in credit risk since origination. The assets which have not undergone any significant deterioration shall be recognized with only 12-month ECLs.
Segmentation
Portfolio segmentation approach will help banks in generating synergies both in the short term and the long term.
Banks can segment their portfolios into:
?Corporate loans (term loans, overdrafts, working capitalloans, LC renance loans)
?Retail loans (consumer, mortgage, vehicle and credit card)
?Agriculture loans (Kharif and Rabi crops)
?Investments (bank, sovereign and corporate)
?International banking division (loans to corporates overseas/other than domestic countries)
?Loans to banks and sovereigns
The retail portfolio shall be segmented by product types or pooled based on various individual and behavioral characteristics. The possible segments or the retail (mortgage, vehicle, credit card and consumer loan) portfolio of a bank can be listed as below:
?Salaried/non-salaried/self employed
?Public/private sector employee
?Income group of the borrower
?Collateral coverage ratio of the facility
ECL:
ECL on financial assets is an unbiased probability weighted amount based out of possible outcomes after considering risk of credit loss even if probability is low. ECL can be defined as the difference between cash flows due under the contract and cash flows that an entity expects to receive. The ECL formula can be defined as following:
?Marginal Probability of default (MPD)
?Loss given default (LGD)
?Exposure at default (EAD)
?Discount factor (D)
Probability of default (PD)
PD is defined as the probability of whether borrowers will default on their obligations in the future. Historical PD derived from a bank’s internal credit rating data has to be calibrated with forward-looking macroeconomic factors to determine the PD term structure.
The forward-looking PD shall reflect the entities’ current view of the future and should be an unbiased estimate as it should not include any conservatism or optimism. The following list of methodologies can be used to generate forward-looking PD term structures:
?Markov chain model
?Parametric survival regression (Weibull model)
?Vasicek single factor model
?Forward intensity model on distance-to-default approach (public-listed firms)
?Pluto Tasche PD model (low/no default portfolio)
Loss given default (LGD)
LGD is an estimate of the loss from a transaction given that a default occurs different future periods.
LGD is one of the key components of the credit risk parameters based ECL model. In the context of lifetime ECL calculation, an LGD estimate has to be available for all periods that are part of the lifetime horizon (and not only for the case of a default within the next 12 months asunder Basel II).The LGD component of ECL is independent of deterioration of asset quality, and thus applied uniformly across various stages. The following methodologies are widely used to estimate LGD:
?Workout LGD
?Market LGD
?Asset pricing model/Implied market LGD
?Market-based model.
Exposure at default (EAD)
EAD is one of the key components for ECL computation.?
EAD can be seen as an estimation of the extent to which the financial entity may be exposed to a counterparty in the event of a default and at the time of the counterparty’s default.
EAD modelling would require the ALM system of the bank to produce either contractual or behavioural cash flows till the lifetime of the loans.
Expected prepayments
EAD shall also be modelled based on historical prepayments and establishing relationships with a change in interest rates to forecast the prepayment factors in order to estimate the expected payments in future scenarios.
Funded exposures
For the funded/single drawdown exposures, the EAD modelling might not pose a challenge as compared to non-funded facilities.
EAD = Drawn line + credit conversion factor * undrawn
Difference between IAS 39 and IFRS 9s