Overview
Expected Credit Loss Modelling for IFRS 9 and ASC 326
Shasat provides specialised ECL and CECL modelling solutions for financial institutions, ensuring compliance with IFRS 9's Expected Credit Loss approach and the Current Expected Credit Loss framework under US GAAP ASC 326. Our expertise combines advanced quantitative modelling, Python-based implementation and deep regulatory knowledge to deliver accurate, audit-ready impairment models.
From initial gap analysis and data quality assessment to full model build, ongoing calibration and post-implementation review, Shasat provides end-to-end support tailored to the specific portfolio, product mix and regulatory environment of each institution. Our Python-based models ensure full transparency, reproducibility and alignment with regulatory and audit expectations.
IFRS 9 impairment framework
The Three-Stage ECL Model
IFRS 9 requires a three-stage approach to recognising expected credit losses, based on changes in credit risk since initial recognition. Shasat assists institutions with defining and validating staging criteria, including both quantitative thresholds and qualitative indicators.
Stage 1
No Significant Increase in Credit Risk
12-Month ECL
Assets performing as expected since origination. Interest revenue recognised on the gross carrying amount. 12-month ECL represents losses expected from default events within the next 12 months.
Stage 2
Significant Increase in Credit Risk
Lifetime ECL
Assets where credit risk has increased significantly since initial recognition. Lifetime ECL reflects all possible default events over the expected life of the instrument. Interest on gross carrying amount.
Stage 3
Credit Impaired
Lifetime ECL
Assets that are credit impaired. Lifetime ECL applies. Interest revenue is recognised on the net carrying amount (gross carrying amount less loss allowance) rather than the gross amount.
Model components
PD, LGD and EAD — The Three ECL Inputs
ECL is mathematically defined as the product of three key risk parameters. Shasat models each component independently and in combination, using advanced statistical techniques calibrated to portfolio-specific data and forward-looking macroeconomic scenarios.
PD
Probability of Default
The likelihood that a borrower will default within a specified period. For Stage 1, the 12-month PD is used. For Stages 2 and 3, the lifetime PD is required. Shasat develops through-the-cycle and point-in-time PD models using regression analysis, survival analysis and machine learning techniques.
LGD
Loss Given Default
The proportion of the exposure that is expected to be lost given that default has occurred. LGD is influenced by collateral type and value, seniority of the exposure, recovery rates and workout costs. Shasat models downturn LGD to reflect stressed market conditions as required by IFRS 9.
EAD
Exposure at Default
The expected outstanding balance at the point of default, including drawn amounts and undrawn commitments. For revolving facilities and credit cards, EAD modelling incorporates Credit Conversion Factors (CCF) to capture the likelihood of additional drawdown before default.
ECL Calculation — Core Formula
ECL = PD × LGD × EAD × Discount Factor
Applied at each reporting date, incorporating forward-looking macroeconomic scenarios and probability weightings across scenarios
IFRS 9 ECL implementation
Our Implementation Approach
Shasat's IFRS 9 ECL implementation is structured to address every stage of the model lifecycle, from initial data review through to ongoing calibration and post-implementation review.
1
Data Quality Assessment
We evaluate data accuracy, completeness and consistency, focusing on rating systems, credit risk methodologies and historical default data required for PD, LGD and EAD estimation.
2
Multiple Quantitative Models
We build ECL models tailored to portfolio product types — retail, corporate, SME, sovereign and structured — using matrices, Monte Carlo simulations, stochastic processes and regression analysis.
3
Scoring Models
We develop and implement advanced credit scoring models to assess credit risk, optimise decision-making at origination and ongoing review, and enhance predictive accuracy across portfolio segments.
4
Impairment Methodology Review
We assess PD, LGD and EAD estimation methodologies, staging criteria, the definition of default, significant increase in credit risk thresholds, and alignment with the institution's credit risk policies.
5
Scenario Development
We create forward-looking macroeconomic scenarios (base, upside and downside) and estimate probability-weighted PDs and LGDs under each scenario, incorporating GDP, unemployment, interest rates and sector-specific variables.
6
Model Development and Validation
We design and validate ECL and CECL models aligned with current regulations and best practices. Python-based implementation ensures full code transparency, auditability and version control.
7
Testing and Calibration
We run back-testing scenarios against historical loss experience and continuously calibrate models to maintain accuracy and relevance as market conditions and portfolio composition change.
8
Disclosure and Compliance Support
We define disclosure requirements for transparent IFRS 7 reporting of credit risk and impairment, including qualitative descriptions of ECL methodology, quantitative reconciliations and sensitivity analysis.
Standards comparison
ECL (IFRS 9) versus CECL (ASC 326)
While both models are forward-looking impairment frameworks, IFRS 9 ECL and ASC 326 CECL differ in several important respects. Shasat supports dual reporters and institutions transitioning between the two frameworks.
Feature
ECL — IFRS 9
CECL — ASC 326
Standard
IFRS 9 (IASB) — applicable to IFRS reporters globally
ASC 326 (FASB) — applicable to US GAAP reporters
Staging
Three-stage model — 12-month ECL (Stage 1) or Lifetime ECL (Stages 2 and 3)
No staging — lifetime expected credit losses recognised at initial recognition for all assets
Loss horizon
12-month ECL at origination for low-risk assets; lifetime on significant deterioration
Lifetime ECL from origination for all in-scope financial assets
Conservatism
Moderate — reflects risk profile and staging
Higher at origination — full lifetime losses recognised immediately
Scope
Financial assets at amortised cost, FVOCI debt, lease receivables, trade receivables, contract assets and loan commitments
Broader scope — also covers held-to-maturity debt securities and purchased credit-deteriorated assets
Methodology
PD x LGD x EAD discounted at EIR, with probability-weighted macroeconomic scenarios
Flexible — DCF, loss-rate, roll-rate, vintage analysis and other methods permitted
Technology and tools
Python-Based ECL Implementation
Shasat builds ECL and CECL models in Python, ensuring full code transparency, reproducibility and ease of audit review. Models integrate directly with institutions' data infrastructure and risk systems.
Python Core language
Pandas and NumPy Data processing
Scikit-learn ML models
SciPy Statistical analysis
Monte Carlo simulation Stochastic
Regression analysis PD and LGD
Survival analysis Lifetime PD
Time series modelling Macro scenarios
Back-testing frameworks Model validation
FAQs
Common Questions
Expected Credit Loss (ECL) under IFRS 9 is a forward-looking impairment methodology requiring financial institutions to recognise credit losses based on expected future defaults rather than waiting for a loss event to occur (as under the incurred loss model of IAS 39). ECL is calculated as the product of Probability of Default (PD), Loss Given Default (LGD) and Exposure at Default (EAD), discounted at the effective interest rate. For Stage 1 assets, 12-month ECL is recognised; for Stage 2 and Stage 3 assets, lifetime ECL applies.
Both ECL (IFRS 9) and CECL (ASC 326) are forward-looking credit impairment models, but they differ materially in scope and timing. IFRS 9 uses a three-stage approach where 12-month ECL is recognised at initial recognition for performing assets and transfers to lifetime ECL on significant deterioration. ASC 326 CECL requires lifetime expected credit losses to be recognised at initial recognition for all in-scope financial assets with no equivalent staging mechanism. CECL is generally considered more conservative at origination because it pulls forward the full lifetime loss estimate immediately.
Shasat uses advanced mathematical and statistical techniques including matrix methods, Monte Carlo simulations, stochastic processes, regression analysis, survival analysis and machine learning approaches. Models are built and implemented in Python, ensuring transparency, reproducibility and alignment with regulatory expectations. Techniques are selected based on portfolio type, data availability and the institution's specific risk profile and system infrastructure.
IFRS 9 classifies financial assets into three stages based on credit risk at the reporting date relative to initial recognition. Stage 1 applies where there is no significant increase in credit risk (SICR) — 12-month ECL. Stage 2 applies where SICR has occurred — lifetime ECL. Stage 3 covers credit-impaired assets — lifetime ECL and interest on net carrying amount. SICR is assessed using both quantitative thresholds (change in PD, rating migration) and qualitative indicators (watchlist status, restructuring, payment patterns). Shasat assists institutions with developing, calibrating and documenting robust staging criteria.
Yes. Shasat provides post-implementation reviews and ongoing model optimisation to address discrepancies identified after go-live, adapt models to changing market conditions, incorporate new macroeconomic data, and respond to regulatory or audit feedback. Services include back-testing against realised losses, model recalibration, scenario updating and preparation for internal audit or prudential regulatory review.
Also from Shasat
Related Valuation Services
IFRS 9 Consulting
Classification, measurement, impairment and hedge accounting implementation
Model and Data Validation
Independent review and challenger model build for credit and pricing models
Credit Risk Valuation
CVA, DVA, FVA and XVA for OTC derivatives under IFRS 13
Financial Instrument Valuation
Derivatives, bonds and convertible loans — IFRS 13 and ASC 820