IFRS 9 ECL Modeling Using Python: PD, LGD & EAD from Scratch

Name: IFRS 9 ECL Modeling Using Python: PD, LGD & EAD from Scratch
Rating: 3.8 (17 reviews)

Learn IFRS 9 ECL modeling using Python with hands-on PD, LGD, and EAD model development and validation

Created bySubhashish Ray

Last updated 4/2026

English

English [Auto],

What you'll learn

Understand the IFRS 9 Expected Credit Loss (ECL) framework and its key components - PD, LGD, and EAD
Develop Probability of Default (PD) models using Python with real-world simulated data
Build and validate Loss Given Default (LGD) and Exposure at Default (EAD) models step-by-step
Apply staging logic and compute 12-month and lifetime ECL in line with IFRS 9 requirements
Incorporate macroeconomic variables for forward-looking scenario adjustments
Automate ECL computation, aggregation, and reporting using Python libraries
Translate regulatory and accounting concepts into practical, data-driven model solutions
Gain confidence to design, document, and explain IFRS 9 ECL models to auditors and regulators
Understand the transition from IAS 39 to IFRS 9 and how ECL improves financial risk recognition
Perform data cleaning, preprocessing, and feature engineering for credit risk modeling using Python
Conduct model validation using performance metrics such as AUC, KS, Gini, Brier Score, and Concordance

Course content

10 sections • 41 lectures • 7h 34m total length

Introduction3:46

Loan Loss Distribution6:11
IAS 39 vs IFRS 9: Incurred Loss vs Expected Loss7:53
What is Expected Credit Loss (ECL)?6:20
Learn how IFRS 9 defines expected credit loss as the discounted value of expected cash shortfalls over a financial instrument’s life, incorporating PD, LGD, and EAD.
Components of Expected Credit Loss (PD, LGD & EAD)10:20
Explore how IFRS 9 decomposes expected credit loss into PD, LGD, and EAD, and how each component reflects default frequency, loss severity, and exposure at default under macroeconomic scenarios.
Default Definition under IFRS 97:41
12-Month ECL vs Lifetime ECL under IFRS 98:17
Explain the IFRS 9 distinction between 12-month and lifetime ECL, and how stage 1 uses 12-month ECL while stages 2 and 3 use lifetime ECL.
Forward-Looking Information and Macroeconomics under IFRS 99:03
Learn how IFRS 9 integrates forward-looking information and macroeconomic scenarios to estimate expected credit losses, using probability-weighted outcomes across baseline, optimistic, and adverse paths, with governance and staging considerations.
IFRS 9 Staging Framework: Stage 1, Stage 2 and Stage 314:01
Significant Increase in Credit Risk (SICR) under IFRS 910:24
End of Section Quiz: Conceptual Foundations

Introduction to the Modeling Dataset17:25
Explore the practical setup for IFRS 9 ecl modeling by building a pd model in python using account master and monthly performance datasets to examine structure, variables, and data splits.
Portfolio Segmentation and Modeling Dataset Preparation11:36
Segment portfolios by credit card products, merge borrower and behavioral data on account id, and prepare the pd modeling dataset with default_flag as the target.
Data Quality Checks and Exploratory Analysis24:24
Default Definition and Modeling Horizon9:31
Define default and set the modeling horizon for IFRS 9 PD modeling, using 90 days past due as the delinquency-based default, with observation and performance windows and a forward-looking target.
Target Variable Construction (Forward-Looking Default)10:13
Train, Validation and Out-of-Time Split8:52
Feature Engineering, Variable Selection & Multicollinearity16:45
Weight of Evidence (WOE) Transformation11:46
Explore how weight of evidence transformation converts variables into a stable, interpretable form for credit risk using logistic regression. The demo covers quantile binning, WOE, IV, and monotonicity.
Logistic Regression PD Model Development19:45
Build a logistic regression model to estimate the 12-month probability of default (PD), covering training data prep, median imputation, model fitting, PD prediction, and coefficient interpretation.
Model Validation (AUC, ROC, KS)15:15
PD Calibration14:47
End of Section Quiz: PD Modeling

Lifetime PD: Concept and Importance4:36
Monthly PD, Marginal PD & Cumulative PD5:44
Build a lifetime PD term structure by moving from 12-month to monthly PD. Derive marginal and cumulative PD, using survival probabilities to assess IFRS 9 ECL with LGD and EAD.
Lifetime PD Curve Construction and Term Structure in Python8:54
Transform the 12-month probability of default into a lifetime PD curve and term structure under IFRS 9, deriving forward, marginal, and cumulative PD for stage 2/3 ECL.
End of Section Quiz: Lifetime PD Modeling and Term Structure

LGD Concept and IFRS 9 Perspective7:24
Explore loss given default under IFRS 9 and its role in the ECL formula (PD × LGD × EAD), including recovery timing, discounting, workout LGD, and secured versus unsecured loans.
Workout LGD: Recovery Cash Flow Modeling and Discounting15:28
Learn to model workout LGD by constructing and discounting recovery cash flows in Python, converting them to present value, and aggregating by account to support IFRS 9 ECL calculations.
LGD Calculation using Workout Method12:31
LGD Validation and Analysis13:28
End of Section Quiz: LGD Modeling

Implementing IFRS 9 Staging: From Concept to Practice3:16
Derive IFRS 9 staging from data by assigning stage 3 to defaults, stage 2 to accounts with days past due, and stage 1 to performing exposures, with rule order preserved.
Applying IFRS 9 Staging Rules13:47
Implement IFRS 9 staging in python by assigning stage one by default, stage three to defaulted accounts, and stage two for delinquent accounts, validate with days past due and EAD.
Stage Migration Analysis and Transition Matrix15:20
Analyze stage migration with a transition matrix to track how accounts move between stages. Leverage IFRS 9 concepts, PD, LGD, and EAD to interpret transitions and exposure across the portfolio.
End of Section Quiz: IFRS9 Staging

Requirements

Basic understanding of banking, credit risk, or finance concepts (helpful but not mandatory)
Familiarity with Python programming fundamentals (variables, dataframes, basic libraries)
Some exposure to statistics or data analysis concepts such as regression, correlation, and probability
Enthusiasm to learn credit risk modeling and apply IFRS 9 concepts practically
A computer with Python installed (Jupyter Notebook, Google Colab, or any IDE)
No prior experience with IFRS 9 or advanced coding is required , all concepts are explained step-by-step

Description

This course provides a complete, hands-on guide to developing IFRS 9 Expected Credit Loss (ECL) models in Python from scratch. It is designed for banking and finance professionals, credit risk analysts, auditors, data scientists, and students who want to understand and implement credit risk modeling based on the IFRS 9 accounting standard.

You will learn how to design, build, and validate all three components of the ECL framework: Probability of Default (PD), Loss Given Default (LGD), and Exposure at Default (EAD). The course also explains important IFRS 9 concepts such as staging, Significant Increase in Credit Risk (SICR) assessment, lifetime versus 12-month ECL, discounting, and the use of macroeconomic scenarios for forward-looking adjustments.

Each section combines theory and practical Python implementation. PD modeling will focus on logistic regression and related validation metrics. LGD modeling will include linear, Beta, and Tobit regression methods, while EAD modeling will cover linear and logistic approaches for Credit Conversion Factor (CCF) estimation. You will use Python libraries such as pandas, numpy, statsmodels, and scikit-learn for data cleaning, variable selection, model development, and validation. You will also explore visualization and evaluation metrics to ensure that each model meets regulatory expectations.

By the end of the course, you will be able to design and implement an end-to-end IFRS 9 ECL model for retail or wholesale portfolios. You will understand how to connect PD, LGD, and EAD models to calculate total ECL and how to interpret results from both accounting and risk management perspectives.

No advanced programming experience is required. A basic understanding of finance, risk, or statistics will help you get the most out of the course.

Who this course is for:

Credit Risk Analysts, Modelers, and Risk Managers
Finance and Banking Professionals
Data Scientists, Machine Learning Practitioners, and Python Developers
Students and Researchers in Finance, Economics, or Quantitative Fields
Auditors, Compliance Officers, and Regulatory Professionals
Aspiring Risk Modelers or Python Learners

IFRS 9 ECL Modeling Using Python: PD, LGD & EAD from Scratch

What you'll learn

Explore related topics

Course content

Introduction1 lecture • 4min

Conceptual Foundations9 lectures • 1hr 20min

Portfolio and Data Understanding3 lectures • 31min

Probability of Default (PD) Modeling11 lectures • 2hr 40min

Lifetime PD Modeling and Term Structure3 lectures • 19min

Loss Given Default (LGD) Modeling under IFRS 94 lectures • 49min

Exposure at Default (EAD) Modeling under IFRS 94 lectures • 47min

IFRS 9 Staging and Credit Risk Assessment3 lectures • 32min

Macroeconomic Adjustment and Forward-Looking ECL2 lectures • 24min

Final Expected Credit Loss (ECL) Calculation and Portfolio Analysis1 lecture • 9min

Requirements

Description

Who this course is for: