IFRS 9 and CECL Credit Risk Modelling and Validation

Publishing Date - January 2019

Table of Contents

1 Introduction to Expected Credit Loss Modelling and Validation

1.1 Introduction 

1.2 IFRS 9 

   1.2.1 Staging Allocation 

   1.2.2 ECL Ingredients 

   1.2.3 Scenario Analysis and ECL 

1.3 CECL 

   1.3.1 Loss-Rate Methods 

   1.3.2 Vintage Methods 

   1.3.3 Discounted Cash Flow Methods 

   1.3.4 Probability of Default Method (PD, LGD, EAD) 

   1.3.5 IFRS 9 vs. CECL 

1.4 ECL and Capital Requirements 

   1.4.1 Internal Rating-Based Credit Risk-Weighted Assets 

   1.4.2 How ECL A_ects Regulatory Capital and Ratios 

1.5 Book Structure at a Glance 

1.6 Summary 

2 One-Year PD

2.1 Introduction 

2.2 Default De_nition and Data Preparation 

   2.2.1 Default De_nition 

   2.2.2 Data Preparation 

2.3 Generalized Linear Models (GLMs) 

   2.3.1 GLM (Scorecard) Development 

   2.3.2 GLM Calibration 

   2.3.3 GLM Validation 

2.4 Machine Learning (ML) Modelling 

   2.4.1 Classi_cation and Regression Trees (CART) 

   2.4.2 Bagging, Random Forest, and Boosting 

   2.4.3 ML Model Calibration 

   2.4.4 ML Model Validation 

2.5 Low Default Portfolio, Market-Based, and Scarce Data Modelling

   2.5.1 Low Default Portfolio Modelling 

   2.5.2 Market Based Modelling 

   2.5.3 Scarce Data Modelling 

   2.5.4 Hints on Low Default Portfolio, Market-Based, and Scarce Data Model Validation 

2.6 SAS Laboratory 

2.7 Summary 

2.8 Appendix A. From Linear Regression to GLMs 

2.9 Appendix B. Discriminatory Power Assessment

3 Lifetime PD 

3.1 Introduction 

3.2 Data Preparation 

   3.2.1 Default Flag Creation 

   3.2.2 Account-Level (Panel) Database Structure 

3.3 Lifetime GLM Framework 

   3.3.1 Portfolio-level GLM Analysis 

   3.3.2 Account-Level GLM Analysis 

   3.3.3 Lifetime GLM Validation 

3.4 Survival Modelling 

   3.4.1 Kaplan Meier (KM) Survival Analysis

   3.4.2 Cox Proportional Hazard (CPH) Survival Analysis 

   3.4.3 Accelerated Failure Time (AFT) Survival Analysis

   3.4.4 Survival Model Validation 

3.5 Lifetime Machine Learning (ML) Modelling

   3.5.1 Bagging, Random Forest, and Boosting Lifetime PD 

   3.5.2 Random Survival Forest Lifetime PD

   3.5.3 Lifetime ML Validation

3.6 Transition Matrix Modelling

   3.6.1 Naive Markov Chain Modelling

   3.6.2 Merton-Like Transition Modelling

   3.6.3 Multi State Modelling 

   3.6.4 Transition Matrix Model Validation

3.7 SAS Laboratory 

3.8 Summary

4 LGD Modelling

4.1 Introduction 

4.2 LGD Data Preparation 

   4.2.1 LGD Data Conceptual Characteristics

   4.2.2 LGD Database Elements 

   4.3 LGD Micro-Structure Approach

   4.3.1 Probability of Cure 

   4.3.2 Severity 

   4.3.3 Defaulted Asset LGD

   4.3.4 Forward-Looking Micro-Structure LGD Modelling

   4.3.5 Micro-Structure Real Estate LGD Modelling 

   4.3.6 Micro-Structure LGD Validation 

4.4 LGD Regression Methods 

   4.4.1 Tobit Regression 

   4.4.2 Beta Regression 

   4.4.3 Mixture Models and forward-looking Regression 

   4.4.4 Regression LGD Validation 

4.5 LGD Machine Learning (ML) Modelling

   4.5.1 Regression Tree LGD 

   4.5.2 Bagging, Random Forest, and Boosting LGD

   4.5.3 Forward-Looking Machine Learning LGD 

   4.5.4 Machine Learning LGD Validation 

4.6 Hints on LGD Survival Analysis

4.7 Scarce Data and Low Default Portfolio LGD Modelling

   4.7.1 Expert Judgement LGD Process 

   4.7.2 Low Default Portfolio LGD

   4.7.3 Hints on How to Validate Scarce Data and Low Default Portfolio LGDs 

4.8 SAS Laboratory 

4.9 Summary

5 Prepayments, Competing Risks and EAD Modelling

5.1 Introduction 

5.2 Data Preparation

   5.2.1 How to Organize Data

5.3 Full Prepayment Modelling

   5.3.1 Full Prepayment via GLMs 

   5.3.2 Machine Learning (ML) Full Prepayment Modelling 

   5.3.3 Hints on Survival Analysis 

   5.3.4 Full Prepayment Model Validation 

5.4 Competing Risk Modelling 

   5.4.1 Multinomial Regression Competing Risks Modelling 

   5.4.2 Full Evaluation Procedure 

   5.4.3 Competing Risk Model Validation

5.5 EAD Modelling 

   5.5.1 A Competing-Risk-Like EAD Framework

   5.5.2 Hints on EAD Estimation via Machine Learning (ML) 

   5.5.3 EAD Model Validation 

5.6 SAS Laboratory 

5.7 Summary 

6 Scenario Analysis and Expected Credit Losses

6.1 Introduction

6.2 Scenario Analysis 

   6.2.1 Vector Auto-Regression (VAR) and Vector Error-Correction (VEC) Modelling

   6.2.2 VAR and VEC Forecast 

   6.2.3 Hints on GVAR Modelling 

   6.3 ECL Computation in Practice 

   6.3.1 Scenario Design and Satellite Models 

   6.3.2 Lifetime ECL 

   6.3.3 IFRS 9 Staging Allocation 

6.4 ECL Validation 

   6.4.1 Historical and Forward-Looking Validation 

   6.4.2 Credit Portfolio Modelling and ECL Estimation 

6.5 SAS Laboratory 

6.6 Summary 

This site was designed with the
.com
website builder. Create your website today.
Start Now