IFRS 9 and CECL Credit Risk Modelling and Validation covers a hot topic in risk management. Both IFRS 9 and CECL accounting standards require Banks to adopt a new perspective in assessing Expected Credit Losses. The book explores a wide range of models and corresponding validation procedures. The most traditional regression analyses pave the way to more innovative methods like machine learning, survival analysis, and competing risk modelling. Special attention is then devoted to scarce data and low default portfolios. A practical approach inspires the learning journey. In each section the theoretical dissertation is accompanied by Examples and Case Studies worked in R and SAS, the most widely used software packages used by practitioners in Credit Risk Management.
Author(s): Tiziano Bellini
Edition: 1
Publisher: Academic Press
Year: 2019
Language: English
Pages: 306
Cover......Page 1
IFRS 9 and CECL Credit
Risk Modelling and
Validation:
A Practical Guide with Examples
Worked in R and SAS
......Page 4
Copyright
......Page 5
Dedication
......Page 6
Tiziano Bellini's Biography......Page 7
Preface......Page 8
Acknowledgements......Page 11
Key Abbreviations and Symbols......Page 12
1.1 Introduction......Page 13
1.2 IFRS 9......Page 14
1.2.1 Staging Allocation......Page 15
1.2.2 ECL Ingredients......Page 17
1.2.3 Scenario Analysis and ECL......Page 18
1.3 CECL......Page 19
1.3.1 Loss-Rate Methods......Page 20
1.3.2 Vintage Methods......Page 22
1.3.3 Discounted Cash Flow Methods......Page 24
1.4 ECL and Capital Requirements......Page 25
1.4.1 Internal Rating-Based Credit Risk-Weighted Assets......Page 28
1.4.2 How ECL Affects Regulatory Capital and Ratios......Page 32
1.5 Book Structure at a Glance......Page 37
References......Page 41
2.1 Introduction......Page 42
2.2.1 Default Definition......Page 44
2.2.2 Data Preparation......Page 46
2.3 Generalised Linear Models (GLMs)......Page 48
2.3.1 GLM (Scorecard) Development......Page 49
2.3.2 GLM Calibration......Page 56
2.3.3 GLM Validation......Page 59
2.4 Machine Learning (ML) Modelling......Page 64
2.4.1 Classification and Regression Trees (CART)......Page 65
2.4.2 Bagging, Random Forest, and Boosting......Page 69
2.4.4 ML Model Validation......Page 74
2.5 Low Default Portfolio, Market-Based, and Scarce Data Modelling......Page 77
2.5.1 Low Default Portfolio Modelling......Page 78
2.5.2 Market-Based Modelling......Page 81
2.5.3 Scarce Data Modelling......Page 85
2.6 SAS Laboratory......Page 86
2.7 Summary......Page 90
2.8 Appendix A. From Linear Regression to GLM......Page 91
2.9 Appendix B. Discriminatory Power Assessment......Page 96
Exercises......Page 99
References......Page 100
Key Abbreviations and Symbols......Page 101
3.1 Introduction......Page 102
3.2.1 Default Flag Creation......Page 103
3.2.2 Account-Level (Panel) Database Structure......Page 105
3.3 Lifetime GLM Framework......Page 108
3.3.1 Portfolio-Level GLM Analysis......Page 109
3.3.2 Account-Level GLM Analysis......Page 115
3.3.3 Lifetime GLM Validation......Page 118
3.4.1 KM Survival Analysis......Page 122
3.4.2 CPH survival analysis......Page 126
3.4.3 AFT Survival Analysis......Page 131
3.4.4 Survival Model Validation......Page 132
3.5.1 Bagging, Random Forest, and Boosting Lifetime PD......Page 136
3.5.2 Random Survival Forest Lifetime PD......Page 139
3.5.3 Lifetime ML Validation......Page 143
3.6.1 Naïve Markov Chain Modelling......Page 144
3.6.2 Merton-Like Transition Modelling......Page 149
3.6.4 Transition Matrix Model Validation......Page 153
3.7 SAS Laboratory......Page 154
3.8 Summary......Page 157
Appendix A. Portfolio-Level PD Shift......Page 158
Appendix B. Account-Level PD Shift......Page 161
References......Page 162
Key Abbreviations and Symbols......Page 164
4.1 Introduction......Page 165
4.2 LGD Data Preparation......Page 166
4.2.1 LGD Data Conceptual Characteristics......Page 167
4.2.2 LGD Database Elements......Page 169
4.3 LGD Micro-Structure Approach......Page 170
4.3.1 Probability of Cure......Page 173
4.3.2 Severity......Page 176
4.3.3 Defaulted Asset LGD......Page 179
4.3.4 Forward-Looking Micro-Structure LGD Modelling......Page 182
4.3.5 Micro-Structure Real Estate LGD Modelling......Page 186
4.3.6 Micro-Structure LGD Validation......Page 187
4.4.1 Tobit Regression......Page 191
4.4.2 Beta Regression......Page 195
4.4.3 Mixture Models and Forward-Looking Regression......Page 198
4.5 LGD Machine Learning (ML) Modelling......Page 199
4.5.1 Regression Tree LGD......Page 200
4.5.2 Bagging, Random Forest, and Boosting LGD......Page 203
4.5.3 Forward-Looking Machine Learning LGD......Page 208
4.5.4 Machine Learning LGD Validation......Page 210
4.6 Hints on LGD Survival Analysis......Page 212
4.7.1 Expert Judgement LGD Process......Page 213
4.7.2 Low Default Portfolio LGD......Page 216
4.8 SAS Laboratory......Page 217
4.9 Summary......Page 221
References......Page 222
Key Abbreviations and Symbols......Page 223
5.1 Introduction......Page 224
5.2.1 How to Organize Data......Page 225
5.3 Full Prepayment Modelling......Page 227
5.3.1 Full Prepayment via GLM......Page 228
5.3.2 Machine Learning (ML) Full Prepayment Modelling......Page 232
5.3.3 Hints on Survival Analysis......Page 235
5.3.4 Full Prepayment Model Validation......Page 236
5.4 Competing Risk Modelling......Page 239
5.4.1 Multinomial Regression Competing Risks Modelling......Page 240
5.4.2 Full Evaluation Procedure......Page 247
5.4.3 Competing Risk Model Validation......Page 251
5.5 EAD Modelling......Page 253
5.5.1 A Competing-Risk-Like EAD Framework......Page 254
5.5.2 Hints on EAD Estimation via Machine Learning (ML)......Page 259
5.6 SAS Laboratory......Page 260
Appendix. Average Closure Rate Shortcut......Page 262
Exercises......Page 263
References......Page 264
6.1 Introduction......Page 265
6.2.1 Vector Auto-Regression and Vector Error-Correction Modelling......Page 267
6.2.2 VAR and VEC Forecast......Page 274
6.2.3 Hints on GVAR Modelling......Page 277
6.3.1 Scenario Design and Satellite Models......Page 278
6.3.2 Lifetime ECL......Page 281
6.3.3 IFRS 9 Staging Allocation......Page 285
6.4 ECL Validation......Page 288
6.4.1 Historical and Forward-Looking Validation......Page 290
6.4.2 Credit Portfolio Modelling and ECL Estimation......Page 291
6.5 SAS Laboratory......Page 298
Suggestions for Further Reading......Page 299
References......Page 300
Index......Page 302
Back Cover......Page 306
