Publicação
Methodology for the analysis of rating migrations within Probability of Default (PD) models
| Resumo: | In credit risk management, Internal Ratings-Based (IRB) models, particularly Probability of Default (PD) models, are used to estimate the likelihood that an obligor will default within a specified time horizon. A key aspect of credit risk management consists on monitoring the stability of the portfolios underlying these PD models, in order to identify changes in borrower’s behaviour and, credit quality, and the overall model performance. These changes may arise from evolving economic conditions, modifications in lending standards, or issues related to data quality. In addition, the examination of rating migrations provides valuable insights into customer transition dynamics, detecting patterns associated with credit improvement or deterioration. The objective of this work is to develop a robust quantitative methodology for the analysis of rating migrations within PD models. This methodology intends to support the systematic assessment of rating migrations over time, effectively capturing their magnitude, frequency, and direction, while also examining shifts in rating distributions, in order to establish appropriate criteria, e.g. via thresholds. These patterns were analysed across multiple models during different macroeconomic conditions and recalibration processes. To achieve this, several metrics were tested, including the Population Stability Index (PSI), Migration Matrix, Matrix Weighted Bandwidth (MWB), Average of Adjustments, Speed and Direction. Additionally, a deeper analysis was conducted to model migration behaviour, providing information into the dynamic characteristics of the models under study. The results indicate that, from the proposed metrics, the PSI, MWB and Average of Adjustments are the ones that exhibit a greater sensitivity to each model characteristics, macroeconomic conditions and recalibration processes. Collectively, these metrics deliver essential insights into the dynamics of risk grade migrations, thereby establishing a comprehensive and robust framework for their evaluation. |
|---|---|
| Autores principais: | Ramirez Celis,Andrea Del Carmen |
| Assunto: | Rating Migrations Calibration Probability of Default Credit Risk |
| Ano: | 2026 |
| País: | Portugal |
| Tipo de documento: | dissertação de mestrado |
| Tipo de acesso: | acesso aberto |
| Instituição associada: | Universidade de Lisboa |
| Idioma: | inglês |
| Origem: | Repositório da Universidade de Lisboa |
| Resumo: | In credit risk management, Internal Ratings-Based (IRB) models, particularly Probability of Default (PD) models, are used to estimate the likelihood that an obligor will default within a specified time horizon. A key aspect of credit risk management consists on monitoring the stability of the portfolios underlying these PD models, in order to identify changes in borrower’s behaviour and, credit quality, and the overall model performance. These changes may arise from evolving economic conditions, modifications in lending standards, or issues related to data quality. In addition, the examination of rating migrations provides valuable insights into customer transition dynamics, detecting patterns associated with credit improvement or deterioration. The objective of this work is to develop a robust quantitative methodology for the analysis of rating migrations within PD models. This methodology intends to support the systematic assessment of rating migrations over time, effectively capturing their magnitude, frequency, and direction, while also examining shifts in rating distributions, in order to establish appropriate criteria, e.g. via thresholds. These patterns were analysed across multiple models during different macroeconomic conditions and recalibration processes. To achieve this, several metrics were tested, including the Population Stability Index (PSI), Migration Matrix, Matrix Weighted Bandwidth (MWB), Average of Adjustments, Speed and Direction. Additionally, a deeper analysis was conducted to model migration behaviour, providing information into the dynamic characteristics of the models under study. The results indicate that, from the proposed metrics, the PSI, MWB and Average of Adjustments are the ones that exhibit a greater sensitivity to each model characteristics, macroeconomic conditions and recalibration processes. Collectively, these metrics deliver essential insights into the dynamics of risk grade migrations, thereby establishing a comprehensive and robust framework for their evaluation. |
|---|