Advanced IRB Approach - Scorecard Development, PD, LGD and EAD Modelling

3 days
Prague, NH Hotel Prague
  • Classic Scorecard Development
  • Representative Sampling and Model Challenges
  • Variable Grouping, Re-coding and Redundancy
  • Multivariate Variable Selection
  • Risk Based Segments versus Rating Classes
  • Horizon-less Application Scorecard Development
  • LGD and EAD Modelling and Scenario Prediction
  • Leading Software Solutions will be used for Case Studies
This course is part two of a two course series. After covering model use and maintenance aspects including validation and re-calibration in the first course, the focus is now on the original creation of risk prediction models.

While the topics in the first course relate more to the Business-As-Usual activities of a credit risk management team, model creation often takes the form of projects, which may sometimes be outsourced to third parties. Having said that, it is strongly advised for everybody working in this field to have a good understanding of all rating system aspects.

The first part of this course explains PD modelling in general and scorecard development in particular. We first cover the general preparatory steps of sampling and partitioning the data. Then we address the various challenges a multivariate model faces and present a variety of common model types that differ in the way they address these challenges.

We then focus on scorecard development in particular, starting with variable grouping and re-coding, moving on to the analysis of variable redundancy and the details of calculating the coefficients of a logistic regression model and finally ending with the calculation of scorepoints.

Two closely related additional topics end the first part of the course. Risk based segmentation uses variable grouping and re-coding from scorecard development to identify segments of cases that share similar value combinations of the most important risk drivers. Horizon-less scorecards address the restrictions imposed by a fixed prediction horizon and predict individual maturation curves instead.

The second part of the course covers LGD and EAD modelling. First, the definition of a workout LGD target variable and the requirements for specific models for already defaulted cases and for downturn adjustments are discussed.

Various LGD modelling approaches are then presented, ranging from simple segment averages to scenario based setups, in which predictions of a workout path are combined with predictions of a respective recovery rate. A large variety of statistical model types are presented that address the specific prediction tasks.

EAD modelling ends the course. Attention is given to the definition of the target variable, especially with regards to the choice of predicting original or undrawn limit usage.

09.00 - 09.15 Welcome and Introduction

09.15 - 12.00 PD Modeling

  • Representative sampling
    • Defining the population
    • Proportional sampling
    • Stratified sampling
    • Over-sampling and adjustments
  • Partitioning
    • Generalization and model complexity
  • Model challenges
    • Missing and erroneous variable values
    • Non-linear variable to target relationships
    • Variable interactions
    • Distribution of target variable
    • Model complexity
    • Model interpretability
  • Model types
    • Linear Regression
    • Logistic Regression and Scorecards
    • Trees and Forests
    • Neural Networks

12.00 - 13.00 Lunch

13.00 - 16.30 Scorecard Development

  • Variable grouping
    • Automatic and manual grouping
    • Criteria for good groupings
    • Weight of Evidence and Information Value
  • Variable relevance: long list to short list
  • Variable re-coding
    • Weight of Evidence and Logit
    • Dummy coding
  • Variable redundancy
    • Intervariable correlation
    • Correlation matrix based clustering
    • Correlation matrix based MDS visualization
    • PCA based variable clustering
  • Logistic regression
    • Calculating coefficients using Maximum-Likelihood
    • Multivariate variable significance
    • Iterative variable selection methods
  • Scorepoints calculation
    • Logodds rescaling based approach
    • Shifted PD based approach

Day Two

09.00 - 09.15 Brief recap

09.15 - 12.00 PD Modeling and Scorecard Development (continued)

  • Risk based segmentation
    • Risk based segments versus rating classes
    • Creating and visualizing risk based segments
      • Weight of Evidence transformation
      • Distance measures
      • K-Means clustering
      • Hierarchical clustering
      • Modularizing a distance graph
    • Profiling risk segments
    • Risk segment based reporting
  • Horizon-less application scorecard development
    • Restrictions imposed by prediction horizons
    • History requirements of horizon-less scorecards
    • Creating the Months On Books variable
    • Modeling maturation curve shapes
    • Approval strategies using horizon-less scorecards
  • Case studies
    • Scorecard development
    • Risk based segmentation
    • Horizon-less scorecard development

12.00 - 13.00 Lunch

13.00 - 16.30 LGD and EAD Modeling

  • Workout LGD
    • Cash flows, recoveries and costs
    • Discount rate
    • Incomplete workouts
    • Negative LGD
  • Default at Observation
  • Downturn LGD
  • LGD Averaging
    • Default weighted or exposure weighted
  • Workout scenarios
    • Cure rate
    • Recovery rate
  • Characteristics

Day Three

09.00 - 09.15 Brief Recap

09.15 - 12.00 LGD and EAD Modeling (continued)

  • Statistical scenario prediction
    • Multinomial and Ordinal Logistic Regression
    • Nominal Tree
  • Statistical LGD and rate prediction
    • Segmentation
      • Regression Tree
    • Linear Regression
      • Input standardization and normalization
      • Target variable normalization
      • Beta-regression
    • Non-linear regression
      • Polynomials and Interactions
      • Grouped variable linear regression
      • Neural Network
    • Pseudo-Logistic Regression and Scorecard

12.00 - 13.00 Lunch

13.00 - 16.30 LGD and EAD Modeling (continued)

  • LGD validation
    • Challenges of LGD Validation
    • LGD model quality measures
  • Exposure at Default modeling
    • Predicting limit usage at default
    • Limit or undrawn limit: CCF vs. LEQ
    • Characteristics
  • Case studies
    • Scenario prediction
    • Regression Tree
    • Linear and non-linear Regression
    • Scorecard

Evaluation and Termination of the Seminar

back to calendar
detailed program
Printer Friendly Format
E-mail Newsletter
Contact Us
  Site Map    Policy Statements    Mobile Version