Econometrics for Banking Credit Risk Modelling Training Course

Course overview

Bank credit risk teams must convert borrower, account, collateral and macroeconomic data into estimates that can withstand model validation, audit, impairment reporting and portfolio decision-making. Weak econometric specification can produce unstable probability of default estimates, misleading sensitivity analysis and poorly calibrated expected credit loss forecasts. This course addresses the practical gap between knowing regression theory and building, diagnosing and explaining credit risk models used in retail, SME and corporate banking portfolios.

Participants work through the econometric foundations of credit risk modelling, including data design, segmentation, logistic regression for probability of default (PD), survival analysis for time-to-default, transition matrices, macroeconomic satellite models and stress testing. They learn to select variables, treat missing values and outliers, test multicollinearity, assess discriminatory power and calibration, interpret coefficients, and document model limitations. The course also examines IFRS 9 expected credit loss (ECL) applications, Basel model governance expectations and the distinction between development, validation and monitoring metrics.

Instruction combines worked banking cases with hands-on modelling labs using realistic loan-book data. Each participant builds and reviews a credit risk modelling workbook or notebook, producing a documented PD model, performance diagnostics, a macroeconomic stress scenario and a concise model-risk briefing for senior stakeholders. The final deliverable can be adapted as a starting point for an internal model-development, redevelopment or challenger-model assignment.

The programme is designed for practitioners who already work with banking data and need a stronger, defensible econometric approach to credit portfolio analytics, impairment forecasting or model risk review.

Course objectives

By the end of this course, participants will be able to:

• Build a borrower-level credit risk modelling dataset with defined target, observation and performance windows
• Estimate and interpret logistic regression probability of default models for banking portfolios
• Apply missing-data, outlier-treatment and variable-binning methods appropriate to credit datasets
• Test model specification using multicollinearity, significance, stability and residual diagnostic techniques
• Evaluate PD model discrimination and calibration using ROC/AUC, Gini, KS and calibration plots
• Construct macroeconomic satellite models linking default rates to economic drivers and stress scenarios
• Calculate IFRS 9 expected credit loss inputs using PD term structures, staging assumptions and scenario weights
• Produce a model development and monitoring pack with assumptions, limitations, validation evidence and governance actions

Benefits of attending

Target competencies

Who should attend

• Credit Risk Analysts — who develop or monitor borrower, account and portfolio risk measures
• IFRS 9 and ECL Modelling Specialists — who require defensible econometric inputs for impairment forecasting
• Credit Portfolio Managers — who use PD, migration and stress results to steer risk appetite and lending decisions
• Model Risk and Validation Analysts — who challenge assumptions, performance evidence and model documentation
• Risk Data Analysts — who prepare banking datasets for scorecards, default models and portfolio reporting
• Internal Audit and Risk Assurance Professionals — who review the control environment around credit risk models

Requirements and prerequisites

Participants should be comfortable with descriptive statistics, hypothesis testing, correlation, basic linear regression and the structure of lending portfolios. Experience handling tabular data in Excel, Python, R or SAS is expected; participants should be able to filter, join, inspect and summarise datasets. Familiarity with credit risk terms such as default, delinquency, PD, loss given default (LGD), exposure at default (EAD) and IFRS 9 is helpful, but detailed model-development experience is not required. The course does not require advanced calculus, machine-learning expertise, prior coding in every listed tool or prior responsibility for regulatory model approval.

Training methodology

The instructor develops each method from a banking use case before participants apply it to loan-level and portfolio data. Short technical briefings are followed by guided labs in Python, R, SAS Viya or Excel-based outputs, with emphasis on interpreting results rather than running code mechanically. Teams review a challenged PD model, compare segmentation choices, and defend a macroeconomic scenario in a credit committee simulation. Daily exercises build toward an individual model pack, followed by an end-of-course action plan for applying the methods to a live portfolio or model-monitoring issue.

Course outline

Day 1: Credit risk data and econometric foundations

• Banking credit risk model lifecycle and model-risk governance
• Default definitions, cure periods and observation-performance window design
• PD, LGD and EAD roles within Basel and IFRS 9 frameworks
• Loan-level data structures for retail, SME and corporate portfolios
• Missing-value analysis, outlier treatment and data-quality controls
• Portfolio segmentation by product, borrower and behavioural characteristics
• Exploratory analysis using default rates, vintages and cohort views

Workshop: Participants profile a loan-book dataset and produce a data dictionary, target definition and initial segmentation proposal for a PD model.

Day 2: Probability of default model development

• Logistic regression mechanics for binary default outcomes
• Weight of Evidence and Information Value for candidate predictors
• Variable transformations, binning and monotonicity assessment
• Feature selection using business rationale and statistical significance
• Multicollinearity testing with correlation matrices and variance inflation factors
• Coefficient interpretation through odds ratios and marginal effects
• Scorecard scaling and conversion from model outputs to risk grades

Workshop: Participants estimate a logistic PD model, select variables and create a risk-grade mapping with documented business justification.

Day 3: Validation, calibration and model monitoring

• Discrimination testing with ROC curves, AUC, Gini and Kolmogorov-Smirnov statistics
• Calibration assessment using observed-versus-predicted default plots
• Back-testing PD estimates across grades, segments and time periods
• Population Stability Index and Characteristic Stability Index monitoring
• Residual diagnostics, influential observations and specification error
• Out-of-time testing and cross-validation for credit risk models
• Model limitations, overrides and challenger-model assessment

Workshop: Participants validate a supplied PD model and prepare a model-performance dashboard identifying recalibration or redevelopment triggers.

Day 4: Macroeconomic forecasting and IFRS 9 applications

• Time-series properties, stationarity and transformations of macroeconomic data
• Macroeconomic satellite models for portfolio default rates
• Lag structures, autoregression and distributed-lag regression
• Scenario design using baseline, upside and downside economic paths
• Transition matrices and migration analysis for staging assessment
• IFRS 9 staging, lifetime PD term structures and probability-weighted scenarios
• Expected credit loss calculation logic and management overlays

Workshop: Participants build a macroeconomic default-rate model and generate baseline and downside ECL scenario inputs for a portfolio.

Day 5: Model governance, communication and applied casework

• Basel credit risk model-use expectations and governance controls
• IFRS 9 model documentation and audit-trail requirements
• Development versus independent validation responsibilities
• Sensitivity analysis and benchmarking against external or challenger estimates
• Model change policy, recalibration thresholds and approval evidence
• Communicating uncertainty, limitations and material findings to committees
• Model inventory entries and ongoing monitoring plans

Workshop: Participants complete and present a credit risk model pack containing the PD model, validation evidence, stress results, limitations and implementation actions.

Tools & standards covered

Python, R, SAS Viya, IFRS 9

A typical training day

Live online deliveries follow the same structure in the East Africa Time zone, with shorter screen blocks and longer breaks.

What the fee includes

How you can take this course

Certification

Participants who complete the full five days receive the Skillset Development Certificate of Completion, stating the course title, course code, dates and delivery format — suitable for professional-development records and employer reimbursement.

Frequently asked questions