Skill Guide

Credit risk modeling fundamentals (PD, LGD, EAD frameworks)

Credit risk modeling fundamentals are the quantitative methods used to estimate the probability of a borrower's default (PD), the loss given that default (LGD), and the total exposure at the time of default (EAD).

This skill enables financial institutions to price loans accurately, set adequate capital reserves (Basel compliance), and make data-driven lending decisions that directly protect the institution's profitability and stability.

1 Careers

1 Categories

9.1 Avg Demand

15% Avg AI Risk

How to Learn Credit risk modeling fundamentals (PD, LGD, EAD frameworks)

1. Master core terminology (PD, LGD, EAD, Expected Loss). 2. Understand the lifecycle of a loan and where default occurs. 3. Learn basic statistical concepts (logistic regression for PD, beta distributions for LGD).

1. Apply these concepts to portfolio segmentation (e.g., retail vs. corporate). 2. Build and back-test a simple PD scorecard using historical data. 3. Avoid common pitfalls like data snooping or using macroeconomic variables without stationarity checks.

1. Architect integrated IFRS 9/CECL models that align PD, LGD, EAD with economic scenarios. 2. Stress-test models under severe but plausible scenarios (e.g., pandemic, rate shock). 3. Mentor junior modelers and communicate model limitations and uncertainties to non-technical stakeholders (e.g., board, regulators).

Practice Projects

Beginner

Project

Build a Basic PD Scorecard for Consumer Loans

Scenario

You are a junior analyst at a consumer bank. Use a sample dataset of past loan applications and outcomes to build a model that predicts default.

How to Execute

1. Clean and preprocess data (handle missing values, create age/income bands). 2. Use Weight of Evidence (WOE) encoding for categorical variables. 3. Fit a logistic regression model. 4. Evaluate model performance using AUC (Area Under Curve) and Gini coefficient.

Intermediate

Project

Develop an LGD Model for Commercial Real Estate Loans

Scenario

The bank's commercial lending portfolio shows high variance in recovery rates after foreclosure. You need to build a model to predict LGD for capital allocation.

How to Execute

1. Analyze historical recovery data, including time-to-recovery and collateral type. 2. Fit a Beta distribution to model LGD (bounded between 0 and 1). 3. Incorporate macroeconomic factors (e.g., commercial property price index) into the model. 4. Validate the model using out-of-time samples.

Advanced

Case Study/Exercise

Design an Integrated IFRS 9 Impairment Model for a Hypothetical Bank

Scenario

A multinational bank is transitioning to IFRS 9. You are tasked with designing a compliant, auditable model that calculates lifetime expected credit losses (ECL) for its corporate loan book.

How to Execute

1. Define stage transfer criteria (significant increase in credit risk). 2. Develop a Point-in-Time (PIT) PD model calibrated to macroeconomic scenarios. 3. Integrate forward-looking information into the ECL calculation (PD * LGD * EAD * discount factor). 4. Run the model under multiple probability-weighted scenarios (optimistic, baseline, pessimistic) and document all assumptions for auditors.

Tools & Frameworks

Software & Platforms

Python (pandas, scikit-learn, statsmodels)R (caret, mgcv)SAS/STATSQL for data extractionExcel for high-level modeling and reporting

Use Python/R for model development and validation. Use SQL to source and manipulate large datasets from data warehouses. Excel remains common for quick calculations and communicating results to business units.

Regulatory & Conceptual Frameworks

Basel II/III IRB ApproachIFRS 9 / CECL (Current Expected Credit Loss)Basel III Standardized ApproachMachine Learning Interpretability Techniques (SHAP, LIME)

Basel frameworks dictate minimum capital requirements and model approaches. IFRS 9/CECL governs accounting for expected losses. ML interpretability tools are increasingly required to explain complex model outputs to regulators.

Interview Questions

Answer Strategy

Test the candidate's understanding of model stability and macroeconomic sensitivity. A strong answer will discuss examining population stability (PSI), testing for overfitting, and recalibrating the model with economic downturn data or incorporating macroeconomic variables as drivers.

Answer Strategy

Tests the candidate's ability to connect theory to asset class specifics. Focus on collateral recovery, time value of money, and workout costs.