EAD Models for Revolving Credit Lines

1. Why Revolving EAD Is Different

Unlike term loans, revolving facilities allow borrowers to draw additional balances right up to default. Exposure at Default (EAD) therefore depends on borrower behavior, credit limit management, and bank actions close to delinquency. Failing to capture these dynamics inflates capital volatility and IFRS 9 provisioning errors.

2. Data Requirements and Segmentation

• Historical daily or monthly balances, limits, and utilization ratios for at least 24 months per obligor.
• Behavioral flags: cash advances, minimum payment patterns, delinquency buckets, cure status.
• Bank actions: limit reductions, blocks, and campaign offers that influence drawdowns.
• Segmentation by product (credit card, overdraft, cheque especial), channel, and customer type (individual, SME).

3. Modeling Credit Conversion Factors (CCF)

CCF represents the share of undrawn commitments expected to be used by the time default occurs. Common approaches include:

1. Static averages per segment, useful when data is scarce but sensitive to structural changes.
2. Logistic or beta regression predicting utilization at default using borrower behavior and macroeconomic variables.
3. Survival analysis tracking utilization trajectory as accounts migrate through delinquency states.
4. Machine learning models (gradient boosting, random forests) capturing non-linear interactions between limit changes and customer behavior.

Whichever model you choose, reconcile total EAD with observed drawdowns during back-testing horizons.

4. Incorporating Behavioral Signals

• Utilization spikes within 60 days before delinquency often anticipate limit max-out; treat them as leading indicators.
• Minimum payment ratios, cash advance usage, and declined transactions can enrich the model with stress signals.
• Account management strategies (proactive limit cuts, collection calls) must be mirrored in the modeling dataset to avoid optimistic CCFs.

5. Stress Testing and Regulatory Expectations

Supervisors expect EAD models to reflect downturn behavior. Practices include stress multipliers on CCF, scenario analysis tied to unemployment or inflation, and overlays for government moratoriums. Document governance triggers that force recalibration when utilization patterns shift materially.