Vintage Analysis for Portfolio Monitoring

1. What is Vintage Analysis?

Vintage analysis groups loans by origination period (monthly or quarterly cohorts) and tracks their performance over subsequent months-on-books (MOB). This longitudinal view reveals whether recent underwriting generates better or worse credit quality than historical standards, enabling proactive portfolio management.

Unlike static snapshots showing current delinquency rates, vintage curves capture full lifecycle behavior—early payment defaults, seasoning effects, and long-term loss realization. This temporal granularity is essential for IFRS 9 lifetime ECL, Basel IRB validation, and strategic decision-making on origination standards.

2. Constructing Vintage Cohorts

Cohort definition:

• Group loans by origination month/quarter into closed cohorts (all loans originated Jan 2025, Feb 2025, etc.).
• Alternative definitions: booking date, first disbursement date, contract signature date—document choice and maintain consistency.
• Minimum cohort size: 100+ accounts for statistical reliability; merge sparse cohorts if necessary.

Performance metrics per MOB:

• Delinquency rates: % 30+ DPD, 60+ DPD, 90+ DPD (charged-off) by months since origination.
• Cumulative default rate: proportion entering Stage 3 by each MOB milestone.
• Roll rates: migration probabilities between delinquency buckets (current → 30 DPD → 60 DPD → default).
• Loss curves: cumulative net charge-offs as % of original balance, tracking realized losses over time.

3. Interpreting Vintage Curves

Seasoning patterns:

• Typical shape: Delinquency rises in early MOBs (honeymoon period ends), peaks at 12-18 MOB, then declines as weak borrowers exit.
• Early payment defaults (EPD): Spikes within first 6 MOBs signal underwriting weaknesses or fraud.
• Long-tail defaults: Deterioration beyond 36 MOB indicates macroeconomic sensitivity or product-specific risks.

Comparative analysis:

• Overlay multiple vintage curves—recent cohorts tracking above historical averages warrant immediate investigation.
• Adjust for portfolio seasoning when comparing young vs. mature vintages (6 MOB vs. 24 MOB).
• Segment by product, channel, credit score band to isolate deterioration sources.

4. Early Warning Indicators from Vintages

Vintage analysis enables predictive monitoring by detecting deviations before portfolio-level metrics react:

• Acceleration alerts: Recent cohorts reaching 5% 90+ DPD faster than historical path (e.g., at 9 MOB vs. typical 12 MOB).
• Roll rate deterioration: Current → 30 DPD transition probability increasing from 2% to 3% for new vintages.
• Loss curve steepening: Cumulative charge-offs for 2025Q3 vintage exceeding 2024 average by >20% at matched MOB.
• Honeymoon shortening: First delinquency appearing at 3 MOB instead of 6 MOB, indicating lower-quality originations.

5. Integrating Macroeconomic Overlays

Vintage performance reflects both underwriting quality and economic conditions at origination:

• Annotate vintage charts with macro indicators (GDP growth, unemployment rate, interest rates) prevailing at origination.
• Identify stress vintages: Cohorts originated during recessions (2020Q2 pandemic, 2008Q4 crisis) show persistent underperformance.
• Adjust expectations: Deterioration during downturn may be cyclical rather than structural underwriting failure.
• Scenario testing: Model vintage curves under forward-looking macro scenarios for IFRS 9 staging and ECL calculation.

6. Vintage Analysis for IFRS 9 and Basel

IFRS 9 applications:

• PD calibration: Vintage default rates provide lifetime default probabilities by origination cohort and seasoning.
• Staging validation: Compare predicted vs. realized Stage 2 migration for closed vintages to assess SICR threshold accuracy.
• Back-testing ECL: Actual losses for matured cohorts validate provisioning adequacy at origination.

Basel IRB validation:

• Long-run default rate (LRDR): Average default rates across full economic cycle measured via vintage analysis.
• Margin of conservatism: Supervisors expect IRB PDs calibrated conservatively relative to observed vintage performance.
• Out-of-time testing: Recent vintages provide holdout dataset for validating rating model discrimination and calibration.

7. Automating Vintage Reporting

Data infrastructure:

• Monthly snapshots: Capture account-level status (balance, DPD, stage) and link to origination cohort.
• MOB calculation: (snapshot date - origination date) in months, handling leap years and partial months consistently.
• Performance aggregation: Calculate cohort-level metrics (delinquency rates, losses) by MOB cross-section.

Visualization best practices:

• Line charts: Overlay 6-12 recent vintages with historical average and confidence bands.
• Heatmaps: Rows = vintages, columns = MOB, color intensity = delinquency rate for pattern recognition.
• Animation: Show vintage evolution over calendar time to illustrate deterioration or improvement trends.
• Automated alerts: Flag vintages deviating >1.5 standard deviations from historical benchmark at critical MOBs.

8. Segmentation Strategies

Homogeneous cohorts improve signal quality:

• Product type: Separate mortgages, auto loans, personal loans—different seasoning curves and risk drivers.
• Origination channel: Branch vs. digital vs. broker—channel quality varies significantly.
• Credit score band: Prime (>720), near-prime (640-720), subprime (<640)—distinct default trajectories.
• Geographic region: State/metro-level cohorts capture localized economic shocks.
• Campaign attribution: Track performance by marketing campaign to assess acquisition ROI.