A credit migration rate measures movement between rating or risk grades over a stated period. Learn transition matrices, cohort calculations, withdrawals, stress analysis, and limitations.
A credit migration rate measures the proportion of borrowers, loans, or securities that move from one credit rating or risk grade to another over a stated period. Migration includes upgrades, downgrades, remaining in the same grade, and movement into default.
For a cohort beginning in grade (i), the migration rate to grade (j) is:
where:
The denominator can change if observations are withdrawn, prepaid, sold, or missing. The method must say whether those outcomes remain separate, are excluded, or are adjusted through a duration-based approach.
A portfolio begins the year with 100 borrowers rated A. At year-end:
| Start: A | End count | One-year rate |
|---|---|---|
| AA | 5 | 5% |
| A | 78 | 78% |
| B | 14 | 14% |
| Default | 2 | 2% |
| Withdrawn | 1 | 1% |
| Total | 100 | 100% |
The one-year A-to-B downgrade rate is:
The A-to-default rate is 2%. If the withdrawn observation were excluded, the denominator would become 99 and every reported percentage would change. Neither convention is inherently correct for every use; consistency and disclosure are essential.
A transition matrix places starting grades in rows and ending grades in columns:
Default is often modeled as an absorbing state, meaning an exposure already in default remains there in the matrix. That convention does not mean real exposures can never cure, restructure, or return to performing status; it reflects the model’s chosen state definition.
The cohort method compares grades at fixed start and end dates. It is intuitive but can lose information about interim moves and can be sensitive to withdrawals and portfolio turnover.
A duration method uses the time each exposure spends in a grade and records transitions as they occur. It can use more information from irregular observation periods but requires reliable event timing and more complex estimation.
Methods can produce different estimates from the same portfolio. The method should match the use, data, and rating process.
If a one-year transition matrix is assumed to be stable and Markovian, an (n)-year matrix can be approximated by:
This assumes the next transition depends only on the current grade and that transition behavior remains stable over time. Credit cycles, rating momentum, time spent in grade, and changing underwriting can violate those assumptions.
Directly observed multi-year cohorts, cumulative default studies, or conditional models may be more appropriate when those effects are material.
Migration affects:
A portfolio can have few defaults while experiencing broad downgrades that increase expected loss, capital usage, collateral calls, and refinancing costs.
Migration behavior depends on the rating philosophy:
Comparing matrices without understanding rating philosophy can create false conclusions about portfolio quality or model stability.
A controlled migration system should document data lineage, rating approvals, default definitions, overrides, withdrawals, grade mappings, estimation method, confidence, stability, backtesting, and use. Basel IRB requirements emphasize retaining PDs, realized default rates, and rating-migration histories to assess predictive power.
This article is educational and does not provide individualized investment, lending, accounting, capital, model-validation, or regulatory advice. Migration estimates depend on rating definitions, portfolio composition, data, horizon, method, and economic conditions.