Credit Risk Migration: Transition Matrices & Ratings Dynamics

Published: April 16, 2026

Article by Ryan O'Connell, CFA, FRM

Credit migration — the movement of borrowers across credit rating categories over time — is the foundation of credit portfolio risk management. When a bank holds a portfolio of corporate bonds or loans, the value of that portfolio depends not just on whether borrowers default, but on whether their credit quality improves or deteriorates. A credit migration matrix quantifies the probability of every possible rating change over a given period, providing the statistical backbone for credit portfolio models like CreditMetrics. Understanding how to read, calculate, and apply transition matrices is essential for anyone working in credit risk, fixed income portfolio management, or credit portfolio modeling.

What Is a Credit Migration Matrix?

A credit migration matrix (also called a rating transition matrix) is a table that shows the historical probability of a borrower moving from one credit rating to any other rating — including default — over a specified time period, typically one year.

Credit migration analysis rests on a Markov chain assumption: the probability of a future rating depends only on the current rating, not on the path that led to it. A bond rated BBB today has the same transition probabilities regardless of whether it was downgraded from A or upgraded from BB. While this assumption simplifies the mathematics considerably, empirical evidence shows it is only approximately true — a point we address in the Limitations section.

A critical feature of transition matrices is that default is an absorbing state. Once an issuer defaults — in the sense defined by probability of default analysis — it cannot migrate back to a performing rating. This means the default column accumulates probability over time, and the default row is always [0, 0, …, 0, 100%].

Rating agencies — primarily S&P Global Ratings and Moody’s Investors Service — publish transition matrices annually based on decades of historical data. These matrices form the empirical foundation for credit portfolio models.

Rating Transition Matrices: How to Read S&P and Moody’s Data

The table below shows an approximate one-year corporate rating transition matrix based on S&P Global’s Annual Global Corporate Default and Transition Study (1981–2023, global corporate issuers). Values are in percentages and represent the probability of migrating from the row rating to the column rating within one year.

Source: Approximate long-run average values based on S&P Global Ratings’ published annual transition studies (global corporate issuers). Rows may not sum to exactly 100% due to rating withdrawals (NR category omitted for clarity). Exact values vary by study period and methodology.

Several patterns emerge from this matrix:

• Diagonal dominance — Most issuers retain their current rating. Higher-rated issuers are more stable: AAA and AA ratings show ~87–88% retention rates, while CCC/C shows only ~49%.
• Negative drift — Downgrades are generally more probable than upgrades. A BBB-rated issuer has a 4.4% chance of falling to BB but only a 3.6% chance of rising to A. General Electric’s multi-year descent from AA to BBB between 2009 and 2019 illustrates how even blue-chip issuers migrate through rating categories over time.
• Default acceleration — One-year default probabilities increase dramatically as ratings decline: ~0.0% for AAA, 0.2% for BBB, 0.6% for BB, 3.0% for B, and 26.0% for CCC/C.

Calculating Multi-Year Transition Matrices

The one-year matrix tells you the probability of rating changes over a single year. But credit risk management often requires projecting over longer horizons — a 5-year bond needs a 5-year cumulative default probability, not just a 1-year figure.

This formula relies on a time-homogeneous discrete Markov chain with default as an absorbing state. The key insight is that matrix multiplication automatically accounts for all possible intermediate paths. When you compute P³, the resulting probability of migrating from BBB to Default over three years includes the path BBB → BB → B → Default, but also BBB → BBB → BBB → Default, BBB → A → BBB → Default, and every other conceivable three-step path through the rating system.

It is important to distinguish two related but different measures:

• Marginal default probability — The probability of defaulting in a specific single year, conditional on surviving to the start of that year
• Cumulative default probability — The total probability of defaulting at any point from now through year n, read directly from the Default column of Pⁿ

Cumulative default probability always increases with the time horizon. Even AAA-rated issuers have a meaningfully higher probability of defaulting over 10 years than over 1 year, because they can migrate through intermediate ratings before eventually defaulting.

Generator Matrices and Continuous Time

For fractional-year transitions or to ensure mathematical consistency when interpolating between annual observations, practitioners use a generator matrix Q:

The generator matrix Q has a specific structure: off-diagonal elements q_ij ≥ 0 represent the instantaneous rate of migration from rating i to rating j, diagonal elements are negative (equal to the negative sum of the off-diagonal elements in that row), and each row sums to zero. This ensures that P(t) remains a valid probability matrix for all t > 0.

In theory, Q = ln(P) — the matrix logarithm of the annual transition matrix. In practice, not every empirically observed transition matrix has a valid generator. This embeddability problem arises because small sample sizes and estimation noise can produce matrices that violate the mathematical requirements of a continuous-time Markov chain. When this occurs, practitioners use numerical approximation methods to find the closest valid generator.

Multi-Year Transition Example

Using Transition Matrices for Credit VaR

Transition matrices are a core input to credit Value at Risk (credit VaR) calculations, most notably in the CreditMetrics framework developed by J.P. Morgan in 1997.

The approach works as follows: for each bond in a portfolio, the transition matrix defines the probability of every possible rating at the end of the horizon. Each new rating implies a different credit spread — an issuer that migrates from BBB to BB will see its spread widen, reducing the bond’s market value. An issuer that migrates from BBB to A will see its spread tighten, increasing the bond’s value. Default produces a loss equal to the exposure minus the recovery value. Market-implied migration probabilities can also be extracted from credit default swap spreads, providing a forward-looking complement to historical matrices.

By mapping every possible migration outcome to a change in portfolio value, the model generates a full distribution of portfolio gains and losses. Credit VaR is then the loss at a specified percentile (e.g., the 99th percentile) of this distribution.

The transition matrix provides the single-name migration probabilities, but a portfolio model also needs default correlation to capture the tendency for credit quality to deteriorate simultaneously across issuers during economic downturns. CreditMetrics uses asset-return correlations, drawn from equity markets, to model correlated migrations across the portfolio.

For a comprehensive treatment of portfolio credit risk models — including CreditMetrics, KMV, CreditRisk+, and the Vasicek single-factor model used in Basel’s IRB approach — see our article on credit portfolio models.

Cohort vs Duration Method

Rating agencies estimate transition matrices using two fundamentally different statistical approaches. Understanding the distinction matters because the method affects the resulting transition probabilities, particularly for lower-rated issuers and short time horizons.

S&P’s Annual Global Corporate Default and Transition Study primarily uses the cohort method, making it the industry standard for published matrices. Moody’s has historically favored the duration approach in some of its research. For most practitioners, the cohort-based matrices from S&P or Moody’s are sufficient. The duration method becomes relevant when you need to estimate transition rates for non-standard time horizons or when dealing with sparse data for rarely-observed transitions.

How to Calculate Migration Probabilities

To project credit migration probabilities for your own portfolio analysis, follow these steps:

1. Obtain the one-year transition matrix — Use the most recent published matrix from S&P Global or Moody’s, matched to your portfolio’s rating system. Ensure the matrix corresponds to the correct issuer universe (global corporate, financial institutions, sovereign, etc.).
2. Raise the matrix to the power n — For an n-year horizon, compute P(n) = Pⁿ through repeated matrix multiplication. Standard numerical software (Python, R, MATLAB) handles this directly.
3. Read the cumulative default probability — The Default column of Pⁿ gives the cumulative probability of default over n years for each starting rating.
4. Extract conditional migration probabilities — If you need the probability of migrating from rating i to rating j in year k, given survival to year k, use the sub-matrix of surviving (non-default) states.
5. Adjust for business cycle effects — Published matrices are through-the-cycle averages. For point-in-time analysis, scale transition probabilities using current economic indicators or use recession-conditioned matrices published separately by the rating agencies.

Common Mistakes

Credit migration analysis involves several subtle pitfalls that can lead to materially incorrect risk estimates:

1. Multiplying the one-year default probability by n for an n-year estimate. A BBB issuer with a 0.20% one-year default probability does not have a 1.00% five-year default probability. The correct approach is matrix exponentiation (Pⁿ), which accounts for intermediate rating migrations. The linear approximation always underestimates the cumulative default probability because it ignores the compounding effect of migration through lower-rated states.

2. Reading rows and columns backward. In a transition matrix, rows represent the starting rating and columns represent the ending rating. Confusing the two reverses the direction of migration and produces nonsensical results. Always verify: the entry at (BBB, BB) means “probability a BBB-rated issuer migrates to BB,” not the reverse.

3. Using corporate transition matrices for structured products. Corporate and structured finance (ABS, CDO tranches) migration dynamics differ significantly. Structured finance ratings tend to be more stable initially but exhibit sharper cliff-like downgrades. S&P publishes separate structured finance transition studies for this reason.

4. Ignoring rating agency differences. S&P and Moody’s use different methodologies, rating scales, and sample populations. Mixing an S&P-rated portfolio with a Moody’s transition matrix introduces systematic bias. Even after mapping to a common scale, residual methodology differences remain.

5. Assuming transitions are independent of the business cycle. Historical transition matrices are through-the-cycle averages that smooth over recessions and expansions. During the 2008–2009 financial crisis, speculative-grade default rates exceeded 10% — far above the long-run average. Using unconditional matrices in stress testing understates tail risk.

Limitations of Credit Migration Matrices

Markov assumption is violated in practice. Empirical research consistently shows that credit ratings exhibit momentum — a recent downgrade increases the probability of further downgrade beyond what the Markov model predicts. This means that two BBB-rated issuers may have different migration probabilities depending on whether they were recently downgraded from A or upgraded from BB.

Through-the-cycle matrices smooth over cyclical variation. Rating agencies calibrate matrices over full business cycles, producing averages that may not reflect current economic conditions. During recessions, actual transition rates — especially downgrades and defaults — can be dramatically higher than the long-run average.

Rating methodology changes over time. Rating agencies periodically recalibrate their criteria, which can shift transition rates independently of actual changes in credit quality. Comparing matrices from different decades requires adjusting for these methodological shifts.

Ratings are discrete approximations. Continuous credit quality is compressed into a small number of categories. Two issuers both rated BBB may have very different underlying credit profiles — one near the top of the category, the other near the bottom. This within-category heterogeneity introduces noise into transition-based risk models.

Limited data for rare transitions. AAA defaults are essentially unobserved in historical data. Transition probabilities for the highest-rated categories are estimated from very small samples, making them unreliable for extreme scenarios. The same applies to multi-notch upgrades (e.g., B to A in one year), which are so rare that their estimated probabilities may be zero despite being theoretically possible.