Credit-Risk Models and Migration

Compare structural and reduced-form credit models, corporate-failure prediction, and rating-migration analysis, including their inputs and limitations.

Credit-risk models translate borrower, market, and rating evidence into estimates of default, credit deterioration, or loss. This section separates structural models, reduced-form models, failure-prediction methods, and empirical rating-migration analysis.

Key Takeaways

  • Different model families answer different questions and should not be compared by name alone.
  • Structural models connect default to the relationship between firm asset value and obligations.
  • Reduced-form models infer default intensity from market and statistical processes without requiring the same asset-value mechanism.
  • Failure-prediction models use financial, behavioral, market, and other indicators to classify or estimate corporate distress.
  • Migration analysis measures movement among credit grades; migration is not the same as default.

Topic Map

TopicBest use
Structural Model of Credit RiskThe model family that relates firm asset value, liabilities, and a default boundary
Merton ModelThe foundational option-based structural model with default assessed at a defined horizon
Jarrow-Turnbull ModelA reduced-form framework using default intensity and recovery assumptions
Corporate Failure PredictionStatistical, accounting, market, or machine-learning indicators of distress and failure
Migration RateObserved or modeled movement between credit grades over a stated horizon

Model Families Compared

Model familyTypical outputMain evidenceImportant blind spot
StructuralModel-implied default risk, distance to a boundary, or risky-debt valueEquity value and volatility, liabilities, rates, capital structureAsset value and volatility are unobservable; simplified debt boundaries can be unrealistic
Reduced-formDefault intensity, survival curve, or price under a stated recovery conventionBond, loan, or credit-derivative prices and term structuresMarket prices also contain liquidity, risk premia, and technical effects
Failure predictionScore, class, rank, or estimated event probabilityFinancial statements, behavior, market data, governance, and qualitative indicatorsLabels, samples, data leakage, and regime change can dominate apparent accuracy
Migration analysisTransition rates or matrices between gradesInternal or external rating histories and default recordsWithdrawals, rating philosophy, sparse grades, and unstable transition behavior

These outputs are not interchangeable. A risk-neutral probability calibrated for pricing should not be treated automatically as a real-world default forecast, and a ranking score should not be read as a calibrated probability unless the model was designed and validated that way.

Choose a Model by Decision

Decision questionRelevant evidence
How does equity value and volatility relate to default risk?Structural or Merton-style model
What default intensity is implied by market prices?Reduced-form model
Which firms show warning signs of distress?Failure-prediction model
How often do credits upgrade, downgrade, or default?Migration matrix or rate
What is the portfolio’s expected loss?PD, LGD, EAD, correlation, concentration, and scenario assumptions

No model label supplies all of these inputs. Start with the decision, horizon, exposure, and available data.

Evidence to Review

  • default definition and prediction horizon;
  • observation date, sample period, and data frequency;
  • financial statements, market prices, ratings, and borrower behavior;
  • treatment of missing data, survivorship, and censored observations;
  • calibration sample and validation sample;
  • recovery, correlation, transition, and macroeconomic assumptions;
  • backtesting, discriminatory power, calibration, stability, and override policy;
  • use in pricing, limits, allowances, capital, stress testing, or monitoring.

Common Mistakes

  • Treating a model score as a fact or guarantee.
  • Comparing probabilities that use different horizons or default definitions.
  • Applying a public-company market model to an illiquid private borrower without adjustment.
  • Using long-run migration rates as a point-in-time forecast.
  • Ignoring model drift, regime change, data leakage, and concentration.
  • Assuming strong rank ordering means probabilities are well calibrated.

Educational Use

These pages explain model concepts and do not provide individualized investment, lending, accounting, valuation, capital, or regulatory advice. Model output depends on definitions, data, calibration, validation, and use; material decisions require current governance and qualified review.

In this section

Choose a subsection first. Deeper term pages live inside each subsection, which keeps large topic hubs readable.

Corporate Failure Prediction

Corporate failure prediction uses financial, market, behavioral, and qualitative evidence to estimate distress or failure risk. Learn model types, Altman Z-score mechanics, validation, and …

Migration Rate

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.

Jarrow-Turnbull Model

The Jarrow-Turnbull model is a reduced-form framework for pricing defaultable securities and credit derivatives using default timing and recovery assumptions.

Merton Model

The Merton model treats corporate equity as a call option on firm assets to estimate debt value and model-implied default risk. Learn formulas, a worked example, and limitations.

Structural Credit Model

A structural credit-risk model links default to a firm's asset value and debt obligations. Learn model mechanics, Merton-style payoffs, inputs, uses, and limitations.

Credit and Default Risk
Sovereign and Political Risk
Browse Risk Management