What's the difference between CreditMetrics and CreditRisk+ for modeling credit portfolio risk?

CreditMetrics and CreditRisk+ represent two fundamentally different philosophies for modeling credit portfolio risk. Understanding the contrast is essential for FRM Part II.

CreditMetrics (JP Morgan, 1997) — The Migration Approach:

CreditMetrics models credit rating transitions, not just defaults. A BBB bond can migrate to BB (downgrade), A (upgrade), or D (default). Each transition changes the bond's value.

Key Features:

1. Uses a transition probability matrix (e.g., from Moody's or S&P)
2. Correlates obligors through asset return correlations (Merton-style)
3. Simulates correlated asset returns, maps to rating transitions, revalues portfolio
4. Captures spread risk from downgrades, not just default losses

CreditRisk+ (Credit Suisse, 1997) — The Actuarial Approach:

CreditRisk+ models defaults as a Poisson process — like insurance claims. It doesn't track rating migrations, only whether an obligor defaults or not.

Key Features:

1. Default follows a Poisson distribution with a stochastic mean
2. No asset return correlations — instead, defaults are driven by common sector factors
3. Closed-form loss distribution (no Monte Carlo needed)
4. Only captures default risk, not downgrade spread widening

Head-to-Head Comparison:

Example — Northfield Credit Partners:

Northfield holds a 500-name corporate bond portfolio. They run both models:

• CreditMetrics: Simulates 100,000 scenarios of correlated asset returns. In each scenario, every obligor's return determines its rating migration. Portfolio loss at 99.9% = $45.2M (includes $12M from downgrades that didn't default)
• CreditRisk+: Uses 3 sector factors (financials, industrials, tech). Generates the loss distribution analytically. Portfolio loss at 99.9% = $38.7M (default losses only)

The $6.5M difference comes from migration risk that CreditMetrics captures but CreditRisk+ ignores.

Which Is Better?

Neither — they serve different purposes:

• Trading books where bonds are marked to market: CreditMetrics (captures spread widening)
• Banking books with held-to-maturity loans: CreditRisk+ is adequate and faster
• Regulatory capital: Basel uses a one-factor Gaussian copula model that's philosophically closer to CreditMetrics

For more on credit portfolio models, explore our FRM Part II course materials.