CreditMetrics (1997, J.P. Morgan) Multi-Obligor Portfolio Risk Lab

Simulate multi-obligor corporate loan and bond portfolios across S&P 8-state rating migrations, forward zero-coupon credit yield curve revaluations, correlated latent asset returns, and 99% CreditVaR.

J.P. Morgan (1997) Rating Migration Matrices Correlated Gaussian Copula CreditVaR & CVaR
Portfolio & Simulation Parameters
CreditMetrics Engine
0.25
0.00 (Independent) 0.40 (Typical) 0.85 (Crisis Systemic)
Name Rating Face ($M) Cpn (%) Mat (Y) Rec (%) Del
CreditMetrics Mathematical Architecture

1. Latent Asset Return & Transition Thresholds:

$$Z_i \sim \mathcal{N}(0, 1), \quad Z_{i, s} = \Phi^{-1}\left(\sum_{m=s+1}^8 P_{i, m}\right)$$

Obligor $i$ ends in rating state $s$ if $Z_{i, s} < Z_i \le Z_{i, s-1}$.

2. Forward Bond Revaluation:

$$V_{i, s} = \sum_{\tau=1}^{T_i - 1} \frac{C_i}{(1 + f_{s, \tau})^\tau} + \frac{F_i + C_i}{(1 + f_{s, T_i - 1})^{T_i - 1}}, \quad V_{i, D} = F_i \cdot R_i$$

3. Analytical Portfolio Volatility:

$$\sigma_{\text{port}} = \sqrt{\sum_i \sigma_i^2 + 2 \sum_{i < j} \text{Cov}(V_i, V_j)}$$

where $\text{Cov}(V_i, V_j)$ is computed via bivariate normal integrals $\Phi_2(Z_{i, s}, Z_{j, s'}; \rho)$.

4. CreditVaR & Expected Shortfall:

$$\text{CreditVaR}_\alpha = \mathbb{E}[V] - q_{1-\alpha}(V_{\text{port}})$$ $$\text{CVaR}_\alpha = \mathbb{E}[\text{Loss} \mid \text{Loss} \ge \text{CreditVaR}_\alpha]$$
Total Portfolio Face $75.0M Notional Exposure
Expected 1Y Value $78.20M Discounted Horizon Value
Credit Volatility ($\sigma$) $1.85M (2.47%) Analytical Benchmark
99% CreditVaR $5.42M (7.23%) 1-Year Tail Loss
99% Expected Shortfall $7.15M (9.53%) Conditional VaR (CVaR)
Diversification Benefit $2.10M (27.9%) $\sum \text{VaR}_i - \text{VaR}_{\text{port}}$

Notice the characteristic fat right tail and positive skewness. Yellow bars denote losses between 95% and 99% VaR; red bars highlight the extreme 99% tail losses.

Scatter of simulated joint standardized asset returns for Obligor 1 and Obligor 2. Higher correlation aligns points along the 45-degree diagonal, driving concurrent downgrades and defaults.

Decomposition of portfolio risk across obligors comparing Notional Face Value, Standalone 99% VaR, and actual Component CreditVaR (marginal covariance contribution).

Obligor Risk Decomposition & Revaluation Profile
1-Year Horizon
Obligor Rating Face Coupon Mat Expected Val Volatility ($\sigma_i$) Standalone VaR Component VaR (%)
Portfolio Sensitivity Analysis (Dual 5x5 Heatmaps)
Correlation ($\rho$) vs Concentration on 99% CreditVaR ($M)
$\rho \downarrow$ \ Portfolio $\rightarrow$ Equal Top-Heavy Bipolar HY-Heavy Base
Base Rating vs Recovery Rate on Credit Volatility ($\sigma_{\text{Credit}}$)
Rating $\downarrow$ \ Rec % $\rightarrow$ 20% 30% 40% 50% 60%
CreditMetrics Quantitative Mastery Quiz
Frequently Asked Questions: CreditMetrics & Portfolio Credit Risk

CreditMetrics is the landmark portfolio credit risk management framework introduced by J.P. Morgan in 1997. It evaluates the credit risk of a multi-obligor portfolio of bonds, loans, and commitments across a specified risk horizon (typically one year). Unlike default-only models, CreditMetrics captures both credit rating migrations (upgrades and downgrades) and defaults, pricing each bond along forward zero-coupon credit curves.

For each possible ending rating state (AAA through CCC), remaining bond cash flows are discounted using the forward zero-coupon credit yield curve corresponding to that rating grade. If the obligor migrates to Default, the recovery rate is applied to par value. This produces an 8-state valuation profile for each obligor, capturing the spread widening economic loss of downgrades.

Portfolio expected value and expected loss depend solely on the marginal transition probabilities of each obligor, remaining constant regardless of correlation. Pairwise asset correlation governs the joint likelihood of simultaneous downgrades and defaults. Higher correlation widens portfolio volatility, fattens the extreme right tail of the loss distribution, and sharply increases 99% and 99.9% CreditVaR.

Standalone VaR is the 99% credit loss quantile of an obligor in isolation, ignoring portfolio diversification. Component CreditVaR (marginal risk contribution) measures the actual dollar portion of total portfolio CreditVaR attributable to that obligor, accounting for its size, rating volatility, and covariance with the overall portfolio. By Euler's allocation principle, the sum of Component VaRs equals total portfolio CreditVaR.

The Merton (1974) model is a single-firm structural model of default based on equity call options. The Vasicek (2002) ASRF model focuses on an infinitely granular portfolio under a single systematic factor to determine regulatory capital. CreditMetrics is a multi-factor, multi-obligor finite portfolio model that evaluates full credit rating migration transitions and discrete mark-to-market economic revaluations.