Why is default correlation so important in credit portfolio management, and how is it measured?

Default correlation measures the tendency for borrowers to default together. It's arguably the single most important — and most difficult to estimate — parameter in credit portfolio management.

Why It Matters Enormously:

Consider a portfolio of 100 loans, each with PD = 2% and exposure = $1M.

• Zero correlation: Defaults are independent. Expected defaults = 2 per year with low variance. Extreme losses are rare because it's extremely unlikely that many independent borrowers default simultaneously.

• Perfect correlation (rho = 1): Either ALL 100 default or NONE do. 2% of the time you lose $100M. The rest of the time, zero losses.

In reality, correlation is between 0 and 1 (typically 0.05-0.25 for corporate portfolios), but even moderate correlation dramatically increases tail risk.

Quantified Impact:

Notice: Expected loss is the SAME regardless of correlation. But tail risk (and therefore capital) increases dramatically. This is why diversification analysis focuses on correlation.

Measuring Default Correlation:

1. Asset Correlation (Merton Framework):

In the structural model, firms default when their asset value falls below a threshold. Asset correlation (rho_A) between firms' asset returns drives joint default probability.

The Basel formula uses fixed asset correlations:

• Large corporates: rho = 0.12 to 0.24 (inverse function of PD)
• Retail mortgages: rho = 0.15
• Other retail: rho = 0.03 to 0.16

2. Historical Default Correlation:

Directly estimate from observed joint default frequencies. Problem: defaults are rare events, so sample sizes are tiny.

3. Factor Models:

Decompose asset returns into systematic (market) and idiosyncratic factors:

R_i = sqrt(rho_i) x M + sqrt(1-rho_i) x epsilon_i

Where M is the common factor (economy) and epsilon_i is firm-specific. This is the foundation of the Basel IRB formula.

FRM Key Points:

• Correlation estimation error is the largest source of model risk in credit portfolios
• Sector and geographic concentration increases effective correlation
• CDO tranche values are extremely sensitive to correlation assumptions (2008 crisis lesson)
• Wrong-way risk: when PD increases simultaneously with EAD or LGD, effective correlation is even higher
• Name concentration (large exposure to single obligors) amplifies the impact of each default

Master portfolio correlation in our FRM Part II Credit Risk module.