FRM Handbook Ch 24: Managing Credit Risk

Book: Financial Risk Manager Handbook Plus Test Bank
Author: Philippe Jorion
ISBN: 978-0-470-90401-5

Chapter 24 is the payoff chapter for the whole credit block. You spent chapters estimating default probabilities, exposures, and recoveries. Now Jorion stacks them into a portfolio loss distribution and asks the question every credit committee avoids: how much capital do you actually need?

From yes/no to portfolio thinking

Old-school credit was binary. A credit officer approved or killed a loan. Concentration limits were crude guardrails at the portfolio level.

Modern credit risk treats each name as a contributor to total portfolio loss. Diversification matters, but so do correlations. And correlations are much harder to estimate for defaults than for stock returns.

The chapter’s framing is clean. Credit loss for one instrument is exposure at default times loss given default, triggered by a default indicator. For a portfolio, sum across names with netting applied per counterparty.

Net replacement value (NRV) in annual reports is the worst-case “everyone defaults now with zero recovery” number. Jorion compares it to using notionals for derivatives risk. Informative upper bound, bad risk metric.

The loss distribution shape

Simulate defaults, exposures, recoveries, and correlations (usually Monte Carlo) and you get a credit P&L distribution that looks nothing like market risk.

Market returns are roughly symmetric. Credit losses are left-skewed. The distribution resembles a short option position. That matches Merton’s insight: a risky bond equals a risk-free bond plus short credit optionality.

Two numbers matter:

Expected credit loss (ECL) is the average loss. Pricing should cover it. Loan loss reserves (provisions) should fund it. For bonds, higher yield compensates for expected defaults. For derivatives, the bank must bake expected loss into the spread.

Unexpected credit loss (UCL) is the tail. Take a high quantile (99.9% is common for credit), subtract the expected loss, and you get the capital buffer. Equity should cover UCL because provisions only cover expected loss.

Correlations change everything

Default correlation controls tail width. Low correlation with many names: losses cluster near the mean. A bank can run higher leverage (Basel’s 8% minimum implies up to 12.5x in simple terms).

High correlation: simultaneous defaults stretch the left tail. At perfect correlation, worst-case loss at a fixed confidence level is basically total notional. No leverage is safe.

Jorion also flags wrong-way and right-way trades. Wrong-way means exposure rises as default probability rises (Asian crisis dollar loans are the textbook case). Right-way means the trade hedges the counterparty’s business. Wrong-way trades inflate tail risk if you ignore the joint movement.

Recovery rates add another correlation wrinkle. Moody data shows recoveries fall when default rates spike (1990, 2001, 2009). Modeling LGD as fixed underestimates credit VaR in stress.

Credit VaR and the model zoo

Credit VaR parallels market VaR but uses credit-specific confidence levels. Regulators often want 99.9% for credit versus 99% for market risk. The horizon is usually one year, matching the time scale of defaults and credit migrations.

The chapter surveys commercial approaches:

• CreditMetrics (J.P. Morgan): migration-based, mark-to-market credit risk
• CreditRisk+ (Credit Suisse): actuarial, default-mode focus
• KMV / Moody’s KMV: Merton-style equity-to-default mapping
• Credit Portfolio View: macro-driven default rates

None are perfect. Jorion is explicit that complexity demands skepticism. Models differ on marking versus default-only loss, correlation assumptions, and treatment of credit spreads.

What stuck with me

Credit risk management is really two budgets. Provisions fund expected loss (an expense). Capital funds unexpected loss (a balance sheet shock).

If you only watch expected loss, you will look fine until a correlated default wave hits. If you only watch individual name ratings, you will miss portfolio concentration and wrong-way exposure.

The chapter does not pretend portfolio credit models are precise. It argues you still need them because the alternative (binary approvals plus gut feel) failed visibly in 2008.

Previous: FRM Handbook Ch 23: Credit Derivatives and Structured Products (Structured Products and CDOs)
Next: FRM Handbook Ch 25: Operational Risk