How much of your money actually comes back after a default?

The recovery rate finishes the expected loss arithmetic. It is the fraction of your claim you get back, and it is the number people estimate most casually and should estimate most carefully.

Definitions, and two ways to measure

Recovery rate = value received ÷ claim, and LGD = 1 − recovery rate.

There are two standard measurement conventions, and they do not agree:

• Trading-price recovery. The market price of the defaulted instrument roughly 30 days after the event. Fast, observable, and the basis for how credit derivatives settle.
• Ultimate recovery. The value actually distributed when the process concludes, discounted back to the default date. Truer economically, but only knowable years later.

When a study quotes "average recovery," which convention it used matters — the two can differ by a lot for the same defaults.

The pattern by seniority

Long-run studies consistently produce the same ordering, though the levels vary by period, region and methodology. Rounded and illustrative:

• Senior secured: around 60%
• Senior unsecured: around 40%
• Subordinated: around 20%

Treat these as shape, not truth. The dispersion around each average is enormous — individual outcomes range from near zero to near par within every single category.

What drives the number

• Collateral quality and coverage. Not just whether there is collateral, but how much value it holds relative to the secured claim.
• Debt ahead of you. The example in the previous lesson showed how completely this dominates.
• Asset tangibility. An airline has aircraft; a utility has a network; a consultancy has people who can leave. Businesses whose value walks out of the building at night recover poorly.
• Jurisdiction and process. Creditor rights, court speed, and whether restructuring or liquidation is the norm.
• The cycle. Recoveries fall when defaults rise, because distressed assets get sold into a market already full of distressed sellers with few buyers.

The correlation that makes credit hard

That last driver deserves emphasis, because it breaks a comfortable assumption. PD and LGD are not independent — they are positively correlated. The years with the most defaults are the years with the worst recoveries.

Watch what that does to expected loss:

• Benign conditions: PD 2%, recovery 45% → LGD 0.55 → EL = 110bp
• Stressed conditions: PD 10%, recovery 25% → LGD 0.75 → EL = 750bp

Default frequency rose 5×. Expected loss rose nearly 7×, because both terms in the product deteriorated together. A model that holds recovery fixed at a long-run average will understate bad years substantially — and understating exactly the bad years is the expensive kind of error.

This is also, in large part, the justification for the risk premium in the spread decomposition. Investors are not merely paid for average losses; they are paid because losses concentrate in the states where they hurt most.

Try it now

Test the sensitivity yourself:

1. Take the spread you built in the first lesson of this course, or rebuild it from the two tables below: the high-quality corporate 5-year yield for August 2026 minus the average of the three Treasury 5-year readings from the same month. Derive the implied default rate using implied PD ≈ spread ÷ LGD, with LGD = 0.60.

2. Redo it with LGD = 0.40 and again with LGD = 0.80.
3. Note how far the implied default rate moves on a recovery assumption alone. Anyone quoting an implied default probability without stating a recovery assumption has told you half a number.