What is a credit default swap, in plain language?
A credit default swap is, in one sentence, insurance on a bond. One side pays a regular premium; the other side promises to cover the loss if a defined credit event happens to a named borrower.
The analogy is genuinely useful, and its limits are where the instrument gets interesting.
The mechanics
- Reference entity — the borrower whose credit is being insured.
- Notional — the amount of protection, say $10m.
- Maturity — five years is by far the most liquid point.
- Credit events — the defined triggers, standardly bankruptcy and failure to pay, with restructuring included under some documentation conventions.
- Settlement — after a credit event, a market-wide auction establishes a recovery price, and the protection seller pays the notional multiplied by (1 − recovery).
- Determinations committee — an industry body that rules on whether a credit event actually occurred, so that thousands of contracts settle consistently.
The premium arithmetic
Take $10m of five-year protection at a market spread of 200bp.
Quoted as a pure running premium, that is 2.00% × $10m = $200,000 per year, paid quarterly at $50,000.
In practice, contracts were standardised so they could be netted and centrally cleared. The coupon is fixed by convention — commonly 100bp for investment grade and 500bp for high yield — with an upfront payment reconciling the fixed coupon to the market spread:
Upfront ≈ (market spread − fixed coupon) × risky annuity factor
With a market spread of 200bp, a fixed coupon of 100bp, and a five-year risky annuity of about 4.5:
- Upfront ≈ 1.00% × 4.5 = 4.5% of notional = $450,000
The buyer pays that upfront and then 100bp running. Economically the same 200bp; structurally fungible with every other contract on the same reference entity.
If a credit event occurs and the auction fixes recovery at 40%:
- Protection payout = $10m × (1 − 0.40) = $6m
What the premium encodes
Exactly what a bond spread encodes: expected loss, a risk premium, and a liquidity component (usually smaller than a cash bond's, because CDS often trades more actively than the underlying bonds).
Which means the spread decomposition inverts directly. This relationship is sometimes called the credit triangle:
spread ≈ default intensity × (1 − recovery)
At 200bp with 40% recovery: implied default intensity ≈ 0.0200 ÷ 0.60 ≈ 3.3% per year. Risk-neutral, for the same reasons as before — it is a price, not a prediction.
Where the insurance analogy breaks
Three ways, all consequential:
- You need not own the bond. Buying protection without the underlying is a way to be short credit, which cash bonds make difficult. Selling protection is a way to take credit exposure without funding a bond purchase.
- It is a traded contract. The premium moves continuously and marks to market. You can enter and exit without any credit event ever occurring.
- It is defined by documentation. What counts as a credit event, which obligations are deliverable, and how settlement works are contractual questions — and the answers have occasionally diverged from economic intuition.
The CDS-bond basis
Protection on an issuer and the issuer's own bond should imply similar spreads. They usually do not exactly. The difference, CDS spread minus bond spread, is the basis.
It is non-zero for structural reasons: buying a bond requires funding while selling protection does not; the contract's deliverable set differs from any single bond; counterparty risk exists; and documentation quirks matter. In stressed periods the basis often goes sharply negative — cash bonds cheaper than protection — typically because balance sheet to hold bonds is scarce exactly when everyone needs it.
Why CDS levels are watched
Because they are quoted directly in spread terms, update continuously, and exist for sovereigns as well as companies, CDS levels function as a real-time credit barometer. A sovereign CDS series is one of the cleanest available reads on how the market is pricing a country's credit — an observation about price, never a forecast of default.
The previous lesson promised an instrument for the common factor, and a single-name contract is not it. Index CDS are baskets: CDX in North America and iTraxx in Europe, each a fixed, equally weighted list of investment-grade names, 125 in the main indices, or a smaller high-yield list, rolled to a new series every March and September. Protection on the index moves with the market's price of risk and pays out name by name as members default. It trades in far larger size than any single name, which is why a manager hedges the market-wide component with the index and handles single names separately.
In the data
The size of the market is published too. The US regulator, the CFTC, reports each week how much protection is outstanding, in dollars of notional; the table is one week in September 2026, all regions together.
Read it two ways. The first pair of rows is the post-crisis question of how much protection is centrally cleared and how much still sits in bilateral contracts; the second pair splits the same total by the quality of what is insured. And read the two dates at the top: the figures describe positions about two and a half weeks before they were published, so this barometer is always a little late.
Try it now
Read CDS as a barometer:
- What the market charges to insure six governments is below; rank them by spread, converting each fraction to basis points as you go.
- For one of them, apply the credit triangle: implied intensity ≈ spread ÷ (1 − recovery), using a 40% recovery assumption.
- Compare that ranking against the ratings of the same six, below. Where the market ordering and the rating ordering disagree, you have found a question worth investigating, not a conclusion.