What Is a Credit Default Swap?

You know what a bond is, you know that issuers sometimes fail to pay it back, and you know that when they do you only recover some fraction of face value. So here’s the question that built an entire market: what if you could buy insurance on that default — a contract that pays you when a borrower goes bust — and what if you could buy it even if you don’t own the bond at all? That’s a credit default swap (CDS): a tradable insurance policy on a borrower’s creditworthiness. It is the most important building block in credit derivatives, the thing the financial crisis was largely fought over, and — once you see it leg by leg — one of the most elegant contracts in finance. Let’s take it apart.

The two legs of a CDS

A CDS is a swap, and like every swap it has two sides exchanging cashflows. Think of it as an insurance contract turned into a tradable instrument. The protection buyer is the one buying cover — they pay a small, regular premium, like insurance instalments. The protection seller is the insurer — they collect those premiums and, in exchange, promise to make the buyer whole if the borrower (the reference entity) suffers a credit event. The bond, loan, or issuer being insured is the reference obligation / reference entity.

So there are exactly two legs:

• The premium leg (a.k.a. the running-spread leg or fee leg): the buyer pays the seller a periodic premium, quoted as an annualized spread in basis points (bp) of the notional. Notional is the face amount of credit being insured — say $10 million. Premiums are paid quarterly, on the standardized IMM dates: the 20th of March, June, September, and December.
• The protection leg (a.k.a. the contingent leg): the seller pays the buyer nothing at all — unless and until a credit event hits. If it does, the seller pays out notional × (1 − Recovery), where Recovery (R) is what the defaulted bond turns out to be worth (per dollar of face). That payment compensates the buyer for the loss on a defaulted bond.

The buyer is therefore long protection and, in credit-speak, short credit — they profit when the borrower’s credit deteriorates. The seller is short protection / long credit — they’re effectively long the bond’s fortunes, collecting carry and praying nothing blows up. (Sound familiar? It’s the short-variance trade’s spiritual cousin: collect small premiums, wear a fat tail.)

Let’s make the premium leg concrete. A $10,000,000 notional CDS at a 200 bp spread costs the buyer 0.02 × $10,000,000 = $200,000 per year, billed as four quarterly instalments of $50,000 each. Over a full 5-year contract, if nothing ever defaults, here’s the buyer’s premium leg:

Five quiet years, $1,000,000 handed over, and the protection leg never fires. The buyer paid a million dollars for peace of mind they didn’t end up needing — which, as any insurance buyer knows, is both the point and the annoyance.

The pitfall: people read “200 bp” as a one-time cost. It isn’t — it’s an annual rate, paid every period until the contract matures or a credit event ends it early. And the spread is quoted on notional, not on the bond’s market price, so a distressed bond trading at 60 cents still has its CDS premium computed on full face value.

Credit events & settlement

So far so good — but the whole contract hinges on one question: what exactly counts as a “default”? If the trigger were vague, every CDS would dissolve into litigation. So the industry standardized it. Under the ISDA (International Swaps and Derivatives Association) definitions, a CDS pays out only on a specifically enumerated credit event. The three core ones for corporates are:

• Bankruptcy — the reference entity files for insolvency / is wound up. (Not applicable to sovereigns, which can’t go bankrupt in the corporate sense.)
• Failure to pay — the entity misses a payment of principal or interest above a materiality threshold, after any grace period. The bread-and-butter trigger.
• Restructuring — the terms of the debt are forcibly changed to the lenders’ disadvantage: maturity extended, coupon cut, principal reduced, ranking subordinated. Controversial, because the bond hasn’t actually “failed” — it’s just been bent. (More on the variants in the next section.)

Sovereigns add extras like repudiation/moratorium (a government declares it won’t pay) and obligation acceleration. The point is that a credit event is a defined legal trigger, not a vibe — “the stock dropped 40%” is not a credit event.

Who decides whether one happened? Not the two counterparties arguing in a car park — the ISDA Determinations Committee (DC), a panel of dealers and buy-side firms that rules officially on whether a credit event occurred for a given name. Their ruling binds every CDS on that entity at once.

Once a credit event is declared, the contract must settle — the protection seller pays up. The amount always boils down to notional × (1 − R), but how you get there comes in two flavours:

• Physical settlement (the old way): the buyer literally delivers the defaulted bonds to the seller and receives par (full face value) in cash. The buyer’s gain is par minus the bonds’ market price — i.e. (1 − R) per unit.
• Cash settlement (the modern default): nobody delivers bonds. The seller simply pays the buyer notional × (1 − R) in cash, where R is determined by a market-wide CDS auction (next section). Cleaner, and it doesn’t require the buyer to go hunting for deliverable bonds.

And the star of the show: Recovery (R) is the value of the defaulted debt after the event, per dollar of face. If the bonds change hands at 40 cents, R = 40%, and the protection pays (1 − 0.40) = 60% of notional. Recovery is the bridge between “default happened” and “here’s the cheque.”

The worked example

Time to watch a whole CDS live and die. The interactive below settles a 5-year single-name CDS on a fixed $10,000,000 notional. Drag the spread slider (50–600 bp) to set the annual premium, the recovery slider (0–80%) to set what the bond is worth after default, and then flip the toggle between Survives and Defaults (in year N). Watch three readouts: the total premium the buyer paid, the protection payout the seller owes, and the buyer’s net P&L. Start at the defaults — set spread 200 bp, recovery 40%, and toggle “Defaults in year 3.”

Let’s nail down the arithmetic the island is showing, with our base case: $10M notional, 200 bp spread, 40% recovery.

The premium leg is 0.02 × $10M = $200,000 per year. The protection leg, if it fires, pays (1 − R) × notional = (1 − 0.40) × $10M = $6,000,000. Now the two paths:

Read both rows carefully, because they are the trade. If the name survives, the buyer dribbles out $1,000,000 over five years and collects nothing — the protection leg never fires, and they’re down a cool million. That’s the cost of insurance you didn’t need.

If the name defaults in year 3, two things happen at once: the seller pays (1 − 0.40) × $10M = $6,000,000, and the premium payments stop (you don’t keep paying insurance on a house that already burned down). The buyer paid only two years of premium — $400,000 — and received $6,000,000, for a net of +$5,600,000. That asymmetry — small, steady outflow versus a huge contingent inflow — is the whole shape of the protection buyer’s payoff: lose a little for years, win big in a crisis.

Flip the perspective and you’ll see the seller has the mirror image: collect $200,000 a year quietly, then potentially write a $6,000,000 cheque in a single bad afternoon. Now adjust the recovery slider in the island: crank R up to 80% and the payout shrinks to (1 − 0.80) × $10M = $2,000,000; drop R to 0% (a total wipeout) and it balloons to the full $10,000,000. Lower recovery, bigger payout — the protection is worth most precisely when the default is most brutal.

The standardized contract (the Big Bang)

Early CDS were bespoke OTC deals: every spread was negotiated, every recovery argued over, settlement was physical and clunky, and each contract had its own quirks. That’s fine for a sleepy market — and a nightmare for a market that, by 2008, was tens of trillions of dollars notional with everyone facing everyone else. So in April 2009, ISDA pushed through the “CDS Big Bang”: a sweeping standardization that turned CDS from artisanal contracts into something that trades like a futures product. Four changes mattered most.

• Fixed coupons + an upfront payment. Instead of a custom spread per trade, contracts now run on standard coupons: 100 bp for investment-grade (IG) names, 500 bp for high-yield (HY) names. But a borrower’s fair spread is rarely exactly 100 or 500 — so the difference is squared up with a one-time upfront payment at inception, sized so the contract’s initial value is zero. If a name’s fair spread is 250 bp but the contract pays a fixed 100 bp coupon, the protection buyer pays an upfront amount to make up the shortfall. Standard coupon + upfront ≈ the old custom spread, but now contracts with the same maturity are fungible.
• The ISDA Standard Model. A single, public pricing model converts between upfront and spread consistently for everyone, using fixed assumed recovery rates for the conversion — conventionally 40% for senior debt and 20% for subordinated debt. (That’s a pricing convention; the actual settlement recovery still comes from the auction.)
• Mandatory auction settlement. The credit-event auction we met earlier became the hardwired settlement mechanism — cash settlement at the auction price is now standard, rather than each pair physically delivering bonds.
• Determinations Committees. The DCs were formalized to rule on credit events and trigger auctions, so the “did a default happen?” question has one authoritative answer.

A follow-up, the “Small Bang” in July 2009, extended the auction mechanism to cover restructuring credit events too (which are messier because there’s no single “defaulted bond” to price).

One subtlety the Big Bang did not fully erase: the restructuring clause. Because restructuring is a soft, contestable trigger, contracts specify which variant applies, and they differ by region:

The headline to remember: North American IG/HY corporates trade No-R (XR) — restructuring is not a trigger, only bankruptcy and failure-to-pay are — while Europe trades Mod-Mod-R (MMR), keeping restructuring in but bounding the deliverables. Two identical-looking CDS on the same firm can pay out differently purely because of this clause.

Why trade a CDS?

A contract this fiddly only exists because it does jobs nothing else does as cleanly. Four big ones:

• Hedge a bond you own. You hold $10M of a company’s bonds and want the yield, but you’re nervous about default. Buy $10M of protection: now if the issuer defaults, your bond loss is offset by the CDS payout. You’ve stripped out the default risk while keeping the position. This is the original, “insurable interest” use — actual fire insurance on a house you actually own.
• Express a bearish view without shorting bonds. Shorting a corporate bond is a pain: you must borrow the bond, pay to hold the short, and bonds are often illiquid. Buying CDS protection (“naked,” with no bond) gives you the same I-think-this-name-deteriorates exposure cleanly — your premium is capped and known, and you cash in if credit worsens. This is the fire-insurance-on-your-neighbour’s-house trade from the intro: a directional credit bet.
• Earn synthetic credit exposure by SELLING protection. Want to be long a credit but don’t want to buy and fund the actual bond? Sell protection. You collect the premium and take on the default risk — economically almost identical to owning the bond (you earn carry, you lose on default), but with no upfront cash outlay to buy the bond and no funding. Selling CDS protection ≈ a leveraged long position in the credit.
• Trade the bond-CDS basis. A bond’s credit spread and its CDS spread should be roughly equal — they’re two prices for the same default risk. When they diverge, the gap is the basis (CDS spread minus bond spread). Relative-value desks trade the basis: buy the cheap one, sell the rich one, and harvest the convergence. The very existence of a clean, liquid CDS price gives the cash bond a benchmark to be measured against.

One catch that earns its own future lesson: when you buy protection, your payout depends on the seller actually being solvent enough to pay you. If the seller blows up in the same crisis that triggers your credit event, your “insurance” is worthless. That counterparty risk — vividly demonstrated when AIG, a giant protection seller, nearly took the system down in 2008 — is exactly what the XVA adjustments (CVA and friends) exist to price, and we’ll get there.

Putting it together

A credit default swap is tradable insurance on a borrower’s credit, and like every swap it has two legs: the buyer pays the premium / running-spread leg (quarterly, in bp of notional, on the 20th of Mar/Jun/Sep/Dec), and the seller pays the protection / contingent leg — worth notional × (1 − Recovery) — but only if an ISDA credit event (bankruptcy, failure to pay, restructuring, plus sovereign extras) is declared by the Determinations Committee. Settlement is overwhelmingly cash at the auction-determined recovery, replacing the old physical delivery. The buyer is long protection / short credit (small steady cost, big payout in a crisis); the seller is the mirror image, like an insurer wearing a fat tail. The 2009 Big Bang standardized everything — fixed 100 (IG) / 500 (HY) bp coupons plus an upfront, the ISDA Standard Model (40%/20% assumed recovery), mandatory auctions, and formal DCs — with the Small Bang adding restructuring auctions, and restructuring clauses (No-R in North America, Mod-Mod-R in Europe) still differing by region. And the reason it all exists: hedging bonds you own, expressing bearish views without shorting, manufacturing synthetic long exposure by selling protection, and trading the bond-CDS basis — all shadowed by the counterparty risk that XVA was invented to price. Master this, and the rest of credit derivatives is just clever combinations of the contract you now know cold.

Next — we’ll move from the single-name CDS you now know cold to CDS indices like CDX and iTraxx: baskets of names that trade as one liquid instrument, and the workhorse of credit hedging and macro positioning.