CDS Indices: CDX & iTraxx

You already know how a single-name credit default swap works: you pay a periodic premium, and if one named company defaults, your counterparty makes you whole on the loss. Lovely. Now suppose you want exposure not to one company but to a hundred and twenty-five of them at once — because you have a view on “credit” as an asset class, or you want to hedge a whole bond portfolio, or you just want a clean, liquid thermometer for how scared the credit market is today. Trading 125 separate single-name CDS — 125 negotiations, 125 tickets, 125 different liquidity puddles — is an operational nightmare. So the market did the obvious thing: it bundled them into one contract with one ticker and one spread. That bundle is a CDS index, and the two giant families are CDX (North America) and iTraxx (Europe and Asia). Think of it as the S&P 500 of credit: one trade, instant diversified credit exposure.

From single names to a basket

The analogy. An equity index like the S&P 500 takes 500 stocks and squashes them into one number you can trade. A CDS index does the same thing for credit: it takes a fixed roster of companies — each with its own single-name CDS — and packages them into one standardized contract. Selling protection on the index is economically like owning a diversified portfolio of corporate credit: you collect premium, and you take a small hit each time one of the names in the basket defaults. Buying protection is the hedge: you pay premium and get paid when names blow up.

The precise definition. A CDS index is a standardized basket of single-name CDS on a fixed list of reference entities, traded as a single contract with one spread and one fixed coupon. The two families you must know:

A few words to define on first use. Investment grade (IG) means a relatively safe credit rating (think BBB− and above); high yield (HY) or sub-investment grade means riskier, lower-rated borrowers (BB+ and below) who pay more to compensate — which is why their indices carry the fat 500 bp coupon instead of 100. Crossover is the European name for the sub-IG basket (the “crossover” zone straddling the IG/HY boundary). One basis point (bp) is 0.01%.

Equal weighting. This is the bit people get wrong. Unlike the S&P 500, which weights companies by market value (Apple counts far more than a small-cap), the standard CDS index is equally weighted: every name is exactly 1/N of the index notional. In CDX.NA.IG that’s 1/125 = 0.8% per name; in CDX.NA.HY it’s 1/100 = 1.0%; in iTraxx Crossover it’s 1/75 ≈ 1.33%. No name dominates — a default by the biggest borrower in the basket hurts exactly as much as a default by the smallest.

How an index trades

Fixed coupon plus upfront. Just like single-name CDS after the 2009 “Big Bang” standardization, index contracts trade with a fixed coupon — 100 bp for the IG indices, 500 bp for the HY/Crossover ones — so that every contract of a given series is identical and fungible. But the fair spread for the basket moves around every second. The market reconciles the fixed coupon with the moving fair spread using an upfront payment: a lump sum exchanged at the start that trues up the difference between the coupon you’re contractually paying and the spread you should be paying.

• Fair spread below the coupon → the protection buyer is overpaying on the running coupon, so they receive an upfront.
• Fair spread above the coupon → the protection buyer is underpaying on the coupon, so they pay an upfront.

Quoting convention. IG indices, where spreads are tight, are quoted in spread (e.g. “CDX IG at 72”). HY indices, where spreads are wide and the contract behaves more like a bond, are quoted in price (e.g. “CDX HY at 105.5,” meaning a price per 100 of notional, like a bond). Same machinery underneath; just a different dialect for IG versus HY.

What a default does. Here’s the mechanical heart of it. When one name in the basket suffers a credit event (default, bankruptcy, etc.), only that name’s 1/N slice triggers. The protection seller pays out on that slice, the defaulted name is removed, and the index notional shrinks by 1/N. The rest of the contract keeps running on the surviving names. So a single default is a partial event — it doesn’t tear up the whole contract, it just lops off one equally-weighted sliver.

Worked example — a default in CDX.NA.IG. Say you sold protection on $125 million of CDX.NA.IG (125 names). One name defaults with a recovery rate of 40% (recovery = the fraction of face value bondholders get back; so the loss given default is 1 − 0.40 = 0.60).

So one default cost the seller (1 − R) on a 0.8% slice — $600,000 — and the contract rolls on against the surviving 124 names on $124 million. The squared-impact drama of variance swaps has no analog here: each name’s contribution is capped at its 1/N slice times its loss given default.

The roll

The analogy. Treasury bonds have an “on-the-run” issue — the most recently auctioned, most liquid bond of each maturity — and a tail of older “off-the-run” bonds that still trade but thinner. CDS indices work the same way, except the refresh is on a calendar: a brand-new series rolls out every six months.

The precise definition. Twice a year — around 20 March and 20 September — the provider launches a new series of each index, with a new series number (Series 41, 42, …). At each roll, the roster is rebalanced: names that have defaulted, been downgraded out of the eligibility bucket, or lost liquidity are dropped, and fresh eligible names take their place. The just-launched series is on-the-run — the most liquid, the benchmark everyone quotes and trades. The previous series don’t vanish; they become off-the-run and keep trading all the way to their own maturity, just with thinner volume and wider bid-ask spreads.

Why liquidity concentrates. New trading flow — new hedges, new positions, dealer market-making — naturally gravitates to the on-the-run series because that’s where everyone else is, so spreads are tightest there. It’s a self-reinforcing Schelling point: people trade the liquid series because it’s liquid, which keeps it liquid.

Rolling a hedge. If you’re using an index to hedge and you want to stay in the liquid, on-the-run contract, you roll your position at each roll date: close the maturing/off-the-run series and reopen the same exposure in the new on-the-run series. That keeps your hedge cheap to trade and tightly priced — at the cost of a small transaction each roll.

The index basis (skew)

The analogy. An ETF can trade at a slight premium or discount to the net asset value of the stocks it holds. A CDS index does the same: the price you pay for the bundle doesn’t have to exactly equal the cost of buying all the pieces. That gap has a name.

The precise definitions. There are two spreads to compare:

• Intrinsic spread — what the index should cost based on its parts: roughly the duration-weighted average of all the constituent single-name CDS spreads. (Duration-weighted because names with longer risky duration contribute a touch more, but “average of the single-name spreads” is the right mental model.)
• Traded index spread — what the index actually trades at in the market as one liquid contract.

The index basis (also called the skew) is the difference:

$\text{basis} = \text{traded index spread} - \text{intrinsic spread}$

A positive basis means the index trades wider (more expensive protection) than its parts; a negative basis means it trades tighter (cheaper) than its parts.

Worked example. Suppose the 125 single-name CDS in an IG index average 62 bp (that’s the intrinsic), and the index itself is quoted at 66 bp (that’s the traded). Then:

$\text{basis} = 66 - 62 = +4 \text{ bp}$

A +4 bp basis: the bundle is 4 bp pricier than the sum of its parts. Small, as bases usually are — but not nothing.

The interactive below makes the gap visible. Each bar is one constituent single-name spread, sorted; the dashed line is the intrinsic average and the solid line is the traded index spread. Drag the “Index demand / technicals” slider — pushing it positive (more demand to buy index protection) lifts the traded line above the intrinsic dashed line, opening a basis; pushing it negative drives the index tighter than its parts. Flip the IG-like / HY-like preset to see how a wider, riskier basket behaves. Read the basis off the bottom as traded − intrinsic.

Why a basis exists. If the parts and the bundle were always identical, the basis would be zero. They aren’t, because of:

• Liquidity and technicals — the index is far more liquid than the average single name, so supply/demand for the bundle (a wave of macro hedgers buying index protection, say) can push the index spread around independently of the underlyings.
• Who’s hedging — demand to hedge “credit broadly” lands on the index, not on 125 separate names, so flow imbalances show up as basis.
• Composition and convexity quirks — duration weighting, differing recovery assumptions, and the upfront/coupon mechanics introduce small wedges.

The index arbitrage. Because the basis should be small, a big basis is a trade. If the index trades far wider than its intrinsic (large positive basis), an arbitrageur buys the cheap thing and sells the rich thing: buy protection via the single names (cheap) and sell protection on the index (rich), pocketing the gap and waiting for it to converge. If the index is far tighter than its parts (large negative basis), do the reverse — buy the index, sell the names. This index-versus-single-names arbitrage is exactly the force that keeps the basis small most of the time; it only blows out when the arb is hard to put on (a crisis, a liquidity crunch, dealers unable to warehouse 125 single-name legs).

A first look at index tranches

The analogy. Imagine the index’s total losses as water filling a tank from the bottom. You can sell someone the right to the bottom few inches (they get soaked first), someone else the middle, and someone else the top (they only get wet in a catastrophe). Each horizontal band is a tranche — a standardized bet on one slice of the portfolio’s losses.

The precise definition. An index tranche is a contract referencing a specific loss band of the index, defined by an attachment point (where its losses start) and a detachment point (where they stop). The common teaching set on CDX.NA.IG carves the 0–100% loss range into four slices:

Losses fill from the bottom up: defaults eat the equity tranche first, and a higher tranche is only touched once every tranche below it has been fully consumed. So the equity tranche = first loss (maximum risk, maximum premium) and the super-senior tranche = last loss (it only suffers in a genuine credit catastrophe).

Why bother slicing? Because different investors want different risk. A yield-hungry investor sells the equity tranche to collect a rich premium and bet defaults stay rare; a conservative one sells the super-senior to earn a thin premium for taking only catastrophe risk. The full mechanics — how correlation prices each slice, why the equity tranche is a correlation bet, the infamous “correlation smile” — are the subject of the next two lessons. For now, just lock in the shape: tranches turn one index into a menu of risk slices, ordered from first loss to last.

Putting it together

A CDS index packages a fixed roster of single-name CDS into one liquid, standardized contract — the S&P 500 of credit. The headline families: CDX.NA.IG (125 IG names, 100 bp coupon, quoted in spread), CDX.NA.HY (100 HY names, 500 bp coupon, quoted in price), iTraxx Europe Main (125 IG names), and iTraxx Crossover (75 sub-IG names). Every name is equally weighted at 1/N of notional, so selling index protection ≈ owning a diversified credit portfolio. Trades use a fixed coupon plus an upfront to reconcile the standard coupon with the moving fair spread; when a name defaults, only its 1/N slice triggers — the seller pays (1 − recovery) on that sliver and the index notional shrinks. A new series rolls every six months (~20 March / 20 September) with a rebalanced roster; the newest series is on-the-run (most liquid), older ones off-the-run (still trading, thinner), and you roll a hedge to stay in the liquid one. The index basis (skew) is traded − intrinsic — the gap between the index’s own price and the duration-weighted average of its parts — kept small by index-versus-single-names arbitrage, and a big basis is itself a trade. Finally, the index’s 0–100% loss range can be carved into tranches — 0–3% equity (first loss) up to 15–100% super-senior (last loss) — a first look at the standardized correlation products we dissect next.

Next — index tranches in full: we’ll take the bottom-up loss waterfall you just met and show how correlation prices each slice, why the equity tranche is really a bet on defaults clustering together, and how the whole structure echoes the securitization machinery at the heart of this topic.