Credit Default Swap

Anatomy of the contract

A credit default swap is a bilateral contract documented under the ISDA Master Agreement. It involves three economic parties — a protection buyer, a protection seller, and a named reference entity — and one third party that is contractually invisible: the issuer of the actual bonds. The reference entity is never a party to the CDS. A bank can buy or sell protection on Ford Motor Company's debt without Ford knowing, agreeing, or being notified. That detachment from the reference entity is one of the most important and most controversial features of the instrument.

The mechanics are simple in summary and intricate in detail. The buyer pays a periodic premium, expressed as an annualised percentage of notional — the CDS spread, quoted in basis points (1 bp = 0.01%) and paid quarterly on standardised IMM dates (March 20, June 20, September 20, December 20). On a $10 million notional CDS at a 200 bp spread, the buyer pays roughly $50,000 each quarter. In return, the seller is obligated to make a contingent payment if a credit event occurs at the reference entity before the contract's scheduled maturity.

Three credit events are defined under standard 2014 ISDA Credit Derivatives Definitions: bankruptcy (the entity files for insolvency protection), failure to pay (a missed payment past the grace period), and restructuring (a unilateral, distressed change to the terms of debt). The first two are uncontroversial; "restructuring" has been the subject of years of contractual wrangling, including a famous 2014 episode in which Goldman Sachs and other dealers disagreed about whether Argentina's failure to pay a holdout-driven judgment constituted a credit event. ISDA's Credit Derivatives Determinations Committees — regional panels of dealers and buy-side firms — now formally rule on these questions, with a published voting record.

JPMorgan, 1994, and the Exxon Valdez

The CDS was invented at JPMorgan in 1994 to solve a specific balance-sheet problem. JPMorgan held a $4.8 billion credit line to Exxon Corporation, opened to help Exxon pay damages and legal fees from the 1989 Exxon Valdez oil spill. The credit line was profitable to maintain but consumed a large amount of regulatory capital under Basel I, which assigned a flat 100% risk weight to corporate exposures regardless of credit quality. A team led by Blythe Masters in JPMorgan's swaps group designed a contract under which the European Bank for Reconstruction and Development (EBRD) — for an annual fee — would assume Exxon's credit risk. JPMorgan kept the relationship and the income; EBRD took on credit risk it wanted as part of its mandate; and JPMorgan freed up regulatory capital.

The structure was generalised over the next several years. Standardisation of documentation began with the 1999 ISDA Credit Derivatives Definitions, and a market that traded almost nothing in 1996 was running into the trillions of dollars of notional by 2001. The 1998 Russia default and the LTCM collapse acted as a stress test that the young market mostly survived; that survival validated the model and accelerated growth.

Pricing — the (1 − R) × λ approximation

The fair par spread is set by a no-arbitrage condition: the present value of the premium leg paid by the buyer equals the present value of the protection leg paid by the seller. Under a constant default intensity λ (probability per unit time that the entity defaults, given that it has survived so far) and a known recovery rate R, the closed-form first-order approximation is

spread (per annum) ≈ (1 − R) × λ

The intuition is direct. The seller's expected annual loss equals the chance of default this year, λ, multiplied by the loss given default, 1 − R. The fair premium just compensates that loss. The risk-free discount rate cancels at first order because both legs are discounted by the same curve.

The full ISDA Standard Model (the open-source pricing convention released alongside the 2009 Big Bang Protocol) replaces the constant-λ approximation with a piecewise-constant hazard-rate term structure calibrated to a strip of CDS quotes at multiple tenors. Standard market convention assumes R = 0.40 for senior unsecured corporate debt, R = 0.20 for subordinated, and R = 0.25 for emerging-market sovereigns; these conventions matter because they affect the implied default probability that traders read off market spreads.

Worked example — what does a 300 bp spread imply?

Suppose the 5-year senior CDS on a B-rated industrial issuer trades at 300 bp. Assume the market convention R = 0.40 for senior unsecured corporate debt.

spread = (1 − R) × λ
300 bp = 0.60 × λ
λ = 300 bp / 0.60 = 500 bp = 0.0500 per annum

5-year cumulative default probability:
P(default in 5 yr) = 1 − exp(−λ × 5)
                   = 1 − exp(−0.25)
                   ≈ 22.1%

So a market spread of 300 bp on a 5-year contract implies — under the standard recovery assumption — roughly a 22% cumulative chance of default over five years. The "risk-neutral" qualifier is important: this probability is calibrated to make the contract fairly priced under a martingale measure, not necessarily the real-world probability. Empirical default frequencies for B-rated issuers historically come in lower; the wedge between risk-neutral and physical default probabilities is a credit-risk premium.

Single-name vs index CDS

The market splits cleanly into two categories:

Feature	Single-name CDS	Index CDS
Reference	One named issuer	Fixed equally-weighted basket
Examples	Ford, AT&T, Italy	CDX.NA.IG, iTraxx Europe
Coupons	Fixed 100 or 500 bp + upfront	Fixed coupon + upfront
Liquidity	Moderate, name-dependent	High — the most liquid credit instrument
Default settlement	Auction per entity	Constituent auction; index rolls
Standard tenors	5y dominates; 1, 3, 7, 10y trade	5y dominates
Notional share (2020s)	~30%	~70%

CDX.NA.IG (125 North American investment-grade names, equally weighted) and iTraxx Europe Main (125 European IG names) are the workhorses. Markit, now S&P Global, administers them; new on-the-run series roll every six months, and the index level is quoted in basis points just like a single-name spread. iTraxx Crossover (75 European sub-IG names) trades at much wider spreads and is the primary hedge for European high-yield credit funds. CDX.NA.HY is the North American high-yield index analogue.

What CDS are actually used for

• Hedging bond positions. A bank holding a $50 million loan to a single corporate borrower can buy single-name CDS on that borrower, transferring the credit risk to the protection seller while keeping the loan on its balance sheet (and the customer relationship intact). This was the original use case.
• Expressing a short credit view without owning bonds. Buying naked CDS protection is the cleanest way to express the view that an issuer's credit will deteriorate. Shorting bonds requires a securities-lending counterparty, has limited supply, and exposes the trader to short-squeeze risk; buying CDS is just a derivative trade.
• Capital relief and synthetic securitisation. Banks bought protection from highly-rated counterparties (monoline insurers, AIG, foreign banks) to reduce risk-weighted assets under Basel II's standardised and internal-ratings-based approaches. Synthetic CDOs, where tranched losses on a CDS portfolio are sold to investors, were a major pre-2008 product.
• Basis arbitrage. The cash-CDS basis — the difference between a bond's spread over the risk-free rate and its CDS spread — is rarely zero. Hedge funds capture the basis by going long the bond and long CDS (positive basis) or short the bond and short CDS (negative basis), funding the position in repo.
• Index macro trades. Buying or selling iTraxx Crossover protection is one of the simplest ways for a macro fund to express a view on European credit conditions without picking individual names. Index CDS are also used to hedge underwriting pipelines at investment banks.

2008 — how CDS broke AIG

The CDS market was a stress amplifier rather than a cause of the 2008 crisis, but one institution stands out: American International Group's London-based AIG Financial Products subsidiary. AIGFP had written roughly $1.6 trillion notional of CDS, of which about $440 billion referenced mortgage-backed securities and collateralised debt obligations. AIG's AAA rating allowed it to write protection at low spreads; the firm's view was that highly-rated structured-credit tranches would essentially never default.

Two features of the contracts proved fatal. First, the CDS were governed by ISDA collateral schedules with strict mark-to-market posting requirements: as the market value of the reference assets fell, AIG had to post collateral. Second, the schedules contained downgrade triggers: a downgrade of AIG itself (the protection seller) required additional collateral posting. As subprime-backed securities collapsed in price through 2007 and 2008, AIG's mark-to-market losses ballooned, and on 15 September 2008 — the same day Lehman Brothers filed for bankruptcy — Moody's and S&P downgraded AIG, triggering more than $30 billion in immediate collateral calls that AIG could not meet.

The Federal Reserve Bank of New York committed an initial $85 billion facility on 16 September; ultimately the rescue grew to $182 billion through the Maiden Lane II and III vehicles, which bought AIG's CDS counterparty exposures at par. The largest beneficiaries were Goldman Sachs ($14 billion), Société Générale ($14 billion), Deutsche Bank ($12 billion), and Merrill Lynch ($7 billion). AIG repaid all support to the Treasury by December 2012, and the rescue ultimately returned a modest profit to taxpayers — though the political and structural consequences reshaped financial regulation.

2012 — the Greek sovereign CDS test

Sovereign CDS — protection on government debt — received their stress test in the 2010–2012 European sovereign debt crisis. The flashpoint was Greece. As Greek 10-year yields rose from 5% in early 2010 to over 35% by early 2012, traders piled into Greek sovereign CDS. The question was whether the Greek debt exchange of March 2012, in which private holders accepted a 53.5% haircut on face value, would be ruled a credit event.

It eventually was. ISDA's EMEA Determinations Committee ruled on 9 March 2012 that the use of the Greek collective-action clause forcing holdouts to participate constituted a "restructuring credit event," triggering CDS payouts. The auction settled at a 21.5% final price, meaning protection sellers paid 78.5% of notional. Total payouts were modest in absolute terms — about $2.5 billion net — because gross CDS positions were small and largely offset within the dealer network. But the episode confirmed that sovereign CDS could in principle pay out, and it accelerated a political reaction: in November 2012, the European Union banned naked sovereign CDS for EU investors on EU sovereigns (Regulation No 236/2012).

Post-2009 reform — central clearing

The Big Bang Protocol of April 2009, followed by the European Small Bang Protocol of July 2009, fundamentally restructured the market. Three changes mattered most.

Standardisation. Spreads moved from name-specific running coupons to two fixed coupons (100 bp for IG-quality names, 500 bp for HY-quality), with the difference between the fair spread and the fixed coupon paid as an upfront cash payment. This made offsetting trades exactly cancellable, enabling enormous notional compression.

Determinations Committees. ISDA established regional Credit Derivatives Determinations Committees with binding authority to rule on credit events, succession events, and auction parameters. Before 2009, individual counterparties argued bilaterally about whether an event had occurred.

Central clearing. ICE Clear Credit (US) began clearing CDX index trades on 9 March 2009 and ICE Clear Europe followed; LCH.Clearnet briefly competed before exiting. A clearing house novates the trade — the CCP becomes counterparty to each side, holds variation and initial margin, and mutualises losses across members. By 2024, more than 90% of index CDS notional and a growing share of liquid single-name CDS notional cleared centrally. The Dodd-Frank Act in the US and EMIR in Europe made clearing mandatory for eligible CDS.

The combined effect was a sharp reduction in gross notional. Multilateral trade compression — orchestrated by TriOptima and LCH — netted offsetting positions that had previously been counted gross. Total CDS notional outstanding fell from $61.2 trillion at end-2007 to about $8 trillion in the mid-2020s, even as the market remained economically healthy.

Variants and adjacent products

• Loan CDS (LCDS). Protection on syndicated loan obligations rather than bonds; tighter recovery assumptions reflect the senior-secured position in the capital structure.
• First-to-default basket. Pays out on the first default in a small basket (typically 5 to 10 names). Cheaper than buying protection on each name; popular in retail-structured products in the mid-2000s.
• CDS option (credit swaption). Option to enter a CDS at a future date and pre-agreed strike spread. Used to express views on credit volatility and to hedge underwriting pipelines.
• Total return swap (TRS). A close cousin: one party pays the total return of a bond, the other pays a floating rate. Functionally equivalent to a financed long position, but documented as a derivative.
• Synthetic CDO. A portfolio of CDS protection (typically 100+ names), tranched into equity, mezzanine, and senior pieces. The pre-2008 ABS CDO was the structured-credit instrument at the centre of the Magnetar and Goldman ABACUS trades.

Common pitfalls

• Confusing notional with exposure. A $1 trillion notional CDS book does not mean $1 trillion of risk. Net positions after offsetting longs and shorts are typically a small fraction of gross notional. AIG's $1.6 trillion was a gross notional figure; the directional exposure that broke the firm was a much smaller subset, concentrated in mortgage-backed securities.
• Treating CDS spreads as default probabilities. CDS spreads embed a risk premium, a recovery assumption, and a liquidity premium. The implied risk-neutral default probability is higher than the physical (historical) probability — typically by a factor of 2 to 3 for investment-grade names.
• Forgetting auction-determined recovery. The "(1 − R)" in the spread formula is an assumption, not a contract term. Actual settlement recovery is determined at auction after the credit event; the auction price can differ markedly from convention. The 2009 CIT Group auction settled at 68.125% (vs the assumed 40%), reducing actual payouts.
• Ignoring counterparty risk pre-clearing. Bilateral CDS exposes the protection buyer to the credit of the protection seller. If the seller defaults, the buyer holds an unsecured claim. This is precisely why AIG's downgrades were existential — every counterparty needed AIG to remain solvent for the protection to be worth its face value.
• Mistaking restructuring conventions. European CDS typically include "modified-modified restructuring" (Mod-Mod-R) as a credit event; North American single-name CDS often exclude restructuring entirely under "Ex-R" or "No-R" conventions. The same underlying issuer can have different protection scopes depending on the contract's regional convention. Always confirm the credit-event suite before pricing.