Credit Derivatives and Default Swaps: The Basics
Credit derivatives are where Chapter 3 begins, and they’re where the CDO story gets complicated. These instruments – primarily credit default swaps – turned the credit market from a buy-and-hold business into a trading business. They made the CDO market possible at the scale it reached. They also introduced risks that many participants didn’t understand.
Chapter 3 is long. This post covers the first half: portfolio risk, how CDSs work, the language disputes that cost people money, settlement mechanics, convertible bond oddities, and how recovery rates are estimated.
Three kinds of credit risk in a portfolio
Tavakoli opens with a useful framework. There are three distinct risks when holding bonds or loans:
Downgrade risk – the issuer’s credit quality deteriorates, the rating agencies downgrade the debt, and the price falls accordingly. Rating agencies publish “transition matrices” showing historical probabilities of moving from one rating to another over time.
Spread widening risk – even without a rating change, the market’s required yield for a given type of debt increases, pushing prices down.
Default risk – the issuer stops paying. The position recovers some fraction of face value.
All three types affected CDOs backed by subprime mortgages in 2005-2007. The BBB tranches that had been rated based on historical corporate debt data were attached to collateral – subprime mortgage loans – with much higher default probabilities and lower recovery rates than the models assumed. By the time derivatives indexes linked to these tranches were trading in the low 50s on a par-of-100 scale, the market was pricing in genuine losses reaching those tranches.
Portfolios also have market risk (interest rate moves) and currency risk (for multicurrency portfolios). Synthetics like CDSs, TRS, and credit spread options can hedge these risks or add to them.
How a credit default swap works
A CDS is a bilateral contract. The protection buyer pays a periodic fee (the premium) to the protection seller. If a specified credit event occurs on the reference entity, the protection seller makes a payment to the protection buyer.
Protection buyer = short credit risk = like selling a bond
Protection seller = long credit risk = like buying a bond (but unfunded)
Key point: unlike insurance, the protection buyer does not have to own the underlying obligation and does not have to suffer an actual loss to get paid. This is what makes CDSs different from insurance legally, and it’s what allows them to be used for speculative positions, not just hedging.
CDSs are over-the-counter transactions. The terms are private and negotiable. “Market standard” ISDA documentation exists, but it is not the only option, and many disputes have arisen precisely because people assumed the standard language covered things it didn’t.
The premium is paid quarterly in arrears in most current contracts. The level is driven by the credit spread of the reference entity – roughly the spread a bond investor would demand to take the default risk.
The basis risk problem: language matters enormously
ISDA has been updating its CDS documentation since the market began, and the debates are not academic. Each language change was driven by real losses in real disputes.
Since 2003, disputed areas have included:
- Bankruptcy and repudiation/moratorium definitions
- Restructuring – none, full, modified (U.S.), modified-modified (Europe)
- What counts as a qualifying reference obligation
- Successor definitions when companies merge
- What counts as a qualifying affiliate
- How to handle novations (contract assignments)
- ABS-specific contract terms
- Settlement method: cash, physical, or pre-agreed amount
Different regions use different standard language. U.S., European, Asian, and Middle Eastern markets have diverged. This creates basis risk: if you’re hedging a position in one market with a CDS written under another region’s standard language, you may not be fully hedged.
The key areas where basis risk enters:
- Credit event triggers (what counts as a default?)
- Settlement mechanics (how is the recovery value determined?)
- Deliverables (what can be delivered to settle the contract?)
Physical settlement vs. cash settlement
In physical settlement, when a credit event occurs, the protection buyer delivers the defaulted bonds (or loans) to the protection seller and receives par. The protection seller ends up holding the defaulted asset at whatever the recovery value turns out to be.
In cash settlement, no asset changes hands. The protection seller pays par minus the recovery value (determined by a dealer poll). The recovery value is typically the average of three to five dealer quotes taken 30 days after the credit event.
Tavakoli has consistently recommended writing cash settlement as an option to physical settlement. Here’s why:
The notional amount of CDSs on a given reference entity can easily exceed the actual outstanding bonds or loans. If everyone with CDSs tries to physically settle at the same time, there aren’t enough bonds to go around. This drives the bond price up, reducing the payment to protection buyers. This is exactly what happened with Delphi and Collins & Aikman defaults.
ISDA eventually introduced a voluntary auction process for settling these squeezes. But Tavakoli’s view: the individual solution (negotiating your own settlement language in advance) may be better than the voluntary process, which was essentially an admission that ISDA’s “standard language” had failed.
For very illiquid assets, neither physical settlement nor a simple dealer poll works well. An auction method is preferable – an explicit auction for the reference asset, with the option for the receiver to take physical delivery if they don’t like the auction price.
Digital, binary, and fixed recovery settlement
There are alternatives to par minus market value.
All-or-nothing (zero-one): Protection seller pays the entire notional if there’s a credit event. The protection buyer gets 100 cents on the dollar. This introduces massive basis risk for corporate credits – recovery rates are rarely zero – and generally makes no sense for CDOs.
Fixed recovery settlement: Pre-agree on a specific cash payment percentage. For instance, the 1997 Glacier Finance deal by UBS/SBC set some payouts at 51% of par (the Moody’s standard average recovery rate at the time for senior unsecured obligations). This makes sense when there’s no specific reference asset, or when the protection seller wants to eliminate the risk of being tagged with the cheapest-to-deliver option, or when daily price volatility after an event is too high to trust a dealer poll.
For emerging market debt, this is particularly relevant. If a sovereign defaults, the currency often collapses simultaneously, making post-default valuations extremely volatile. Agreeing in advance on a payment amount removes that uncertainty.
The cheapest-to-deliver option
This is one of the more important concepts in credit derivatives, and it costs investors money regularly.
When a credit event occurs and physical settlement applies, the protection buyer can typically deliver any qualifying obligation of the reference entity. If there are multiple deliverable instruments (bonds at different maturities, loans, convertibles), the protection buyer will deliver the cheapest one – the one trading at the lowest price.
From the protection seller’s perspective, they’re on the hook for par, but they receive whichever instrument is cheapest. This difference is the cheapest-to-deliver option, and it has real value.
Deal arrangers at banks understand this. In synthetic CDOs, where the CDO is selling credit protection to the arranging bank, arrangers try to negotiate the broadest possible deliverable language for the CDOs they structure. They want to be able to deliver the cheapest instrument. When the arranger is the protection seller (selling credit protection elsewhere), they try to limit deliverables to the most expensive ones.
CDO investors should either limit the range of deliverable obligations or demand a higher spread to compensate for the cheapest-to-deliver option. Many don’t do either, because they don’t read the documentation carefully enough.
Convertible bonds: a special problem
Convertible bonds – corporate bonds that can be exchanged for equity shares at a preset price – are deliverable in many CDS contracts. This creates complications.
When the stock price is high relative to the conversion price, the convert behaves like equity. When the stock is low, the fixed-income characteristics dominate. The conversion premium tells you how far out of the money the equity option is.
Hedge funds (convertible arbitrageurs) buy convertibles in large quantities – estimated at over 50% of new issuance. They often strip the equity option out, leaving a deeply discounted zero-coupon bond, and sell that bond as a callable par asset swap. The bond is callable by the original arbitrageur (who still holds the equity warrant) – they can call it back at any time and force the asset swap investor out.
The protection for the asset swap investor: a lockout period (usually six months minimum) during which the arbitrageur can’t call. This is negotiated; earlier market practice often didn’t include a lockout at all.
Convertibles deliver into CDS contracts and can cause significant basis risk. After Enron’s default, deeply discounted stripped convertibles recovered at about the same rate as ordinary bonds. Moody’s is still cautious about convertibles as deliverables and adds stress tests to CDO models that allow them. Typically CDOs that allow convertibles as deliverables limit them to 10% of the portfolio.
Negative basis trades
Supply and demand affect CDS pricing in ways models can’t always predict.
Example from the book: Fiat issued a $2.2 billion convertible in December 2001, convertible into GM shares. Arbitrageurs bought CDS protection on Fiat in anticipation – and the CDS spread widened from about 130 bps to 190 bps in just over a week. The cash Fiat bonds at the same maturity remained unchanged because they were locked in buy-and-hold portfolios. The CDS was suddenly trading at a dramatically wider spread than the equivalent cash bond.
This created a negative basis trade opportunity: buy a five-year Fiat par asset swap (earning LIBOR plus cash bond spread), buy five-year CDS protection on Fiat (paying 190 bps), and lock in a positive net spread. You’re fully hedged on credit risk and earning the difference.
Basis trades like this appear regularly whenever CDS spreads and cash bond spreads diverge significantly.
Default and recovery rates: what the models use
Two key inputs for any CDS or CDO valuation: probability of default and recovery rate in the event of default.
Rating agencies publish cumulative default rate tables by rating category. This data is primarily U.S. and European corporate credits. It’s used to estimate default probabilities for credits with equivalent ratings – a shortcut that works reasonably well for similar asset types but breaks down when applied to novel asset classes like subprime mortgages.
Recovery rates are more variable. Moody’s defines loss severity as par minus recovery rate, measured at market value approximately 30 days after default.
Key factors in recovery:
- Seniority – secured loans typically recover much more than senior unsecured bonds, which recover more than subordinated debt
- Loan covenants and insider access – banks with lending relationships have better information and more control over restructuring
- Asset quality – in recent years, excessive leverage in both mortgage lending and corporate leveraged buyouts has pushed recovery rate assumptions down
Tavakoli’s practical advice: don’t just use rating agency data. Do your own due diligence on the underlying collateral and form your own estimates of default probability and recovery value. Historical data is always backward-looking, and it doesn’t capture novel collateral types or changed underwriting standards.
The expected loss formula
Once you have probability of default and recovery rate, expected loss is straightforward:
Expected Loss = Probability of Default × (1 - Recovery Rate) × Notional
This is a mean – an average. For worst-case scenario analysis, use worst-case default probability and worst-case recovery rate.
The formula is simple. The inputs are not.
The second half of Chapter 3 covers total return swaps, synthetic CDOs, pricing, information asymmetry, pay-as-you-go contracts for ABS, and the indexes that eventually let hedge funds bet against the subprime market.
Previous: Trusts, Conduits, and the Myth of Bankruptcy-Remote Entities
Next: Total Return Swaps, Synthetic CDOs, and Credit Indexes