Credit Derivatives

Cumulative default curves are credit's actuarial tables — and the gap between the IG and HY curves is what the spread pays for.

• High yield front-loads: the first three years after issuance are the most dangerous.
• Ratings lag markets — spreads blow out before downgrades arrive.
• Loss = default × (1 − recovery): a loan recovering 70 hurts less than a bond recovering 30.
• The CDS calculator above runs this in reverse — implied default rates out of spreads.

Spreads grind tighter for years and explode for weeks — the cycle has a script, and "average spread" is a fiction.

• Late cycle: complacency compresses spreads below any sensible loss estimate.
• Panic: spreads overcompensate for depression-grade defaults — capital is scarce exactly when compensation is best.
• Discipline = sizing across the asymmetry: less risk when spreads say "nothing can go wrong", liquidity ready for when they say "everything already has".

A credit instrument's true identity is its place in the repayment queue — spread, seniority and documents together, or you've read half a sentence.

• The queue: secured loans → senior unsecured → subordinated → hybrids (AT1) → equity.
• Historic recoveries down the stack: ~65% secured, ~40% senior unsecured, ~25% subordinated, ~0 below.
• Same company, different risk: a good loan and a bad bond can share an issuer.
• Structural subordination: lend to the holding company and you stand behind every creditor of the subsidiaries that own the assets.
• The queue is contestable now — J.Crew/Serta-style document games rearrange it mid-life; document quality trades as its own factor.
• Stress question: at what enterprise value does my layer stop being covered? That's credit's strike price.

Thirty years from an insurance workaround to a systemic institution and back:

• 1994 — JPMorgan structures the first CDS (Exxon's Valdez credit line): default risk becomes transferable.
• 1997 — BISTRO: the first synthetic securitisation — the CDO era's blueprint.
• 2004 — CDX and iTraxx standardise index trading; credit becomes a macro asset class.
• 2005 — the GM/Ford correlation crisis: the first warning that tranche models mis-handle idiosyncratic shocks.
• 2008 — AIG: $500bn+ of sold protection without collateral; the poster child of the crisis.
• 2009 — the "Big Bang" protocol: fixed coupons, auctions, central clearing — CDS grows up.
• 2012 — the London Whale loses $6bn in index tranches at a bank hedging itself.
• 2016–now — the liability-management era: J.Crew, Serta; documents become the battlefield. SRT/capital-relief trades revive synthetic tranching as a regulated tool.

• ISDA and the Determinations Committees: regional committees of dealers and buy-side firms rule on whether a credit event happened — the closest thing this market has to a court, and the trigger for every CDS payout.
• Auctions: after a credit event, an industry auction sets one recovery price so all contracts settle at the same number instead of thousands of bilateral arguments.
• Index sponsors: S&P Global (Markit) constructs and rolls the CDX and iTraxx families — the indices through which most credit macro risk trades.
• Clearing: ICE Clear Credit clears the bulk of index and single-name CDS.
• Loan market bodies: the LSTA in the US and the LMA in Europe standardise loan documentation and settlement — slow, manual, and improving.
• CLO managers and rating agencies: the buyers who set marginal demand for leveraged loans, and the agencies whose tranche ratings determine what those buyers may hold.

The sentence that prevents most credit mistakes: the spread is not the return. Subtract expected loss (default probability times one minus recovery) before calling anything cheap — the calculators above do it in seconds.

1. Where in the capital structure, and behind whom? Structural subordination hides behind clean-looking senior labels.
2. Does the spread cover the expected loss? Run the credit triangle first (calculator above), then judge what is left as compensation.
3. Read the documents. Covenants, permitted baskets, unrestricted-subsidiary definitions and voting thresholds decide what happens in stress.
4. When does the borrower need markets again? The maturity wall, not the current ratio, is what kills companies.
5. At what enterprise value does my layer break? That is the credit equivalent of a strike price — know it before you need it.
6. Who else owns it? Index membership and rating-boundary holders create forced sellers on downgrade.

• A rating used as a substitute for analysis — ratings lag spreads, and 2008 demonstrated their limits at the tail.
• Cov-lite paper priced as if it were covenanted: the missing tripwires are worth real basis points.
• Complexity that grows with each layer — re-securitisation multiplied labels, not safety.
• Correlation assumed away: pooling only diversifies risks that are genuinely independent.
• Documents that permit asset transfers out of the collateral pool — the J.Crew lesson, now a standard negotiating point.
• Liquidity assumed in a market that trades by appointment — loan settlement is measured in weeks, not days.