Credit Derivatives

CDO & Synthetic Tranches

Also known as: Collateralised debt obligation, Synthetic CDO, Index tranche

Slicing a pool of credit risk into layers of first-loss and last-loss — the machine that concentrated 2008, and the tranche market that outlived it.

Asset class
Credit (structured)
Instrument type
Tranched claims on a credit portfolio (cash or synthetic)
Traded
OTC; index tranches on CDX/iTraxx are the liquid survivors
Typical users
Correlation desks, hedge funds, yield-hunting institutions
A mezzanine tranche's loss: untouched until portfolio losses reach the attachment point, wiped out by the detachment point.
AttachDetachMezzanine (3–7%)Equity (0–3%)SeniorPortfolio lossTranche loss
1 · SnapshotThe one idea to remember
Key intuition: tranching doesn't reduce risk — it sorts it. The senior investor is betting defaults come one by one; the equity investor is betting they don't come at all. The whole structure prices one question: do the names default independently, or together?
2 · BeginnerWhat is it, really?

Take a portfolio of 100 bonds or loans. Instead of selling investors a share of the whole pool, sell them layers of the losses. The equity tranche absorbs the first defaults (say, losses up to 3% of the pool) in exchange for a fat yield. The mezzanine takes the next slice (3–7%). The senior layers above only lose if defaults burn through everything beneath them — which is why rating agencies once stamped them AAA. That is a collateralised debt obligation.

The design is genuinely useful — it's how a pool of risky loans can fund itself partly at safe-asset prices (a CLO is exactly this, built on loans, and it worked fine through 2008). The catastrophe version was the ABS CDO: pools made of subprime mortgage bonds, re-tranched, then re-re-tranched ("CDO-squared"), with AAA labels multiplying while the underlying risk was the same correlated bet on US house prices. When that one bet failed, every layer failed together.

The synthetic variant skips the bonds entirely: the portfolio is just a list of names referenced through credit default swaps. No cash raised, no assets bought — pure transfer of tranche-shaped credit risk. Synthetics made the exposure infinitely replicable, which is how a mid-sized mortgage market generated outsized losses. It's also what survives today, in cleaner form: standardised index tranches on CDX and iTraxx.

3 · IntermediateHow it works in practice

The mechanics of a tranche

A tranche with attachment \(A\) and detachment \(D\) absorbs portfolio losses \(L\) between those points:

$$ \text{Tranche loss} = \frac{\min(\max(L - A, 0), D - A)}{D - A} $$

A 3–7% tranche on a 100-name pool (40% recovery): untouched through the first ~5 defaults, then each further default eats ~15% of the tranche; by ~12 defaults it's gone. Protection sellers receive a running spread on the surviving tranche notional, exactly like a CDS on the slice.

Correlation is the price

Spreads across the capital structure encode default correlation:

  • Low correlation: defaults arrive scattered → equity gets hit in almost every scenario (expensive), seniors are near-immune (tight).
  • High correlation: all-or-nothing world → equity sometimes survives untouched (relatively cheaper), seniors suddenly carry real tail risk (wider).

Equity is short correlation, senior is long it. Desks quote tranches in implied correlation the way options quote implied vol — and trading "the correlation smile" across attachment points is its own discipline.

The index tranche market — the liquid remnant

Standardised slices of CDX.IG / iTraxx Main (0–3, 3–7, 7–15, 15–100) trade with real two-way flow: hedge funds run equity-vs-senior relative value, dealers hedge correlation books, and "bespoke tranches" on custom name-lists revived quietly in the late 2010s as yield-starved buyers returned. Volumes are a fraction of 2007, documentation is cleaner, and the buyers now mostly know what convexity they're selling. Mostly.

Worked example — why AAA wasn't: a 2006 mezz ABS CDO bought BBB subprime bonds that each lose ~everything if national house prices fall 15%. Correlation among them: near 1 (same bet). Tranching a portfolio of one repeated bet changes nothing — every layer was the same BBB risk with a different label. The models assumed correlation ~0.3; the world delivered ~1. That gap, times $500bn, was 2008's structured-credit bill.
4 · AdvancedPricing & valuation

The Gaussian copula and its scar tissue

The market standard (Li, 2000) couples names through a single factor:

$$ X_i = \sqrt{\rho}\, M + \sqrt{1 - \rho}\, Z_i, \qquad \text{default}_i \iff X_i < \Phi^{-1}(p_i) $$

Conditional on the market factor \(M\), defaults are independent — making tranche expected losses semi-analytic. Its failures are canon: a single \(\rho\) can't fit all tranches simultaneously (hence base correlation, the smile-fitting patch); tail dependence is understated exactly where seniors live; and \(\rho\) calibrated to CDS spreads told you nothing about house-price correlation. "The formula that killed Wall Street" is unfair to the formula — it was a quoting convention treated as a risk model.

Correlation-desk risk anatomy

  • Leverage of the mezz: a 3–7 tranche has delta ≈ 4–8× to the index spread — spread convexity that flips sign as losses approach attachment.
  • Idiosyncratic vs. systemic: equity tranches are long single-name dispersion (one surprise default is catastrophic); seniors only care about the systemic factor — the two ends literally trade different risks on the same portfolio.
  • The 2005 correlation crisis as the template: GM/Ford downgrades spiked idiosyncratic risk; hedge funds long equity/short mezz lost on both legs as the smile twisted — the reminder that "market-neutral" tranche books are short a hidden cross-gamma.
  • 2012, the London Whale: a $6bn loss in index tranches, post-crisis, at a bank hedging itself — proof the instrument's convexity, not its era, is the hazard.

Where the risk lives now

The post-crisis descendants: CLOs (cash-flow tranching of loans — conservative cousins that sailed through 2020), SRT / capital-relief trades (banks buying mezz protection on their own loan books from funds — synthetic tranching as regulatory tool, the fastest-growing corner), and index tranches as the macro-credit-tail instrument. The lesson institutionalised since 2008 is narrow but real: tranche the cash flows of diverse pools, not the labels of a single correlated bet.

The formulas above are standard textbook formulations, simplified for teaching. They explain the mechanism — they are not a valuation tool, and they will not reproduce a dealer’s price.

5 · Desk notesHow practitioners think about it
Practitioner note: for any tranche, ask what single macro variable would push portfolio losses through the attachment point — then price your tranche as a digital option on that variable. If the answer is "one national housing market" or "one sector's refinancing wall", the copula's ρ is a costume, and you are selling a far-out-of-the-money put on one number.

More Credit Derivatives

CreditCDS

Credit Default Swap

Insurance on a borrower's default — and the market's sharpest real-time gauge of credit fear.
CreditCDX · iTraxx

CDS Index

Default protection on 100+ names in one trade — the S&P 500 of credit risk.
CreditCLN

Credit-Linked Note

A bond with a CDS hidden inside: earn an enhanced coupon for carrying someone else's default risk.