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.
1 · SnapshotThe one idea to remember
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.
Point at a line to read what it is doing.
How do I read this chart?
Losses on the underlying pool run across, losses taken by each slice up. Read it as three step functions rather than three lines: a tranche is untouched until the pool's losses reach its attachment point and then takes everything until its detachment point. The rating on a slice is a statement about where those two numbers sit, not about the borrowers.
Where this is explained properly:
The axes carry no scale on purpose. Every line here is a stylised shape drawn to show a mechanism, so there is no number to read off one — and any figure taken from it would be invented.
- 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
Which risks decide the outcome
Not how risky this is, and not a rating — there is deliberately no total. It says which of five failure modes drives what happens here, in the same order on all 129 products so they can be compared. This publication's own reading; see the notice below.
- Marketmatters
- Creditdecides it
- Liquiditydecides it
- Fundingmatters
- Operationalbarely applies
What decides it here. Correlation decides the tranches, and correlation is an assumption rather than an observation. 2008 is what happens when the assumption and the world disagree.
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:
What the symbols mean
- Lleverage, or a loss given default
- Dduration: how far a bond's cash flows sit in the future
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.
4 · AdvancedPricing & valuation
The Gaussian copula and its scar tissue
The market standard (Li, 2000) couples names through a single factor:
What the symbols mean
- rhocorrelation between two things
- Phithe normal distribution's cumulative function
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
Now say it back
Close the page and give CDO & Synthetic Tranches in four sentences. It takes a minute and it is the only way to find out whether reading it was enough.
- Who wants what — two parties wanted opposite things badly enough to write it down.
- What the contract obliges, and when — not the payoff; the obligation.
- Where the money comes from — name the source, or you have described a hope.
- What makes it lose — the ordinary way, not the dramatic one.
Put CDO & Synthetic Tranches beside any other instrument →
Where this instrument shows up elsewhere
- MediumSubprime & the CDO Machine, 2008Case StudiesHow a national bet on house prices was relabelled AAA thousands of times — and what happened when the one assumption…
- MediumWhich Desk Trades WhatPrepEleven trading seats and six that sit next to them: what each one actually touches, the single number it lives by,…
- HardSecuritisationConceptsTurning streams of loan payments into tradable bonds — the machine behind MBS, ABS, CLOs and CDOs, its 2008 failure,…