Synthetic CDOs and the Madness of Correlation
Book: Structured Finance and Collateralized Debt Obligations | Author: Janet M. Tavakoli | Publisher: John Wiley & Sons (2008) | ISBN: 978-0-470-44344-6
Chapter 14 is where Tavakoli explains how synthetic CDO trading books became what she calls invisible hedge funds inside investment banks. The mechanics are fascinating and deeply uncomfortable.
The Synthetic Cash Windfall
After the defaults of WorldCom, Enron, Kmart, and Adelphia, banks holding super senior tranches should have been showing losses on their books. The AAA tranches of affected deals were downgraded to AA. Logic says the super senior marks should have moved too.
They didn’t.
Bank traders worked out mark-to-market methodology with risk managers who were often junior to and financially dependent on those same traders. Since super senior tranches almost never traded in the market, there was no independent price to challenge whatever the trading desk said. Some creative desks traded tiny slices of super senior positions at pre-agreed prices with counterparties, then used those prices as “market evidence” to validate their existing marks.
In plain language: they made up the prices and then manufactured fake evidence to support them.
Synthetic Equity and the 60 Percent IRR
When credit spreads were wide in the early 2000s, the first-loss (equity) tranche of an investment-grade CDO could generate returns of over 60 percent IRR under Moody’s base-case loss assumptions.
But here is the thing. Much of that 60 percent wasn’t real equity return. It was the 44 bps stripped off the super senior tranche and rerouted into equity cash flows. The bank arranger robbed the super senior to pay the equity. The equity looked amazing. The super senior got paid almost nothing for the risk it held.
By the mid-2000s credit spreads had tightened significantly. The same equity math now produced single-digit returns. Tavakoli asks: how can arrangers justify this?
They couldn’t. But they kept doing it. Senior bank managers had no idea.
Her advice is blunt: if you are a buyer of equity risk, never accept fixed payments from an arranger. Always arrange your own deal. Retain the excess spread on the full notional. She recommends targeting at least 25 percent base-case IRR and walking away if you can’t get there.
Bespoke Tranches: Single-Tranche CDOs
Single-tranche CDOs (STCDOs) emerged around 2000. The premise sounds reasonable: sell just one tranche of a CDO to a specific investor, then hedge the remaining risk using credit default swaps.
The reality was a mess.
When an arranger sold only the mezzanine tranche of a $5 billion CDO, the remaining $4.75 billion of risk stayed on the trading book. The hedge? Sell credit default protection on about $1 billion of the underlying names. This is the delta hedge.
Tavakoli explains why this hedge is fundamentally broken. The sold mezzanine tranche is a horizontal slice of the deal’s risk. The delta hedge is a vertical slice. They do not behave the same way. When credit spreads on specific names gap out, the mezzanine price for a rated tranche can stay flat while the CDS positions move against you. The losses are real. The hedge provided nothing.
If everything went fine, the arranger made slightly more than a fully hedged deal. If things went wrong, the arranger had taken on the equivalent of large, unhedged equity risk without knowing it.
One head of structuring at a Canadian bank told her: “If my boss knew what I was really doing, he’d fire me.”
Banks’ Invisible Hedge Funds
This section is one of the most important in the book.
Traders running STCDO books were effectively operating leveraged credit funds inside investment banks, invisible to senior management. They had:
- Large one-sided long positions in credit risk
- Hedge ratios they controlled and could manipulate
- The ability to manufacture synthetic income by adjusting those ratios
- A call option on upside (higher bonuses if bets paid off) with limited personal downside (someone else takes the loss)
The misalignment of incentives was structural. Risk managers were paid less than traders and often depended on traders for future employment references. Senior bank managers didn’t understand the products well enough to challenge the numbers.
The Madness of Correlation
Correlation trading became a mass financial delusion, in Tavakoli’s words.
Here is the problem. When calculating credit losses, you have three variables: default probability, recovery rates, and correlation. Of these three, correlation is the least important. Yet the industry built entire careers around modeling it.
Correlation traders borrowed the language of conventional options markets and called it “delta hedging” and “spread convexity.” Tavakoli calls this a false analogy. In the Treasury market, you can calculate the exact price of a bond for any interest rate. The math is reliable.
In credit markets, you are guessing at pairwise default correlations using historical data that may not apply to current conditions. You are estimating recovery rates from averages that ignore how bad conditions can get. You are using models so unstable that small changes to inputs produce wildly different hedge ratios.
One investment bank showed her their delta calculations. The deltas across all tranches added up to 118 percent of the notional. They had to fudge the numbers to make them sum to 100. When a model needs to be faked to pass basic arithmetic checks, it is not a model you should be using.
Her prescription: “A major improvement in the financial markets would be to spend zero time, money, and resources on correlation and spend all of one’s time, money, and resources on better estimates of default probabilities and recovery rates.”
Synthetic Notional vs. Actual Risk
Dealers wanted to report market size as the notional of tranches they actually sold. If they sold a $250 million mezzanine tranche of a $5 billion deal, they reported $250 million.
Tavakoli argues this is misleading. The arranger’s true risk position is equivalent to the entire $4.75 billion they didn’t sell. Reporting only the sold tranche tells you nothing about the risk concentration.
By mid-2005, bespoke tranches were reported at $169 billion. If those tranches represented 5 to 10 percent of the underlying CDO notionals on average, the full underlying exposure was somewhere between $1.7 and $3.4 trillion.
Found Money and Moral Hazard
The found money problem is almost embarrassingly simple once you see it.
A trader who wants a bigger bonus can manipulate the hedge ratio. Sell slightly more credit default protection than the model says, report the same risk, pocket the additional income as revenue. Risk managers who cannot independently verify the hedge ratios have no way to challenge this.
This isn’t even illegal. It is just the natural outcome of a system where:
- Senior managers don’t understand the products
- Risk managers are financially subordinate to traders
- Hedge ratios cannot be independently verified
- Large bonuses depend on reported revenue
Tavakoli describes the correlation trade as having all the elements needed to manufacture income while deferring risk into the future. Fast cash flows now, potential losses later, and a likely move to a different firm before the bill comes due.
Previous: Super Senior Sophistry: The AAA Disappearing Act
Next: A Tour of Structured Finance Products: The Good, Bad, and Weird