CDO Risk and Valuation: When Models Meet Reality
Book: Structured Finance and Collateralized Debt Obligations | Author: Janet M. Tavakoli | Publisher: John Wiley & Sons (2008) | ISBN: 978-0-470-44344-6
Chapter 5 hits different. This is where Tavakoli stops explaining how CDOs work and starts explaining why they blow up.
The chapter opens with a blunt statement: “Diversification reduces risk” is the least useful finance nursery rhyme for CDO investors.
That’s a bold claim. Let’s break down why she makes it.
The Diversification Myth
Everyone learns in Finance 101 that spreading your money across assets reduces risk. Add a third asset to a two-asset portfolio and the volatility drops. The math checks out. Tavakoli shows the standard calculations. Expected return stays the same, variance falls, value-at-risk improves.
But here’s the catch: the math only works if you actually understand what you’re diversifying into.
Tavakoli’s line on this is sharp: CDOs can help investors diversify into their areas of incompetence.
So if you don’t understand high-yield loans from the telecom sector, and you buy a CDO that holds them, you haven’t reduced risk. You’ve just hidden it. Adelphia, Enron, WorldCom, Global Crossing - the first half of the 2000s was a parade of CDO-held corporate bonds going to zero. Not because diversification failed mathematically, but because investors didn’t know what they owned.
The best investors actually diversify less. They own what they understand deeply. Diversification is a defensive tool, and like any defensive tool, it only works if you know what you’re defending against.
Modern Portfolio Theory: The Bane of CDOs
Modern Portfolio Theory assumes returns are normally distributed. That’s already a problem.
Credit losses are not normally distributed. They’re skewed - there’s a long fat tail of bad outcomes. As one risk manager put it after looking at the lopsided shape of returns: “I’ve been skewed.”
The bell curve shape that MPT relies on doesn’t describe what happens with credit risk. There are expected losses (what you plan for), unexpected losses (the volatility of those plans), and then there’s tail risk - the stuff that falls outside even your stress scenarios. Fraud, for instance.
And here’s the kicker: a five standard deviation credit event should, under the normal distribution model, happen once every 7,000 years. In the actual market, it happens once or twice a decade. Banks with large concentrations in Enron, WorldCom, or Global Crossing might say it happened even more often than that.
Monte Carlo models don’t solve this problem. They’re only as good as the scenarios you imagine when building them. And imagining the worst requires understanding the fundamentals - the actual business, the actual balance sheet, the actual risk.
Mark-to-Market Hazard
Here’s a moral hazard that sounds abstract but has real teeth.
CDO managers often report their own performance, and investors can’t independently verify it. If a manager holds distressed loans that don’t trade publicly, they can mark those loans at whatever price seems convenient.
Tavakoli describes the case of Beacon Hill, a hedge fund where the SEC alleged managers artificially marked up their mortgage-backed securities shortly after renegotiating a management contract that boosted their performance fees. The portfolio prices were rising on paper. Market prices were dropping in reality.
She also describes receiving a cold call from a fund selling a private placement - a $1 billion market value CDO with high-yield mezzanine, distressed debt, bank loans, and equities. The fund manager earned 20 percent above an 8 percent hurdle. Investors had no way to verify the pricing. And yet someone was calling strangers trying to sell it.
Cash Flow Hazard
The mark-to-market problem has a cousin: the cash flow hazard.
When a CDO manager has a claim on the equity cash flows, the interests of the manager and the noteholders can diverge badly. If losses eat through the equity tranche and the excess spread goes to the manager rather than back to investors, the manager now has an incentive to trade into higher-yielding but riskier credits. The more spread they can generate, the more they earn, even as the portfolio deteriorates.
Tavakoli describes a real cash CDO where exactly this happened. The portfolio started investment grade. Aggressive trading for excess spread left it rated junk. The single-A investor threatened to sue. The manager settled.
Arrangers play games too. They include Ford Motor Credit in a “finance risk” bucket while simultaneously filling the “automotive risk” bucket, letting both risks accumulate while gaming diversity scores. They arbitrage ISDA documentation so their side of the trade gets favorable terms and the CDO investor gets the worst language available.
Global Derivatives Risk and Loans
By end of 2006, global notional derivatives outstanding had grown from $41 trillion in 1995 to $386-$416 trillion. Credit default swaps went from insignificant to nearly $29 trillion by the same measure.
The loans picture is equally concentrated. If you had $1 billion in interest rate swaps with Enron, net credit exposure after netting was around $43 million. A $1 billion loan to Enron was a $1 billion exposure. Loans don’t net. And banks don’t mark loans to market the way they mark trading books.
This is the blind spot Tavakoli identifies: enormous credit exposures deemed solid or strategically important get almost no scrutiny, while small but troubled credits get lengthy risk presentations. Exposures to Marconi, Adelphia, Global Crossing, and WorldCom were often in the $500 million to $1 billion range, and loan officers argued these were fast-moving trains worth boarding. Until they crashed.
The Leverage Paradox
Debt has no upside. That’s the core of this section.
When you borrow to buy equity, there’s a chance the equity rises and bails you out if things go wrong. Debt doesn’t work that way. If a debt security drops in value because of fraud or a systemic problem - like the subprime mortgage market - the value destruction is permanent. There’s no upside potential to bail out the leveraged position.
Yet banks kept extending credit lines to hedge funds buying fully priced CDO tranches, turning a blind eye to what would happen in an unwind.
Even CDO equity behaves like debt, not corporate stock. It can’t create new products or generate revenue growth. The upside is capped by the structure.
Fraud: Expect It
Tavakoli devotes several pages to fraud, and her conclusion is simple: expect it, plan for it, and build reserves for it.
The RBG metals fraud is a detailed example. A network of metals companies - RBG Resources, Hampton Lane, Allied Deals, SAI Commodity - duped a group of major international banks including JPMorgan Chase, FleetBoston, and Westdeutsche Landesbank. They built up trading volume over years, generating small wash trades to increase credit lines, then issued fraudulent bills of lading for nonexistent metals. Banks authorized letters of credit. The metal didn’t exist.
The head of RBG dressed well, drove expensive cars, sat in plush offices. The veneer was largely rented. One bank introduced to RBG through a broker did essentially no independent verification.
The lesson Tavakoli draws from Robert Cialdini’s work on influence: we are susceptible to counterfeit social evidence. If everyone else is lending, we assume it’s fine. We lend too. The solution isn’t cynicism - it’s conscious alertness to that bias.
Due diligence isn’t a cure. Bernie Cornfeld once poured a thin layer of oil over tanks full of water. Due diligence investigators dipped their sticks, got oil on them, and signed off on the loans.
But due diligence combined with fraud reserves is a plan.
Tavakoli’s Law and Hedge Funds
Hedge funds became massive investors in CDOs, buying already-leveraged tranches and leveraging them further through total return swaps and synthetic financing. Prime brokers had no recourse to the managers themselves - only to fund assets. Managers earned fees whether the fund performed or failed.
Tavakoli’s Law: If some hedge funds soar above market averages, others must crash and burn.
The math is straightforward. Passive investors get average returns. Active investors who outperform must be offset by active investors who underperform in aggregate. Somewhere in the hedge fund universe, for every massive winner, there are massive losers.
After LTCM’s collapse in 1998 - the fund was run by two Nobel laureates and a former Fed vice chairman, and it still imploded after Russia defaulted - the lessons seemed obvious. But by 2007, hedge fund assets had tripled to around $1.5 trillion. And the risks were arguably worse than at LTCM, because structured products allowed hidden leverage beyond what prime broker surveys could detect.
Amaranth Advisors is the object lesson here. A fund with $9 billion in assets traded leveraged positions in natural gas futures, went long winter contracts while short fall and spring contracts in sequential years. Spreads moved five to ten standard deviations. Amaranth needed to sell illiquid assets into a market that had realized it was distressed. In September 2006, the fund skidded from $9 billion to $3 billion.
Tavakoli called this the “Dead Man’s Curve” trade. The classic phrase: “The last thing I remember, Doc, the market started to swerve…”
Brain Damage Theory
This section is Tavakoli at her most pointed.
A 2005 paper in the Journal of Economic Literature argued that people with frontal brain damage make better investors because they’re more willing to accept reasonable risk. The study showed that brain-damaged subjects would take a 50-50 bet between winning $300 or losing $200, while most people without brain damage would only accept that bet if the upside were $400.
The authors concluded that reduced loss aversion made for superior investment decisions.
Tavakoli’s response: “I am not making this up.”
Her counterargument is that the market doesn’t present us with known probabilities and defined outcomes. What looks like acceptable risk in a controlled experiment is simply a lower margin of safety when you’re operating in the real world with uncertain probabilities and variable outcomes. When markets turn bad and you haven’t managed the risk, that “superior decision” gets relabeled an inferior one.
Amaranth’s Brian Hunter looked like a genius taking controlled risks - until he wasn’t.
The Alternative to Model Reliance
All financial models require assumptions. When the models aren’t flawed, the assumptions often are. When the assumptions aren’t flawed, the application might be.
Correct mathematical equations are frequently misapplied. You can be astonishingly precise while being frighteningly inaccurate.
The alternative Tavakoli advocates isn’t abandoning models. It’s scenario analysis - building hedges and stress tests around the scenarios you genuinely can’t afford to be wrong about, rather than trusting a model’s ability to predict outcomes it was never designed to predict.
Understanding the fundamentals - the balance sheet, the actual cash flows, the actual collateral - combined with scenario stress testing and fraud reserves, is the approach she advocates throughout the book.
The chapter ends with a preview of what’s coming: a key determinant of success is understanding the default probability of individual portfolio assets and the structural and documentation features of the financing.
That’s where the next chapter begins.