Collateralized Debt Obligations

Credit Default Swaps let you trade the default risk of a single reference entity. A collateralized debt obligation (CDO) does something structurally different: it pools hundreds of credit exposures — corporate bonds, bank loans, or (most famously) residential mortgages — into a single portfolio, then slices the pool's losses, not its cash flows directly, into layers of seniority. The remarkable (and, as later modules will show, sometimes dangerous) result is that a pool built entirely from risky assets can produce a slice that a rating agency is happy to call AAA.

The waterfall

Picture the whole pool's notional value as a single vertical bar, from 0\% at the bottom to 100\% at the top. The bar is cut into horizontal tranches, each defined by an attachment point (where its loss exposure begins) and a detachment point (where it ends):

Tranche	Attachment	Detachment	Absorbs losses…
Equity	0%	3%	first — the "first-loss piece"
Mezzanine	3%	7%	only once equity is wiped out
Senior	7%	15%	only once mezzanine is wiped out
Super-senior	15%	100%	only in a catastrophic scenario

As the underlying pool suffers losses (from defaults among the bonds or loans inside it), those losses flow through the structure like water down a waterfall: they fill the equity tranche first, and only once equity's entire notional is exhausted do losses begin spilling into mezzanine, and so on upward. For a given tranche with attachment a and detachment d, the fraction of that tranche's own notional lost, as a function of the pool's aggregate loss percentage \ell, is:

\text{tranche loss \%} = \frac{\max\big(0,\ \min(\ell, d) - a\big)}{d - a}.

Drag the equity tranche's size below and watch how thickening it (pushing mezzanine, senior, and super-senior's attachment points upward) protects everything above it against a given pool loss — this single number is the whole engine behind "manufacturing" a safe senior tranche.

Worked example

A CDO pools loans totaling \$500 million, structured with the attachment points above: equity 0$–$3\% (\$15M), mezzanine 3$–$7\% (\$20M), senior 7$–$15\% (\$40M), super-senior 15$–$100\% (\$425M).

Suppose the underlying loan pool loses 5\% of its value — \$25 million in defaults. Follow the waterfall:

Equity absorbs the first \$15M (its entire notional) — wiped out, a 100% loss for equity holders.
The remaining \$25\text{M} - \$15\text{M} = \$10M spills into mezzanine (\$20M tranche): 10/20 = 50\% loss.
Senior and super-senior are untouched — the pool's 5\% loss never reaches senior's 7\% attachment point, so both take a 0\% loss.

Even though every dollar inside this pool came from the same shaky loan book, the senior and super-senior investors emerged completely unscathed from a loss that wiped out equity and halved mezzanine. That's the tranching mechanism working exactly as designed.

Two ingredients combine to make it possible, at least on paper. First, subordination: the senior tranche only loses money after every dollar of equity and mezzanine beneath it has already been wiped out, so it needs a genuinely severe, correlated wave of defaults to touch it at all — not just a handful of unlucky borrowers. Second, diversification: pooling hundreds of loans is supposed to average away the idiosyncratic risk of any one borrower, the same statistical logic an insurer relies on. Put together, a senior tranche's expected loss can look, on paper, dramatically safer than any single loan in the pool — safe enough, historically, for rating agencies to stamp it AAA even when every underlying loan was BBB or worse.

That word "correlated" in the first sentence is doing enormous work, and it's exactly the thread the 2008 financial crisis module pulls on: this very tranching mechanism, applied at massive scale to pools of subprime mortgages, is central to how the crisis unfolded.

The diversification argument for a safe senior tranche silently assumes that defaults across the pool are roughly independent (or only mildly correlated) — the way flu cases in unrelated households might be. Mortgage defaults are not like that: they share a common driver, the housing market and the broader economy, so when conditions turn bad, they tend to turn bad for many borrowers simultaneously. Under high correlation, the "diversification" that was supposed to protect the senior tranche evaporates — losses stop looking like a smooth, averaged-out curve and start looking like an all-or-nothing cliff. Underestimating default correlation is widely regarded as the single most consequential modeling error behind the mispricing of mortgage-backed CDOs before 2008 — an AAA label is only as good as the correlation assumption baked into the model that produced it.