Tranching & the Cash-Flow Waterfall

Take a pile of a thousand mediocre car loans, or middling corporate bonds, or wobbly mortgages — collateral a rating agency would politely stamp BBB and a pension fund is legally barred from touching. Now perform a trick: out of that exact same pile, manufacture a bond so safe it earns a AAA rating, sell it to the pension fund, and sell the leftover explosive bits to a hedge fund that wants the fireworks. No new loans were added. Nobody’s credit improved. Yet you’ve conjured a triple-A bond out of triple-B raw material. This is tranching, and it’s the closest thing structured finance has to alchemy. The machine that does it is the cash-flow waterfall, and the fuel is a stack of protections called credit enhancement. By the end of this lesson you’ll know exactly how the trick works — and exactly where the trapdoor is hidden.

Slicing the pool: attachment & detachment

The analogy. Picture the pool’s losses as floodwater rising in a building. The basement floods first, then the ground floor, then the upper floors. A tranche is a floor of that building. The equity tranche lives in the basement — it drowns first. The senior tranche has the penthouse — water only reaches it after every floor below is already submerged. Where your floor starts flooding and where it’s fully underwater are the two numbers that define everything about your tranche.

The precise definitions. A tranche is a slice of the pool’s losses (and the cash flows that mirror them), defined by two points on the 0–100% loss axis:

• The attachment point is where the tranche starts taking losses. Below it, your tranche is untouched.
• The detachment point is where the tranche is completely wiped out. Above it, there’s nothing left of you to lose.
• The thickness of a tranche is detach − attach — how much loss it can absorb across its own range.
• The subordination of a tranche is the total percentage of losses sitting below it that must be fully burned through before it loses a cent. Subordination equals the attachment point. It is the cushion of junior capital protecting you, and it is precisely the credit enhancement that earns a senior tranche its rating.

A standard three-tranche stack looks like this:

The loss formula. Given a total pool loss (defaults as a % of the pool), the loss hitting any one tranche is a clamp:

$\text{tranche loss} = \mathrm{clamp}(\text{poolLoss} - \text{attach},\; 0,\; \text{thickness})$

In words: subtract your subordination (the attachment point) — that’s the part of the loss the floors below you already soaked up. If the result is negative, you’re dry (loss = 0). If it exceeds your thickness, you’re fully drowned (loss = thickness). Anything in between is the chunk you personally eat. To express that as the percentage of your own tranche wiped, divide by thickness:

$\%\text{ of tranche wiped} = \frac{\mathrm{clamp}(\text{poolLoss} - \text{attach},\, 0,\, \text{thickness})}{\text{detach} - \text{attach}}$

The interactive — drag the flood up. The picture below is this whole section in one control. The default stack is Equity 0–5%, Mezzanine 5–15%, Senior 15–100%. Grab the Portfolio loss slider and pull it up from 0%: watch the losses fill from the bottom. The equity tranche turns to ash first; only once it’s 100% gone does the mezzanine start bleeding; and the senior tranche sits there serenely until losses pierce 15%. The per-tranche readouts show each slice’s thickness and exactly what fraction of it has been wiped, and the structure highlights whichever tranche is currently absorbing the marginal loss. Push the slider to 4%, then 8%, then 20%, and narrate to yourself which floor is underwater.

Worked example — pool loss of 8%. Let’s run the formula for every tranche when defaults take out 8% of the pool.

So an 8% pool loss completely vaporizes the equity tranche, chews through 30% of the mezzanine, and leaves the senior tranche perfectly intact. The senior holder didn’t lose a cent on an 8% default wave — because 15 points of subordination stood between them and the water. That is credit enhancement doing its job.

When to use this lens

Any time you see a structured product — a CLO, an ABS, a CDO, a CMBS — reach for attachment/detachment first. Those two numbers tell you, in one glance, how much pain a tranche can shrug off (its subordination) and how much it can take before it’s gone (its thickness). Everything else is detail.

The cash-flow waterfall

The analogy: a champagne tower. Stack champagne glasses in a pyramid and pour from the top. The top glass fills first — that’s the senior tranche getting its coupon before anyone else sees a drop. Once it’s full, the overflow cascades to the next level (mezzanine), and only the final dregs reach the bottom glass (equity). But flip the tower upside down for losses: spill something nasty in and the bottom glass overflows first. One tower, two directions: cash pours down from the top, losses rise up from the bottom. Memorize that and you understand the waterfall.

The precise definition. The cash-flow waterfall (formally, the priority of payments) is the contractually fixed order in which collateral cash is distributed each period. It almost always runs in two stages:

1. The interest waterfall distributes the interest the pool collected: senior fees and the trustee first, then the senior tranche’s coupon, then the mezzanine coupon, and finally any leftover interest as residual to the equity tranche.
2. The principal waterfall distributes principal repayments and recoveries — paying down the tranches’ outstanding balances, again in priority order (with some twists below).

Sequential vs pro-rata principal. How principal gets allocated across tranches is the key design choice:

• Sequential pay: all principal goes to the senior tranche first until it’s fully retired, then to the next tranche, and so on. This steadily builds subordination for the juniors over time — as the senior balance shrinks, the same dollar cushion underneath it becomes a larger percentage of the remaining structure, so surviving juniors get safer the longer the deal runs.
• Pro-rata pay: principal is split proportionally across tranches, so everyone amortizes together and the relative thicknesses stay roughly constant. Deals often start pro-rata in good times, then switch to sequential if a performance trigger is breached (e.g. delinquencies or losses cross a threshold), slamming the cash back toward the senior to protect it when the pool sours.

Worked intuition. Say the senior is $850M of a $1,000M deal, with $150M of juniors below it (15% subordination). Pay $200M of principal sequentially to the senior: it shrinks to $650M, the deal is now $800M, and the $150M of juniors is suddenly 18.75% subordination (150 / 800). The seniors that remain got safer just by being paid first. Under pro-rata, that $200M would have been split across all tranches and subordination would barely have budged.

Credit enhancement

Subordination is the headline form of credit enhancement, but real deals stack four protections. Think of them as concentric moats around the senior tranche.

1. Subordination. Already covered: junior tranches absorb losses first. The senior’s attachment point is its subordination. This is structural enhancement — it comes from how the liabilities are sliced, costs nothing extra, and is the largest moat in most deals.

2. Overcollateralization (OC). The deal holds more collateral par than it issued in liabilities. If you fund $480M of notes with $500M of loans, the extra $20M is overcollateralization — a 4% cushion of “spare” assets whose cash flows belong to the structure but back no bond. Losses eat the spare collateral before they impair the notes.

3. Excess spread. The collateral yields more than the structure pays out in coupons plus fees plus realized losses. That positive carry — the excess spread — is the first line of defence: every period, small losses get quietly absorbed by the surplus before they ever touch a tranche. Unlike a fixed reserve, it’s a flow, refreshed each period. In good months the surplus may be trapped and used to pay down senior principal (de-leveraging the deal) or build the reserve.

4. Reserve / cash-collateral account. A pot of cash set aside at closing (or built up from trapped excess spread) that the trustee can tap to cover shortfalls. It’s the literal rainy-day fund — funded capital sitting in an account, ready to plug a missed payment.

Worked example — sizing the cushions. A deal holds $500M of collateral against $480M of issued notes. The overcollateralization is $500 - 480 = 20$ ($20M), i.e. $20 / 500 = 4\%$ of the pool. On top of that, the collateral yields 2% more per year than the deal pays in coupons and fees — that’s 2% of annual excess spread, roughly $10M a year of surplus that eats small losses before the 4% OC cushion is even touched, and certainly before any tranche is impaired.

Worked example: who loses what

Let’s put it all together with one master table. Same stack — Equity 0–5%, Mezzanine 5–15%, Senior 15–100% — across four pool-loss scenarios. To turn percentages into dollars, assume a $1,000M pool: equity is $50M (0–5%), mezzanine is $100M (5–15%), senior is $850M (15–100%). For each tranche we compute the loss as $\mathrm{clamp}(\text{poolLoss} - \text{attach},\,0,\,\text{thickness})$ in pool-percent, then convert to ”% of that tranche wiped” and to dollars.

Read it left to right and the personality of each tranche jumps out. At a 2% loss only the equity bleeds (40% of it). By 6% the equity is entirely gone and the mezzanine has just started (10%). At 12% the mezzanine is 70% destroyed but the senior is still pristine. Only at 20% — after losses have torched all of equity and all of mezzanine, a brutal 15 points of subordination — does the senior finally take its first, modest 5.9% hit. The senior’s job is to be boring, and across three of four scenarios it succeeds perfectly.

How BBB becomes AAA (and the catch)

Now the payoff. Why can a pool of BBB loans support a large AAA tranche? Because the AAA tranche only loses money if cumulative defaults exceed all the subordination beneath it. If 15% of junior capital must be vaporized before the senior takes its first dollar of loss, then the senior is, in effect, betting that fewer than 15% of the pool defaults (after recoveries). The individual loans might each have a meaningful chance of going bad — that’s why they’re BBB — but the senior claim on the pool fails only in the far worse scenario where an unusually large fraction of them default together. Diversification is doing the heavy lifting: spread the bet across a thousand loans and, if their defaults are roughly independent, the chance that more than 15% blow up at once is genuinely tiny. Tiny default probability earns a tiny spread and a triple-A stamp. That’s the alchemy: pooling plus tranching converts average credit quality into a thin sliver of near-certain safety on top of a thick layer of risk you’ve shunted down to the equity and mezzanine.

The catch — correlation. That entire argument rests on a single load-bearing assumption: the defaults are not too correlated. Diversification only protects the senior if the loans fail independently — a default here, a default there, statistically unrelated. But suppose they’re all mortgages in the same overheated housing market, or all corporates exposed to the same recession. Then a single macro shock makes them default together, in a clump. Correlated defaults destroy the diversification: instead of a steady trickle that the equity and mezzanine comfortably absorb, you get either very few losses (boom) or a catastrophic wave that smashes straight through 15% of subordination and into the supposedly-safe senior (bust). Crucially, higher correlation moves risk UP the stack — it makes the equity tranche safer (more chance of zero losses) and the senior tranche more dangerous (more chance of the wipeout scenario). A AAA tranche priced for low correlation is radically mispriced if correlation turns out to be high.

Putting it together

Tranching slices a single loan pool into stacked claims defined by an attachment point (where a tranche starts losing) and a detachment point (where it’s wiped out); its thickness is detach − attach and its subordination equals its attachment — the cushion of junior capital that must burn before it loses a cent. Losses flow bottom-up — equity first, then mezzanine, then senior — via clamp(poolLoss − attach, 0, thickness), while cash flows top-down through the waterfall (priority of payments): senior coupons first, equity gets the residual, with principal allocated sequentially (senior first, building junior subordination over time) or pro-rata (until a performance trigger flips it sequential). Four moats of credit enhancement protect the structure — excess spread (the running surplus, first line of defence), overcollateralization (collateral par > liabilities par), subordination (junior absorbs first), and a reserve account (funded cash backstop). Together they let a pool of BBB collateral support a large AAA senior tranche, which loses only if defaults pierce all its subordination — a near-impossibility if defaults are roughly independent, and a near-certainty if they aren’t. That last clause — correlation — is the trapdoor, and the subject of the next lesson.

Next — default correlation: we’ll formalize the one variable that decides whether a senior tranche is a fortress or a time bomb, meet the copula models that tried to pin it down, and trace exactly how a wrong correlation number detonated the AAA-rated heart of the 2008 crisis.