How to analyse a Structured-Credit Waterfall, Step by Step

A waterfall is two queues running in opposite directions: collections fill the pieces of a structure from the most senior downwards, and losses climb from the most junior upwards. Read one by listing the pieces and their sizes, computing what stands beneath each piece with the pool named as the base, then walking a stated loss up from the bottom until it stops. The order is knowable. The odds are not.

A structure is not a set of shares in a pool. If it were, every piece would give up the same proportion of itself the moment the pool gave up anything, and the pieces would be one instrument sold in three sizes. A queue separates them. There is one queue for money arriving and a second queue, pointing the other way, for losses arriving. Everything a reader can honestly say about a structured pool comes out of those two queues and the sizes of the pieces standing in them. Both queues have to be read, and the sizes read with them.

What is actually flowing, and in which direction?

Two things move through a structured pool and they move against each other. Money collected from the underlying borrowers enters at the top and fills the pieces from the most senior downwards: the senior piece is served, then the mezzanine piece, then the equity piece. Losses enter at the bottom and climb: the equity piece is reached, then the mezzanine piece, then the senior piece. The two directions come before any number. Almost every misreading of a structured pool takes one direction and answers the other direction's question with it.

The everyday version is a queue at a ticket counter on a festival morning. The person at the front is served first, and the person at the back is served last. Now the counter runs out of tickets. Who is turned away first? The person at the back. One queue, two ends, and the position that is best for being served is the position that is worst for missing out. A structured pool is that queue written down and paid for, with the senior piece at the front for money and at the back for losses.

The two directions are not an accident of how somebody drew a diagram. The two directions are the arrangement itself. A holder of the senior piece is paid before anybody else out of collectionsThe money that actually reaches the pool from the underlying borrowers in a period, as distinct from what those borrowers owe., and in exchange that same holder accepts a lower return than the equity piece is promised. A holder of the equity piece is paid last and reached first, and accepts that in exchange for the largest promised return. The yield a piece should carry is settled separately, under the pricing of credit risk. The order itself is the part a reader can check.

Kept separate from the beginning, the two arrows prevent half the errors on this subject. A reader who has been told the senior piece is paid first, and who then reasons that it must therefore lose least, has reached a true conclusion by an argument that does not hold. The conclusion happens to be true because the arrangement was written so, and not because being first in one queue implies anything about the other. On a differently written structure it would be false, and the reader who reasoned that way would have no idea.

Does step one really begin with an addition?

It does, and it is the step most readers skip. Step one is to write down every piece of the structured pool, its size in rupees and its share of the pool, then add the sizes and set the total against the pool. The addition is the whole of step one, and it looks like housekeeping until a structure fails the check. A set of pieces that does not account for the pool has either a piece nobody mentioned or a figure somebody mistyped, and every step after this one inherits the gap without showing it.

On the invented pool used throughout this guide, the senior piece is Rs 960 crore, the mezzanine piece is Rs 180 crore and the equity piece is Rs 60 crore. Add them: Rs 960 crore and Rs 180 crore and Rs 60 crore make Rs 1,200 crore, the pool exactly. In whole rupees that is Rs 12,00,00,00,000/-, and the same addition closes there too, with nothing left over in either direction. The shares are 80.0 per cent, 15.0 per cent and 5.0 per cent of the pool, and those three close on 100.0 per cent with no residual. A set of shares very often does not.

Why does a missing piece matter so much? Because the pieces below a given piece are the only thing standing between the pool's losses and its holder, and a piece nobody mentioned is either standing there unrecorded or is not standing there at all. Suppose the list handed over showed the senior piece and the equity piece and no mezzanine piece, against the same Rs 1,200 crore pool. The addition gives Rs 1,020 crore and falls Rs 180 crore short. The gap is known before its cause is known, and that is the correct order to find out in. The addition takes ten seconds on paper, and it is the only step in this guide that can invalidate all the others.

What stands beneath a piece, and what is the base?

Beneath a piece stands the total size of every piece the losses reach before they reach it. The total beneath a piece is the amount the structured pool can give up before that piece feels anything at all, and computing it is step two. The senior piece has two pieces under it, the mezzanine piece and the equity piece, so its cushion is Rs 180 crore and Rs 60 crore added, or Rs 240 crore. The mezzanine piece has one piece under it and no more, so its cushion is the Rs 60 crore of the equity piece and stops there. The equity piece has nothing under it at all, so the very first rupee the pool gives up reaches it.

Now name the base out loud. A share of the pool and a share of a piece are two different numbers, and only one of them is what stands beneath anything. Set Rs 240 crore against the pool and it reads 20.0 per cent. Set the mezzanine piece cushion of Rs 60 crore against that same pool and it reads 5.0 per cent. The base is the pool in both readings, and it is the pool for a reason a reader can hold on to: the loss arrives at the pool. The loss arrives as a shortfall in what the underlying borrowers pay, it is measured against the pool that lent to them, and it is only afterwards handed to the pieces. Measuring the cushion against anything other than the thing the loss lands on gives an answer to a question nobody asked. Setting the same rupee amount against a piece instead answers a different question, and that question is worked out separately.

Two things follow immediately and both are worth pausing on. First, the figure is fixed on the day the structure is made. Sizes do not move by themselves. Second, the cushion is the only quantity in the whole procedure that describes a holder's position rather than the pool's. A reader who computes one number about a structure usually computes this one. The cushion deserves that status. The sentence readers attach to it does not deserve the same trust, and that sentence is the subject of everything below.

Why are the two orders written as two separate lists?

The order in which cash is paid runs from the senior piece downwards. The order in which losses are taken runs from the equity piece upwards. The two orders are not one list read in two directions: they are two lists, written deliberately so that whichever piece is paid first is hit last. Knowing that a piece is paid first says nothing about how much loss stands beneath it, and knowing how much stands beneath it says nothing about when that piece is paid. Two questions, two lists, and they are answered in different units: one in rupees, one in dates.

The two lists are where a careful reader loses time, so it is worth being blunt about which question each of them answers. Whether the money owed to one holder arrives before somebody else's is the payment order, and it is a question about sequence in time. How much the pool has to give up before a rupee of that holder's own is gone is the loss order, and it is a question about size in rupees. The two lists happen to be mirror images in this structured pool. The two lists are mirror images in most structures a reader will meet, and mirroring is the arrangement people pay for. But the mirroring is a fact about the document, not a law, and the only way to know it holds is to read both lists rather than reading one and assuming.

An everyday version follows, and it is not a perfect one, so it is worth stating carefully. A household paying its bills at the end of the month has an order it pays in: the school fee, then the electricity, then the shopkeeper it has known for years. If the salary is short, the order it stops paying in is not simply that list reversed. The household might stop paying the shopkeeper first, and stopping the shopkeeper happens to reverse the list, or it might stop the electricity instead. Which one it stops depends on who will wait. The household has two orders and they are settled separately. A structured pool has the same two orders, and the only difference is that both of them are written down in advance instead of being decided in the moment.

Where does a stated loss stop as it climbs?

Step four is a walk and it is the step that turns a list of sizes into a reading. Take a loss stated as a share OF THE POOL, turn it into rupees, and hand it to the pieces from the bottom. Each piece absorbs up to its own size and passes whatever is left upwards. Nothing else happens, and nothing else is needed: no assumption, no view, no forecast. The walk is arithmetic on figures already in hand, and arithmetic on figures already in hand can be trusted in a way that nothing else on this subject can.

Run at 10.0 per cent of the pool, ten per cent of Rs 1,200 crore is Rs 120 crore, or Rs 1,20,00,00,000/- in whole rupees. Hand it to the equity piece: the equity piece is Rs 60 crore, so it absorbs Rs 60 crore and is exhausted, and Rs 60 crore travels upwards. Hand that to the mezzanine piece: the mezzanine piece is Rs 180 crore, comfortably larger than what has arrived, so it absorbs the whole Rs 60 crore and stops the walk there. The senior piece absorbs nothing. Add the three amounts back: Rs 60 crore and Rs 60 crore and nothing is Rs 120 crore, the loss the walk started with. The closing check is not decoration. The check is what makes a waterfall a queue rather than an opinion, and it holds at every loss that can be stated.

Notice what the walk at 10.0 per cent produced: two amounts of Rs 60 crore, one absorbed by the equity piece and one absorbed by the mezzanine piece. A reader who spots two identical figures and is told nothing about them will usually assume one is a misprint, so it is worth naming. The repetition is forced arithmetic and not a coincidence. Rs 120 crore of loss, less the Rs 60 crore that the equity piece could take, leaves exactly Rs 60 crore, and the equity piece happens to be that size because it was made that size in step one. The Rs 60 crore that reaches the mezzanine piece is one third of the mezzanine piece, so two thirds of that piece is still standing when the walk ends.

What happens when the walk is taken far enough to reach the top?

Pushed further, the walk reaches the heart of the procedure. At a loss of 20.0 per cent of the structured pool, or Rs 240 crore, the equity piece absorbs Rs 60 crore and is gone, the mezzanine piece absorbs Rs 180 crore and is gone, and the senior piece absorbs nothing at all. Not one rupee. The figure should look familiar, and it is worth saying why rather than letting a reader wonder: Rs 240 crore IS what step two computed as standing beneath the senior piece. The largest loss at which a piece absorbs nothing is the same number as what stands beneath it, so steps two and four are one fact seen from two sides.

Now go one further. Everything above stops at exactly the moment a reader most needs it to continue. At a loss of 28.0 per cent of the pool, or Rs 336 crore, the equity piece and the mezzanine piece are both exhausted and Rs 96 crore reaches the senior piece. Rs 96 crore is 10.0 per cent of the Rs 960 crore senior piece, so a tenth of the senior piece has gone. The three absorbed amounts, Rs 60 crore and Rs 180 crore and Rs 96 crore, add back to Rs 336 crore, and the check closes here exactly as it closed at 10.0 per cent. The senior piece is not safe. It is late. Those are different properties, and only one of them is written into the structure.

The shape of the relationship is worth seeing on its own. A reader's intuition gets it wrong before any arithmetic is done. The amount a piece absorbs is not proportional to what the pool loses. It is flat, then steep, then flat again. The equity piece absorbs at the full rate of the pool's loss from the very first rupee, and stops dead once it has given up its own Rs 60 crore. The mezzanine piece absorbs nothing while the equity piece is being used up, then absorbs at the full rate until it too is gone, then nothing. The senior piece is motionless across the whole first fifth of the pool's loss and then begins. Draw it and the shape is unmistakable. Describe it in a sentence and half the readers will still be picturing three lines rising together.

What do the four declared walks look like set side by side?

The whole worked instance sits in one place here. The structured pool holds Rs 1,200 crore of receivables, funded by a senior piece of Rs 960 crore, a mezzanine piece of Rs 180 crore and an equity piece of Rs 60 crore. Four losses are walked, each stated as a share OF THE POOL and each set out as a position to look at. Not one of the four is expected, likely, typical or stressed. How often a loss of a given size arrives at a pool is a matter of counted frequency, and no count stands behind these four. They are four places to stand and look at the same three pieces.

Down the equity piece column the figures stop moving after the first row. Nothing is left of that piece to take. Down the mezzanine piece column they stop after the third. Down the senior piece column the entries are empty until the last row, where Rs 96 crore appears. The senior piece column is the whole argument in one strip of figures: three quarters of it is nil, and then it is not, and nothing anywhere in the arithmetic said when the last row would arrive rather than the first.

What can divert cash away from a piece?

Everything so far treats the structure as a set of sizes, and a set of sizes does not move. Step five is where the structure starts moving. Most structured pools carry tests written into them that change where money goes when something in the pool deteriorates: cash that would otherwise have travelled down to a junior piece is redirected instead to repay a senior one, and it keeps being redirected for as long as the test is failed. The amount standing beneath a piece is fixed on the day the structure is made. A test that diverts cash moves something else: where the money goes while the structure is running.

Reading a structure therefore means finding every such test and asking three things about each one, in this order. What does it measure? What level makes it trip? And what happens the moment it does? The first question is usually answered somewhere in the document. The second is a number and is the one most often quoted without the first. The third question matters most and is skipped most often. The answer sits in a different part of the document from the other two. A test whose trip level is known and whose consequence is not is worth almost nothing.

The everyday version of this is a joint household account with a rule attached. Two earners agree that the school fee is paid first every month, and they also agree that if either salary is late, nothing at all goes to the discretionary jar until the fee has been paid twice over. Somebody reading only the standing instruction would say the jar gets whatever is left. Somebody who has also read the second rule knows that in exactly the months when money is short, the jar gets nothing rather than less. The second rule is a diversion test, and it matters because it does its work precisely when the ordinary reading is least reliable.

Notice what this does to a reading built only on the sizes. A holder of the equity piece who has computed that nothing stands beneath them has read their loss position correctly and has read nothing at all about their cash position. A test can switch that cash off entirely while the equity piece is still fully intact. A holder of the senior piece gains from the same test, and gains most in the conditions where the sizes alone would have said nothing had changed yet. Neither of those facts is visible in Rs 960 crore, Rs 180 crore and Rs 60 crore. Both of them are visible in the document that also contains the tests, and that is why finding the tests is a step of the procedure rather than a footnote to it. Where the pool is being administered by a servicerThe party that collects each instalment from the underlying borrowers and passes the money on to whoever is due it, without owning the loans., the measurement that trips a test is usually taken from that party's own reporting, and that is another reason to know what is being measured.

What can this reading not tell at all?

Step six is to write down what the reading cannot say, and that list is long enough to be the output rather than a caveat appended to it. The five steps above genuinely produce this much: the order, the sizes, what stands beneath each piece, where any stated loss stops, and the tests that can move cash while all of that is true. Every one of those is checkable by somebody else from the same figures, and being checkable is what makes them worth having.

Now the other list. The reading cannot say how likely any of the four walked losses is. Saying so would need a loss distributionA table or a curve saying how often losses of each size have arrived. Nothing on this platform holds one for any pool. , and no record of how often a loss of any size has arrived in a pool of anything stands behind these figures. The reading cannot say whether the losses in this pool would arrive together or spread themselves out. Nothing in the sizes describes the correlationHow closely two things move together. Applied to a set of borrowers it asks whether trouble tends to reach many of them at the same time. between the underlying borrowers, and that single unknown is the difference between a pool that gives up 2.0 per cent every year and a pool that gives up nothing for years and then gives up 28.0 per cent at once. The reading cannot say when money arrives. No schedule of collections by period stands behind this pool at all. The order is the teaching. The odds are a different kind of fact, and they come from counting rather than from arithmetic on sizes.

The three absences are not the same kind of gap, and the difference between them is worth being precise about. The first is missing evidence: somebody, somewhere, has counted how often pools of a given kind have given up given amounts, and those counts are not in this record. The second is missing structure: even with the counts, a statement about this pool would need to say how its borrowers relate to one another, and inventing that relationship would be inventing the answer. The third is missing timing: the walks above have no dates on them at all, and a loss of Rs 240 crore arriving in year one and the same loss arriving spread over seven years are different events for every holder, though the arithmetic here cannot tell them apart. Three absences, three reasons, and none of them is repaired by working the arithmetic harder.

What does the finished procedure actually produce?

The six steps together leave a sentence of a particular shape, and it is worth writing that shape out because it is the deliverable. A holder's own piece feels nothing until the pool has given up a stated amount, itself a stated share of the pool; after that the holding absorbs every further rupee until it is gone; and the following named tests can move cash away from it before any of that happens. The sentence is checkable by somebody else, reproducible from the sizes alone, and it contains no forecast anywhere in it. Anything more confident than that sentence has arrived from somewhere other than the structure, and the honest response is to ask where.

The sentence is useful because of what it refuses to include: it does not call a holding safe, it does not call a loss unlikely, and it does not say what a holding is worth. It says where a stated loss stops. A reader who wants more than that is not being unreasonable, and the extra they want genuinely exists in the world: somebody does count how often losses of each size arrive, and somebody does study whether trouble reaches borrowers together. None of it is in this record, so none of it is in the sentence, and supplying it quietly would be handing over a number that came from nowhere with the authority of everything above it.