Collateral, Guarantee and Credit Enhancement Compared

Collateral is an asset the borrower pledges, so a claim has something to reach if payment stops. A guarantee is a second party's promise to pay when the borrower does not. Credit enhancement is protection built inside a structure, usually by placing other pieces beneath the one held in the order losses are taken. All three answer different questions, so which one is held decides what has to be asked next.

Underneath that sits one idea worth holding on to before any of the detail. None of the three is an improvement to the borrower. Each one is a separate arrangement bolted alongside the promise to pay, written at the same time or later, and each is tested only at the moment the promise fails. Failure is the worst imaginable moment to discover how an arrangement was drafted. So almost every question worth asking is about what somebody established before the support was needed, rather than about the name the support carries.

What do collateral, a guarantee and credit enhancement all have in common?

Every one of the three is a second answer to a question the borrower's own position has already answered badly.

Think about what a lender has in front of them before any support appears. There is a contract saying what is owed and when. There is some cash generation that is meant to pay it. There is a size, measured against that cash generation. There is some distance between where the readings sit today and where they would have to get to before the answer changed. Where all four of those come back comfortable, nobody asks for anything else. Support is what gets asked for when one of them does not.

So the presence of credit support is itself information: it says that somebody, at some point, sat down with the borrower's own position and was not comfortable with it. The reflex runs the other way. A borrowing described as secured, and as carrying a promise from a second party, sounds stronger than one described as neither, and in the narrow sense of what a claim can reach it may well be. But the arrangements did not appear because everyone was relaxed.

The everyday version runs like this. A landlord takes a deposit of three months, asks for a co-signer on the agreement, and wants a further three months of rent up front. Not one item on that list describes the flat. The rooms, the floor, the rent and the term describe the flat; the deposit, the co-signer and the advance describe what the landlord concluded about the tenant after asking. A tenant who reads the deposit as evidence that the flat is a good one has read the wrong half of the notice.

Now translate it. One worked example of the ratios a lender takes before any support is discussed runs on four declared amounts, and those four amounts belong to no borrower at all. Rs 800 crore is the debt. Rs 300 crore is the cash held. Earnings before interest, tax, depreciation and amortisation come to Rs 200 crore across a year, and the interest on that debt to Rs 40 crore across the same year. The four amounts give a gross leverage of 4.00 times a year of earnings, a net leverageA reading that sets what is owed less the cash held against a year of earnings. How it is struck, and what it refuses to settle, is covered separately. of 2.50 times the same year of earnings, and an interest coverage of 5.00 times. How those readings are struck and what each one does not settle is covered separately. The narrower point is this: the four readings are the borrower's own answer. Credit support is somebody else's answer, written because the first one was not enough.

What does collateral add to a claim, and who provides it?

Collateral is an asset the borrower pledges. If payment stops, a claim then has something specific to reach.

The same four headings run through all three supports below, so they come one at a time here. The comparison only becomes readable when they are filled in the same order every time.

Who provides it: the borrower itself, out of what it already has. That is the first thing that separates collateral from the other two, and it does more work than it looks. No new party enters the picture. No money arrives from anywhere. The borrower has simply agreed that one of the things it already had is now attached to this particular claim rather than being freely available to everybody.

Collateral attaches to a named asset, or to a described pool of assets. Not to the borrower in general. A claim over a business as a whole and a claim over one identified machine are two very different objects, and the difference shows up on the day somebody has to go and collect.

When it is tested: at the moment payment stops and somebody has to reach the asset and turn it into money. Until then it does nothing whatsoever. There is no month in which collateral pays a coupon.

Collateral changes what is recovered. Recovery is the whole of it, and the flat statement is worth making because the natural reading is wider. Collateral does not make the borrower pay, and it does not make failure any less likely; it changes what is left afterwards and nothing at all about whether afterwards arrives.

The line between whether failure arrives and what is left afterwards is the one everything below is built on. Whether a borrower keeps paying is settled by what cash the business generates and what else that cash has to cover. A pledged machine adds nothing to that cash. On the day the money is short, the machine is still a machine. The pledge decides who gets to reach that machine once the money has run short, and how much of it they reach before anybody else does.

What must be established about collateral before it counts for anything?

Six questions stand between being told that an asset has been pledged and actually holding a claim on one.

The definition is the easy half. The six questions below it are where lenders actually lose money.

Is the asset identified specifically, or described so loosely that it could be almost anything? A pledge over one named machine, with a serial number, is an object. A pledge over the current assets of the business is a category whose contents change every week. Such a pledge is perfectly ordinary, and completely different from a pledge over one named machine. The wording decides which of the two is in hand.

Has the chargeThe legal hook that ties a claim to one particular asset. Until it has been created and entered on a register, a lender has a description of an asset rather than a hold on one. over it been created and registered, and is the registration something a lender has seen with their own eyes rather than been told about? Registration is the single most consequential item in the whole comparison. A described pledge and a registered charge are separated by a step somebody has to actually take, and the step happens outside the lending document. How a charge over an asset is created, registered and ranked is set by the Ministry of Corporate Affairs at mca.gov.in.

Does anybody rank ahead on the same asset? An asset can carry more than one claim, and the order between them is not decided by who asked first or who is owed most. Two lenders can hold a perfectly valid charge over the identical machine and receive very different amounts from selling it. Where the ranking of claims comes from is covered separately.

Can the asset be sold separately from the business it sits inside? Separability is a trap that reads as a technicality and is not one. Suppose the pledged asset is the only kiln in a cement works. Selling it recovers money and stops the operation that was supposed to generate the cash to pay the lender. The pledge is real, it is registered, nobody ranks ahead, and exercising it destroys the operation that was going to repay the rest. A pledge over a delivery van does not carry that problem. The two get written up in the same sentence and behave nothing alike.

Who valued it, on what basis, and how old is that valuation? A number in a lending file has an author and a date, and both of them matter more than the number does. A number with no author, no basis and no date attached to it is not a valuation, and a lender who accepts one has accepted a figure rather than an answer.

How long does reaching it take? An asset that can be sold in a week and an asset that can be sold after a process running for years are both collateral. A claim that has to wait longer than the lender can wait has, in practice, less than it was told it had.

What does a guarantee add, and how is it different from collateral?

A guarantee is a second party's promise to pay if the borrower does not.

Run the same four headings, in the same order, and watch every one of them come out differently.

Who provides it: somebody other than the borrower, and that is the whole point of the arrangement. Collateral rearranges what the borrower already had. A guarantee brings in a party who was not previously involved and who has now taken on an obligation of their own.

A guarantee attaches to the obligation, rather than to any asset. Nothing is pledged, nothing is registered against a machine, and nobody has to identify anything with a serial number. The guarantor has promised in respect of the borrowing itself.

When it is tested: when the borrower fails and the guarantor is called. Note the two steps in that sentence. The failure and the call are not simultaneous, and the gap between them is where a good deal of the argument happens.

A guarantee changes who is being analysed. A guarantee does not remove a credit question, it adds one. The claim is only as good as the weaker of the two arrangements, and the weaker arrangement is not always the one written by the weaker party.

The difference between a weak party and a weak document carries more weight than it looks. Two things sit behind a claim carrying a guarantee. The guarantor is a party with their own capacity to pay. The guarantee is a document with its own wording. A very strong guarantor giving a very narrow promise can be worth less than a modest guarantor giving an unconditional one. Analysing only the party, or only the document, leaves half the arrangement unexamined.

The everyday version is one most households have seen. Somebody co-signs a relative's loan. The lender did not stop analysing at that point; they started analysing the co-signer, asked what the co-signer earns, what else the co-signer has already promised, and what the co-signer would still be able to pay in the year the borrower could not. A guarantee inside a lending file works exactly that way, and the only difference is scale.

What must be established about a guarantee before it counts?

Five questions, and the fifth one swallows the other four.

Is it for the whole amount, or for a stated part of it? A promise covering forty per cent of a borrowing and a promise covering all of it are both guarantees and are described by the same word in most conversations about them.

Is it a promise to pay when asked, or a promise to make good whatever is left after the borrower has been pursued to the end? Those two are years apart in practice, and they are separated by who does the work in the meantime. The first sends the guarantor a demand and expects money. The second sends the lender through a process against the borrower first, and only then produces a number the guarantor has to meet. A claim valued today on money that arrives at the end of the second route has been valued on something rather different from what the lender assumed.

Is it capped, and is the cap on the amount or on the period? Both kinds of cap exist. An amount cap says how much. A period cap says for how long the promise stands, and a promise can quietly expire before the borrowing does.

Does the guarantee survive a change to the borrowing it supports? Lending arrangements get rearranged all the time, and a rearrangement agreed between the lender and the borrower can release a guarantor who was not asked. A release discovered afterwards is a very expensive way to learn that the document had a condition in it.

And then the question containing the rest: what does the enquiry say about the guarantor? Run the same five questions on the second party that were run on the first, in the same order. Ask what is owed by that party, what pays it, how large it is set against that, how much room there is, and what stands behind that party in turn. A promise from a party nobody has examined is a promise nobody has valued.

What is credit enhancement, and why is it not a promise from outside?

Credit enhancement is protection built inside a structure rather than promised from outside it.

The same four headings again, and this time they come out stranger than either of the first two.

Who provides it: the structure, by arrangement. Most commonly by placing other pieces beneath the one held in the order losses are taken, sometimes by a reserve funded at the start and held back, and sometimes by funding a pool with less money than the pool is worth. One thing is missing from that list: nobody is promising anything to anybody.

Enhancement attaches to one piece of a pool, rather than to a borrower. A reader coming from the first two supports has to reset here. Collateral and a guarantee both sit alongside a borrowing made by somebody. Enhancement sits inside a funded structure and describes the position of one piece relative to the others.

When it is tested: continuously, as losses arrive, rather than once at the end. Collateral waits for a failure. A guarantee waits for a failure and then for a call. An order of loss is working from the first rupee that goes missing, quietly, without anybody being called.

Enhancement changes who in the queue takes the loss first. And here is what is easiest to miss and most worth having: most credit enhancement adds no money from anywhere. The pool loses exactly what the pool was always going to lose, and the arrangement decides only the order in which the pieces feel it.

The everyday version is a shop with three people standing behind one till. Takings for the day are whatever they are. If the three have agreed that the first person absorbs the first shortfall, the second absorbs the next, and the third is touched only when both of the others are exhausted, they have not increased the takings by a single rupee. The three have decided the order of disappointment. Placing one claim deliberately below another is subordinationPlacing one claim deliberately below another, so that whatever goes wrong reaches the lower one first. How claims are ranked against each other is covered separately., and it is the commonest form credit enhancement takes.

How much protection stands beneath each piece of the invented pool?

Only one of the three supports can be worked in figures, and that one is an order of loss. Collateral and a guarantee each need a document that states terms. An order of loss needs nothing beyond the sizes of the pieces.

The structure below is stated in full: a pool of Rs 1,200 crore of receivablesAmounts that other people already owe, arising from goods or services already supplied. Money expected in, rather than money in hand., funded by a senior piece of Rs 960 crore, a mezzanine pieceThe middle piece of a funded structure. It sits above whichever piece takes losses first and below whichever takes them last. of Rs 180 crore and an equity piece of Rs 60 crore. The three pieces are 80.0, 15.0 and 5.0 per cent of the pool, and unusually they add to 100.0 with nothing left over, so a reader can add the column below with no hedge attached to it.

Now the enhancement arithmetic: one subtraction, one addition and nothing cleverer than that. Beneath the senior piece stand the mezzanine and equity pieces together: Rs 180 crore plus Rs 60 crore is Rs 240 crore, or 20.0 per cent of the pool. Beneath the mezzanine piece stands the equity piece alone: Rs 60 crore, or 5.0 per cent of the pool. Beneath the equity piece stands nothing at all: Rs 0 crore and 0.0 per cent, printed in the table below as a nought rather than left out of it.

Name the base aloud at every use. Twenty per cent of the pool is not twenty per cent of the senior piece, and the two readings differ by the size of that piece. The base is the next block, and quoting the wrong one is the commonest error in reading enhancement.

And now the refusal, in the same block as the arithmetic: not one rupee of that Rs 240 crore came from outside the pool. Every piece was used to fund the receivables in the first place. The pool will lose whatever the pool loses, and the arrangement decides only who feels it first. How likely any of it is goes unstated. The sizes of the pieces say nothing about how often losses of any size arrive in a pool, so the senior piece is nowhere described as safe.

Why does the same Rs 240 crore read as two different numbers?

Because a share is a share of something, and the something changes the number without the arrangement moving at all.

Now one thing changes and nothing else. The numerator stays identical, set against the one held instead of against the pool. Nothing in the structure has moved: the same three pieces are funding the same receivables in the same order.

The rupee route says the same thing without any rounding in it at all. The two readings sit 5.0 points apart, and 5.0 points of a Rs 1,200 crore pool is Rs 60 crore, precisely the equity piece. A share quoted with no base attached is therefore not a small untidiness. A bare percentage is a number waiting to be attached to whichever quantity the next reader happens to have nearest.

One more base, this time showing the trap from the awkward side. Beneath the mezzanine piece stands the equity piece alone, Rs 60 crore. Against the pool that is 5.0 per cent, a clean figure that prints exactly. Against the mezzanine piece itself it is 33.33 per cent, and that one does not terminate: Rs 60 crore over Rs 180 crore runs on forever and is printed here rounded to two places. Anybody who then multiplies 33.33 per cent back by Rs 180 crore lands slightly below Rs 60 crore and wonders where the rupees went. A printed figure is a display, not an input, and the amount is what the work runs on.

So ask for the amount and the base together, every time.

What must be established about credit enhancement before it counts?

Five questions again, and the last of them is the one nobody asks.

Which form is it? An order of loss and a funded reserve behave differently and are both called enhancement. One is a position in a queue; the other is an amount of money actually set aside and held. Being told that a piece is enhanced is being told which kind of arrangement is involved, and nothing else.

How much of it stands beneath the one held, expressed as an amount and as a share of the pool rather than as a share of that piece? Both readings are arithmetically correct and they are constantly swapped, as the two blocks above show. Ask for the rupee amount alongside the percentage and the swap becomes impossible.

Is it there at the moment that piece needs it, or only after everything else has been settled? A reserve released back to somebody once a test is passed is not standing beneath that piece on the day the test fails. Timing is part of the arrangement, not a detail of it.

Does it replenish once it has been used, or is it spent? A reserve topped up out of later collections and a reserve drawn down once are the same size on day one and nothing alike in year three.

And the hardest one to ask: does the enhancement depend on the same thing going right that the protected piece already depends on? Protection that fails in exactly the circumstances that would call on it is protection in name. Answering the question needs a record of what makes a pool lose money, and no such record sits inside the structure itself, so the question stands open.

Are the three supports one protection at three strengths?

The three are not one protection at three strengths, and getting that wrong is the most expensive mistake in the comparison.

The tempting picture is a ladder. Collateral at the bottom, a guarantee above it because a named party sounds better than a machine, credit enhancement somewhere else because it sounds technical. On that picture a lender asks how much support there is and stops.

Look at the three definitions again with the ladder in mind and it falls apart immediately. Collateral changes what is recovered and leaves the chance of failure exactly where it was. A guarantee changes who is being analysed and opens a second enquiry. Credit enhancement changes who in the queue is hit first and leaves the total loss in the pool precisely where it was. Three different parts of the problem, three different pieces of evidence, three different next steps.

Set them out on the same six headings and the separation becomes obvious. No two of the three ever fill the same cell.

How does the support held decide the next question?

Because the support type is a branch in the enquiry rather than a score on it.

The branch is the practical form of everything above, and it is why the last row of that table matters more than any of the others. Being told that a claim carries support tells a lender which of three completely different investigations they now have to open. Holding collateral sends the lender to a register and to whoever else has a claim on the same asset. Holding a guarantee sends the lender back to the beginning with a different party's name at the top of the file. Holding a piece of a structure sends the lender to the arrangement itself, to the amount standing beneath that piece and to the base the amount was quoted on.

A lender who has been told there is support and stopped asking has learned only that somebody, once, was worried.

What can none of the three supports do?

Four things, and this guide closes on them rather than on a comfort.

None of the three makes a borrower generate more cash. Every rupee that pays a borrowing on time comes from the business, and no arrangement written alongside the promise adds to it. Pledging a machine does not sell more cement. A co-signer does not lift anybody's takings. An order of loss inside a pool does not collect a rupee more from anybody who owes the pool money.

None of the three changes what is owed, or when it falls due. Dates, amounts and terms sit exactly where they sat before the support was written, and they sit there afterwards too.

And none of the three is tested until the arrangement it supports has already failed. Every one of them is therefore a claim about the future behaviour of a document. The honest description explains why the six questions and the two sets of five above are almost entirely about drafting, registration and timing rather than about strength. The value of a support on the day it is written is the value of the sentences in it, read by somebody who has assumed the worst.

And the fourth thing none of them can do: none of the three can be worked on Palash Cements Limited at all. No record states what that issuer has pledged, to whom, on what terms, or whether any party has ever stood behind its borrowing. So the support column for that issuer stands empty below, with the reason printed inside it rather than a plausible entry.

An empty row with a reason inside it is worth more to a reader than a confident entry typed to make a table look complete. An empty row also names exactly what to go and find: the security document, the charge register, and the name of any party standing behind the borrowing. All three exist somewhere for a real borrowing.

How do four different readers actually use this?

The three supports are read differently depending on who is doing the reading, and the differences are practical rather than theoretical.

Before the money goes out is the only moment at which anything can still be changed, so a lender writing a facility uses the six questions then as a checklist. The registration is either arranged now or argued about later. Whether the pledged asset can be sold apart from the business is either asked now or discovered during a sale. A lender who asks these afterwards is not doing credit work, they are doing archaeology.

An analyst reading a bond document from outside uses the same six questions as a list of what to look for and, just as usefully, as a list of what is missing. A document that describes a pledge and never mentions a register has told the analyst something. So has a guarantee whose wording is summarised in three lines with no statement of whether it pays on demand. The absence is the finding.

An investor holding one piece of a structure has no borrower to run five questions on, so the enhancement questions are almost the only ones left. The form, the amount standing beneath, the base that amount was quoted on, whether it replenishes and whether it is available at the moment it would be needed are what matter. The duties placed on a trusteeA party appointed to hold rights and act for a whole group of holders together, so that each of them does not have to act alone. Those duties are set by the securities regulator. acting for the holders sit alongside that, and they are set by the Securities and Exchange Board of India (SEBI) at sebi.gov.in.

And a household asked to co-sign a relative's borrowing is on the other side of exactly this arrangement. The lender is about to run the five questions on them. The lender asks what is owed by that household, what income pays it, how large the existing obligations are set against that income, how much room there is if something changes, and what stands behind them in turn. Anyone about to sign as a guarantor is being asked to become the second borrower, and the questions above are the ones being asked about them. Reading the guarantee before signing it is the same work a lender does, done from the other chair.

Why the three supports do not sit on one scale

Nothing about the three supports moves along a single dimension.

A slider moves one quantity along a scale, and a scale carries a claim of its own: that the thing at one end and the thing at the other end are the same thing at two strengths. The claim of a single scale is false for these three. Collateral is not a weaker guarantee. A guarantee is not a stronger reserve. Sliding between them would merge three distinct arrangements and undo the distinction in a single gesture.

One relationship inside credit enhancement is genuinely worth watching move: how far a loss travels up an order before it reaches a given piece. A travelling loss belongs where a structure is read rather than where three supports are compared, and it is covered separately. Nine points stand in its place, each one a moment at which the distinction has to be performed rather than assumed, and at one of them every option is arithmetically true with only the base to tell them apart.

Palash Cements Limited, its five year bond and the pool of receivables in the drawings are invented.
Educational material. Not advice on any investment, tax, budget or market position.