Clearing Margin Architecture: The Layers, Not the Numbers

Clearing margin architecture is a sequence of layers, each absorbing a loss the one before it could not, in a fixed order that is published in advance. Its purpose is that the failure of one participant does not become the failure of the arrangement. The ordering is the design: what each layer holds matters far less than the sequence in which the layers are consumed.

Every pricing argument up to this point priced something. A payoff arrived as a set of numbers, an argument established what it had to cost, and it stopped there. Not one of those arguments asked what happens if the side that agreed to deliver those numbers does not deliver them. The silence is a division of labour rather than an oversight. Clearing margin architecture is the other half of it.

The other half has a name and a shape. Clearing margin architecture is not mathematics, and no part of it can be derived from anything else in this subject area. The architecture is an ordering, agreed in advance and written down, that decides which resources absorb a failure and in what sequence. The ordering is the whole of what there is to learn. The quantities attached to any part of it are set and published by the clearing body itself, and where they come from is set out below.

What is clearing margin architecture for?

Begin with the arrangement rather than the resources. When two sides agree a position, somebody has to stand between them so that neither has to investigate the other. ClearingThe arrangement standing between two sides of an agreed position, leaving each to face the arrangement rather than the other side. is that standing between. Once clearing is in place, each side faces the arrangement rather than the person opposite them. Whether that particular person is good for the position stops being a question either side has to answer.

Notice what has happened to the problem rather than assuming it has gone away. The problem has been concentrated. Every side now faces one place. If that one place cannot meet what it has taken on, everybody is affected at once instead of one pair being affected. Concentrating a problem is only an improvement if the concentrated version is then engineered, and clearing margin architecture is that engineering. The architecture exists precisely because the arrangement has made itself the point through which everything passes.

Here is the everyday version, and it carries everything that follows. Consider a building where each flat is separately responsible for damage it causes, and there is also a common fund everybody pays into for damage that goes beyond what one flat can cover. Nobody in that building believes the common fund makes damage impossible. The fund is not there to stop a burst pipe. The fund is there to keep one burst pipe in one flat from becoming an unpaid repair bill for the whole building. The arrangement is about containment, and containment is a different promise from prevention.

So the purpose can be stated in one line, and it is worth learning as a line rather than as a feeling. The purpose is that the failure of one participant does not become the failure of the arrangement. A defaultA participant failing to meet an obligation it has taken on. The event the whole sequence is built to respond to. is treated as a thing that will occasionally happen, not as a thing the design prevents. Everything that follows, the resources collected, the order they are consumed in, the fact that the order is written down before anybody needs it, follows from taking that sentence seriously.

The resources themselves have a name. MarginResources collected against a position and held against the participant they came from, available if that participant fails to meet what it has taken on. is what is collected against a position and held against the participant it came from. The last clause is doing real work and most readers skim it. Held against the participant it came from means that this money has a name on it, and that name matters enormously once the sequence starts running.

What does every price rest on, and where do the quantities come from?

Every price established anywhere in this subject area rests on one thing that is almost never said out loud. The argument runs: here is a payoff, here is a bundle of already priced things that delivers the same numbers in every outcome, therefore the payoff costs what the bundle costs. Read it again and look for the hidden clause. The bundle delivers those numbers. Delivers. The argument assumes, without ever pausing on it, that what has been agreed will actually arrive.

The assumption is not free and it is not automatic. Somebody has to build and maintain the machinery that makes it a reasonable thing to assume, and the machinery costs resources to run. Clearing margin architecture is what maintaining that assumption looks like when it is built out of resources instead of out of words. Naming that machinery is the whole of what pricing theory can say about it, and everything quantitative belongs to the body that operates the arrangement.

And here is the boundary, drawn now rather than at the end. The mathematics has nothing whatever to say about how much of this machinery there should be. Not a little to say, not something approximate to say. Nothing. A quantity of margin is a policy decision made by the clearing body, informed by judgements about how far things might move and how long a position takes to close. No no-arbitrage argument settles any of those judgements, so no derivation anywhere in this subject area ends in a quantity of margin.

The structure is teachable and the quantities are not. A stated number would present a policy choice as though it were a result, and it would be wrong the moment the clearing body revised that choice.

What are the layers, in order?

A layerOne stage of the sequence, consumed in full before the next stage is touched at all. is one stage of the sequence. The published order in which those stages are consumed has a name of its own, the waterfallThe published order in which the layers are consumed, so called because what one layer cannot absorb falls to the next., and the name is a good one: what one layer cannot absorb falls to the next, and never sideways. Here is the sequence, named by what it is rather than by what it holds.

First, the margin collected from the participant whose position moved against it, held against that participant alone. Second, further margin collected from that same participant, held not against what has already happened but against what might happen between the failure being recognised and the position being closed. Third, the resources that same participant had already contributed to a shared pool. Fourth, the clearing arrangement's own committed resources. Fifth and last, the shared pool contributed by every other participant.

Read the list once more and notice something about the first three. All three are resources belonging to the participant that failed. The sequence works through everything with that participant's name on it before it reaches anything belonging to anyone else, and it does so in a specific internal order of its own. The property is deliberate and it has a name worth carrying: own resources firstThe design property that a failing participant's own resources are exhausted before anybody else's are reached at all.. Own resources first is the single most important thing about the shape, and the reason it was built that way comes below.

What does each layer absorb, and in what sequence?

Each layer answers a different question, and the questions are what distinguish them. A reader is tempted to think of the layers as differing only in size, as though they were buckets of increasing capacity. The layers do not differ in size alone. Each layer differs in what kind of loss it is there to meet, and that difference is why the order is the one it is rather than some other one.

The first layer meets movement that has already happened. The position moved against the participant, that movement is a fact, and the resources collected against that position are the first thing reached. The second layer meets movement that has not happened yet. Between the moment a failure is recognised and the moment the position is actually closed out, time passes, and things move during it. The stretch of time between recognition and close-out is the only reason the second layer exists at all, and it is why the second layer is separate from the first rather than being more of the first.

The third layer is where the character changes without the ownership changing. The third layer is still the failing participant's own resources, but it is the portion of them that had been committed to the common arrangement rather than posted against a particular position. The fourth layer is the arrangement's own capital, and the fifth is mutualisedContributed by many participants and available to absorb the failure of any one of them, so a loss reaching this layer is shared rather than borne alone.: contributed by everybody, available against any one failure.

Look at what the picture shows about the untouched layers. The fourth and fifth are untouched not because they were spared by anybody's decision but because the third was never exhausted. No layer is reached by judgement, negotiation or discretion; a layer is reached only when the layer before it has nothing left in it. That is what makes the arrangement a sequence rather than a set of resources somebody chooses among under pressure.

Why is a sequence not the same thing as a pile?

The distinction between a sequence and a pile is the one most readers slide past, and it is worth slowing down on. Imagine two arrangements that hold identical resources contributed by identical people. In the first, a loss is worked through those resources in a published order, one holding at a time, each exhausted before the next is touched. In the second, a loss is spread across all the holdings at once, each giving up a share. The resources are the same in both. The arrangements are not remotely the same thing.

The difference is who knows what, and when. In the sequence, every participant knows where it sits and knows the order, so every participant knows in advance whether a given failure will reach it at all. In the pile, every failure touches everybody a little, and nobody knows. A sequence converts an uncertainty about who is affected into a certainty about where each participant stands, and that conversion is most of the value of the design.

The building analogy holds here too. A building where every repair is charged to the common fund and split across all flats is a pile. A building where the flat responsible pays first, and only what is left over reaches the common fund, is a sequence. Both buildings might hold exactly the same money. The second building asks a question the first never asks. Whose problem was this? The answer comes in resources before it comes in words.

Why is the sequence itself the design rather than a detail?

Now take the property named earlier and look straight at it. The failing participant pays first, and it pays twice. Its margin goes first, and then its own contribution to the shared pool goes next, and only after both of those are gone does the sequence reach anything belonging to anybody else. Two separate pockets, both with the same name on them, both emptied before the boundary is crossed.

Why build it that way? Because of what it does to the incentive of every participant before anything goes wrong. If a participant's own resources are the first thing consumed when that participant fails, then the cost of taking on more than it can carry lands on it first and lands on others only afterwards. Putting a participant's own resources at the front of the sequence is how the arrangement makes each participant bear the consequence of its own position before it becomes anybody else's problem. Reordering those layers changes that, whatever else stayed the same.

The other half of the design is that the order is published in advanceFixed and written down before it is needed, rather than decided at the moment when everybody has a stake in the answer.. Think about when the order would otherwise have to be settled. The order would be settled during a failure, at the exact moment when every participant has a direct interest in the answer and a reason to argue for whichever ordering spares it. A rule agreed then is not a rule; it is a negotiation between people who all want different things and are all under pressure.

The fire escape route in a building is painted on the wall while nothing is happening. The route is not decided during a fire, and not because anybody doubts the good sense of the people in the building. A route agreed in advance is a fact everybody can rely on, and a route agreed during the fire is an argument. The clearing arrangement makes exactly that choice, for exactly that reason, and writing the order down before it is needed is not administrative housekeeping. Writing it down is the mechanism.

Where does this touch the pricing argument, and where does the connection stop?

Vagueness on this point produces bad reasoning in both directions, so it is worth being precise: there is exactly one point of contact. Every price established anywhere in this subject area assumes the position will be honoured. The clearing arrangement is the machinery built because that assumption is not free. The contact point is that assumption, and the relation is a dependence rather than a derivation.

Now the limit, and it is sharp. The mathematics does not say how much of the machinery to build. The machinery is answering a question the mathematics never asks. The pricing argument asks what a payoff must cost given that it will be delivered. The clearing arrangement asks what happens when it is not. The two questions are different, and no amount of work on the first one produces an answer to the second.

The dependence runs one way and it is worth stating in that direction. Prices lean on the arrangement; the arrangement does not lean on the prices. In a derivation that concludes with a price, the entire apparatus described here is standing silently underneath the word delivers, holding it up. In a clearing rulebook, nothing was produced by a derivation, and expecting one is a category error rather than a gap in the reading.

What goes wrong when the structure is read as an assurance?

The cost of the mistake is a confidence the design itself does not claim to support, and it is worth naming what that confidence usually does to a reader. The confidence stops them asking further questions. Somebody who believes the structure settles the matter never looks at what the arrangement actually publishes, never asks how far the layers reach, and never notices that those things move over time. The design is unusually honest about its own limits, and a reader who is less honest than the design has not been misled by it.

Where do the thresholds, rates and dates come from?

No threshold, rate, percentage, minimum, multiplier or effective date attaches to any layer in the account above, and none appears below it. The layers are named by what they do rather than by what they hold, and the drawings carry no axis of values. Every such quantity belongs to the body that publishes it, and an account of the structure alone is what remains correct a year after it is written.

The reason is simple once stated. The structure and the quantities are different kinds of thing and they move at different speeds. The structure is a design decision that changes rarely, if at all, and can therefore be written down and taught. The quantities are policy parameters set by the clearing body, reviewed and revised as conditions change, and published by that body rather than derived by anybody else. A teaching account that prints one of those quantities is wrong from the moment somebody revises it, and the reader has no way of knowing when that happened.

One qualification. The dates in the reference table below record when those sources were consulted, and say nothing about when any part of a clearing arrangement takes effect.

How does somebody reading a clearing rulebook actually use this?

Here is where the structure earns its keep. Somebody opens a published arrangement for the first time and finds a long document full of defined terms, cross references and quantities. Knowing the shape in advance turns that document from an undifferentiated wall into four questions with findable answers, and the questions are the same whichever arrangement is being read.

1. Find the list and check that it is a list rather than a poolThe part of the document that names the resources available against a failure is the place to look. Resources named as a set of stages are a sequence. Resources named as a single fund drawn on at discretion are something else, and everything below changes.
   The word to look for is the order, not the amounts.
2. Read the order, and read it as the design rather than as a listThe stages are recorded in the sequence the document gives them. That sequence, not the inventory, is what determines how a failure travels and whom it reaches.
   Two arrangements with the same stages in a different order are different arrangements.
3. Find the boundary between own resources and everybody else'sMark the point in the sequence where the failing participant's own resources are exhausted and somebody else's begin. Everything before that point is a participant bearing its own consequence; everything after it is shared.
   Ask how many separate holdings of the failing participant sit before that point.
4. Take every quantity from the publisher at the moment it is neededThe document will carry quantities. They are read from the current published text and dated at the moment of reading, and no quantity is carried across from a description written by somebody else.
   A quantity without the date it was read on is not usable.

The same discipline runs the other way against a claim about clearing made somewhere else. A claim that a position is covered invites three questions: which layer is meant, where that layer sits in the order, and where the quantity was read and when. The three questions separate a person who has read the published arrangement from a person who has absorbed a general impression that clearing makes things fine. The general impression is very common, and it is exactly the misreading those three questions catch.

The layered structure and its order are invented as a generic teaching example, not taken from any named body.
Educational material. Not advice on any investment, tax, budget or market position.