How to Read a Derivative Contract: Nine Things to Find

The document is read in nine hunts, not in the order it is printed: the two parties exactly, the headline figure and whether it ever moves, each leg and where its rate comes from, every date, the counting method on each leg, the wording for a missing benchmark reading, the wording for a failing counterparty, and the three questions the document never answers.

A derivative agreement is assembled in the order lawyers put paper together and it reads best in the order money travels. The two orders are not the same, and following the printer's order is why so many capable people close the file having formed an impression rather than a set of findings. A reading built around the money asks one question first: what will move between these two parties, and when. Every one of the nine hunts below is an input to that answer.

Why not simply start at the front and read straight through?

Consider a rental agreement. Rent, term and notice period will happen every month, so those three get read with real attention. The wording about the building becoming uninhabitable will probably never apply, and it is skimmed or skipped. The reading is shaped by the pagination and by the reader's own stamina, and the parts given least attention are exactly the parts that would decide everything on the one day they applied.

A derivative agreement rewards a reader who imposes an order on it instead of accepting the one the printer supplied. Paper of this kind is drafted from the general to the particular. Cover and parties first, then a definitions section, then a long stretch of standing wording that appears in every copy of that document type regardless of who signed it, and only then the terms that decide what money moves. A reader working forwards therefore meets a defined terma word the document gives its own specific meaning to, which can differ from what the word means in ordinary speech. before knowing which of the defined terms will matter, and reaches the economics with their attention already spent.

The routine crosses the document rather than walking down it. Five kinds of place get visited, in an order that has nothing to do with where they were printed. One kind of place it never lands on at all is the boilerplatethe standing wording a document carries in every copy, drafted once and reused, rather than negotiated afresh for each arrangement., and the standing wording is not unimportant, only not what this twenty minutes is for.

Nine hunts follow. Each one is a single search with a single recorded result. A reading which cannot be shown to somebody afterwards is an impression, so every result is written down rather than kept in the head. Two of the nine are spent on wording that will probably never be triggered, and those two are the ones most worth keeping.

One: who are the two parties, exactly?

Find the full legal namethe exact registered name of a company, which is often not the shorter name it trades under or the name on its building. of each side, and then find which of the two the document calls by which label the whole way through. Documents of this kind rarely repeat the names after the front page. A document picks a label for each side and uses it for the rest of the text. Every obligation read afterwards is attached to a label rather than to a name, and a reader who has not pinned the two together spends the rest of the reading half wrong about who owes what.

Every single later finding is a statement about one of these two, so the pairing is found first rather than assumed, and a reader who has the labels crossed will attribute half the obligations to the wrong side without ever noticing. The pairing costs a minute and it is the cheapest error prevention in the whole routine.

There is a second thing to find in the same place, and it catches people who work at large organisations. The entity that signed has to be the entity actually of interest. A group can have many registered entities, the cash flow that matters can sit in one of them, and this agreement can have been signed by another. The name governs, not the logo.

One more thing can sit in this same hunt. Some arrangements run under a longer framework document signed once, with each individual deal recorded in a shorter confirmationthe short document recording the agreed terms of one particular deal, as distinct from the longer framework signed once and used for many.. Where that is the shape, find both pieces of paper. The pair together is the agreement and neither one alone is.

On the agreement worked through here, both sides are named in full: Chitrakoot Cements Limited and Saranga Capital Limited. Chitrakoot Cements is labelled the fixed payer throughout, and Saranga Capital is labelled the floating payer throughout. Both pairings go into the notes before anything else is read.

Two: what is the headline figure called, and does it ever move?

The largest number on the cover, the word the document uses for it, and then four words written beside it before any further reading: this is not a cash flow. On this agreement the figure is Rs 1,000 crore and the document calls it the notional. Every other quantity in the whole document is a rate, a date or a method for counting time. The only rupee amount either party will ever actually hand over is a difference produced by putting all of those together.

There is a check that settles this in seconds and does not depend on knowing anything about derivatives at all. Find the part of the document that lists what gets paid and when, and look for the headline figure in it. On this agreement it is not there, and it is not there on any agreement of this shape. The name on the cover is a label somebody chose. The list of movements is evidence, and evidence beats a label every time.

Write three things down at this point and then stop: the figure itself, the word the document uses for it, and whether that figure appears in the list of movements. The three lines are the whole product of this step, and a reader who carries the figure forward in their head rather than on paper is the reader who later quotes it to somebody as though it were an amount at stake.

Three: what does the fixed leg say, and where does its rate come from?

Each of the two legs gets its own hunt and both hunts ask for the same four things: which party pays this leg, what rate it uses, where that rate comes from, and how often it falls due. The interesting result is the difference between the two answers rather than either answer on its own, so keeping the shape identical across the two legs is deliberate.

On the fixed leg the four answers are quick. Chitrakoot Cements pays it. The rate is 7.20 per cent a year. The rate comes from the document itself, meaning somebody negotiated it and typed it in. The leg falls due once a period. The whole of the fixed leg is inside the paper itself. Nothing happening outside the document can ever change what was written down, and that is why this hunt is over in a minute.

The rate is recorded with its period attached, always. A rate written down without saying what stretch of time it refers to is not a finding, it is half of one, and the person who reads the note six months later has no way of recovering the missing half from what was left behind.

Four: what does the floating leg say, and where does its rate come from?

Now run the identical four questions on the other leg and watch the second and third answers change character completely. Saranga Capital pays it. The rate is not in the document. The document carries an instruction for finding a rate instead, and the leg falls due once a period against the same headline figure.

The fixed leg carries a number and the floating leg carries a procedure, and a reader who writes a number in the floating leg's box has written down something the agreement does not contain. The mistake is a real one rather than a pedantic one. A note written that way ends up saying the arrangement pays six per cent. The agreement actually says it pays whatever a particular reading turns out to be on a particular morning. For the first period the reading happened to be 6.00 per cent a year, and for later periods it is not knowable from this document at all.

A vague note here is worth nothing, so the instruction comes out in pieces. Which reading. Taken on which day. At what time of that day. Published by whom. And whether anything is added to it or taken off it before it is used. Five answers, and a blank in any of the five should be uncomfortable. A blank in this list is a question that has to be put to somebody at the exact moment there is least time for it.

Five: which dates does the agreement fix, and what does each one do?

Four kinds of date matter and each does a different job. The effective date, the point from which period one begins to run. The reset dates, one standing at the front of each period, where the floating leg's reading gets captured. The payment dates, one standing at the back of each period, where money actually crosses. And the end date, beyond which nothing further happens.

Here is the trap, and it is the reason this step produces work rather than a lookup. Agreements of this kind very often state a rule for producing dates instead of printing the dates themselves. The list the reader needs has to be built before it can be used. A rule can be short and still leave ten dates to write out, and a reader who copies the rule into their notes and moves on has recorded an instruction while believing they recorded a calendar.

So the list gets produced. Then comes the check the step exists for: the last date the list produces must be the end date the document states. If those two disagree, either the rule was read wrongly or the document has a problem, and both of those are far cheaper to find on a quiet afternoon than on a payment morning.

One more thing belongs in this hunt. Some of the dates the rule produces will land on days the money system is shut, and the document carries its own business day conventionthe document's own rule for shifting a date that lands on a day the money system is closed, so that the shifted date is fixed by the paper rather than by habit. saying which way such a date moves. Record what this document says, in this document's own words. The usual convention is not what governs, and it is not what goes into the notes.

Six: how is a period counted, and is it counted the same way on both legs?

The counting method named for each leg goes into the notes, both of them, even when the two look identical. The counting hunt sits on its own rather than folded into the two leg hunts, and the separation is the entire point. Two methods recorded side by side make any difference between them visible in one glance, and a difference between them changes what every settlement on this arrangement comes to. The same two methods recorded in two separate places, twenty minutes apart, hide the difference completely.

The hunt's silences are as telling as its questions. No view is wanted on what the methods do, or which is more common, or which one favours whom. All the hunt asks for is two names copied into two adjacent boxes and then looked at together. Copying two names is a two minute job with an outsized payoff, and it is skipped constantly because both entries usually turn out to be the same and the reader concludes afterwards that checking was unnecessary.

On the agreement worked through here the two entries agree and the first period is a full one, giving a day count fraction of 1.0000 on both legs. A full opening period is convenient rather than typical, and it is what lets the arithmetic later in this guide stay short enough to follow without pen and paper.

Seven: what does the agreement say happens when the benchmark is not there?

Hunt for the wording that tells the floating leg what to use when the reading it normally leans on is not published on a reset date, and for the separate wording covering the benchmark ceasing to be published at all. Such wording is the fallbackthe wording saying what a floating leg is to use when the reading it normally depends on is not available on the day it is needed., and it is worth understanding why it is so consistently skipped: everything found in hunts two through six will be used every single period, and this will probably be used never.

Fallback wording is negotiated rather than standard, so knowing what one agreement says about a missing reading tells precisely nothing about what the next one says. Negotiated rather than standard: that is what to carry away from this hunt. The finding therefore has to be recorded for each arrangement separately, and a colleague who summarises theirs from memory should not be relied on.

The document's own construction gets copied into the note, even when the wording strikes the reader as obvious. The value is not in understanding it now. The value is that on the morning the reading does not appear, the answer is already in the notes instead of somewhere in a document that three people are opening at once.

Eight: what does the agreement say happens when the other party is not there?

The second rarely used stretch of wording. Hunt for four things: what counts as a failure by either side, what the other side may do about it, how the arrangement is brought to an early terminationthe ending of an arrangement before its stated end date, because something has gone wrong rather than because it has run its course., and how anything owed at that moment is worked out.

The boundary in this hunt is not decorative. The note records what the document says and nothing whatsoever beyond that. A party's actual rights, and what either side has to place or report along the way, depend on conditions set outside the paper entirely. The Reserve Bank of India settles those conditions at rbi.org.in, and they change. The finding here is a quotation from the document itself, not a conclusion about anybody's rights.

Hunts seven and eight together are the pair almost nobody can summarise about their own arrangement, and that is worth sitting with for a second. A capable person asked what their agreement says the floating leg uses if the reading is missing will usually pause. Asked what their counterpartywhoever stands opposite in an arrangement settled privately between two parties, and whose ability to pay sits behind every figure in it. failing would trigger, they will usually offer a general impression rather than a finding. Both answers were available the whole time, in a document sitting in a drawer.

Nine: what is not in the document at all?

A reading routine that ends with the document closed leaves a feeling that the arrangement is now known. The arrangement is not known, and the ninth hunt exists to stop that feeling forming. The hunt looks for what the paper never addresses, and there are three such gaps.

The first is the one every reader eventually asks: what is this arrangement worth today? Producing that figure takes a dated schedule of expected readings for the benchmark, covering every period that has not yet happened, obtained from whoever publishes such things. No such schedule stands behind the agreement worked through here, and so its value today is never computed. A signed agreement will not supply it either. The document states what will be computed; it never states what a stream of future computations is worth today.

The second and third absences are of a different kind. The agreement never states what somebody has to report about its existence, nor which body receives that report, nor how soon. Nor does it state what either side has to place as collateral behind it. Both are set outside the document and both move, and that is why they are routed rather than written.

Write those three unanswered questions down beside the nine findings. A reader who does not record the gaps stops remembering that the gaps are there. Recording the gaps is the closing instruction of the whole routine, and it turns the end of a reading into a short list of errands rather than a feeling of completeness.

What do the nine findings come to when they are put together?

Nothing in the routine asked for a computation. But the findings, laid next to each other, produce the first period's arithmetic on their own. Watching them do it is the proof that the reading was a reading rather than a skim.

The hunts found a headline figure of Rs 1,000 crore. The hunts found 7.20 per cent a year printed against the leg Chitrakoot Cements carries. An instruction sat where the other leg's rate would be, and following it at the first reset produced 6.00 per cent a year. And they found a counting method giving a day count fraction of 1.0000 across a full opening period.

Run the printed rate across the headline figure for a full period and the fixed leg lands at Rs 72.00 crore gross. Put the reading through the same arithmetic and the other leg lands at Rs 60.00 crore gross. Neither of those two amounts is ever handed over. The amount that crosses is the difference between the two, Rs 12.00 crore. The leg Chitrakoot Cements carries is the heavier of the two, and the Rs 12.00 crore moves out of Chitrakoot Cements towards Saranga Capital.

There is a second route to the same figure and it is worth taking. Two routes agreeing is how the findings are known to have been read correctly. How far 7.20 per cent a year stands above 6.00 per cent a year is 1.20 percentage points, and carrying that distance across the headline figure for a full period gives Rs 12.00 crore once more. Running it backwards is the same check in reverse. Setting the Rs 12.00 crore back against the notional that produced it returns 1.2 per cent, the very distance the check began with.

One caution attaches to that arithmetic and it is not a small one. Only the first period can be worked at all. The second period needs a second reading of the benchmark, taken at the second reset date, and no such reading exists behind this guide. Any figure printed against period two would have to be invented, and an invented figure is indistinguishable from a real one once it is written down.

Who actually runs this routine, and what do they do with the findings?

Four kinds of reader run something like this, and the same nine findings get used very differently by each. Seeing that is what stops the routine feeling academic.

A treasury officer at a company like Chitrakoot Cements runs it because they will be the person telephoned on a reset morning. Their findings live in a one sheet note beside the agreement, and the two entries they most want on it are the ones from hunts seven and eight, precisely because those are the two they will never be able to recall under pressure.

A lender looking at the same company runs it for a different reason. The lender is not interested in whether the arrangement is clever. A lender wants to know what leaves the company's account and when, and whether anything in the wording could bring a large obligation forward. Hunts five and eight carry almost all of the weight for a lender, and the headline figure carries almost none. The first move a lender makes with a report is to check whether the writer confused the two.

An analyst outside the company runs a version of it on whatever the company has published, and mostly finds the routine stopping early for want of a document. An incomplete reading is a finding too. Recording that hunts three, four and six could not be completed is far more useful than guessing at them, and it is the honest form of the note.

And a household runs the same shape of routine without calling it that. On a floating rate home loan, the four questions on the floating leg are exactly the four to ask: which reading the rate follows, when it gets refreshed, who publishes it, and what the lender adds on top. Most people can answer the first and the last of those four and very few can answer the middle two. The same gap leaves a treasury desk reading its own agreement at speed on a bad morning.

What can the finished reading still not tell?

Somebody who has run all nine hunts can describe this arrangement completely, name every date it fixes, and work its first period from their own notes. Such a reader still cannot say whether anybody should be inside it.

Look at everything that would have to be settled before anybody could answer that. One is what this party has already committed to, and on what footing. Another is what its own money coming in and going out does when a benchmark reading moves, and over what stretch of time. A third is what it would actually do on the two difficult days that hunts seven and eight are about. None of that is in the document, none of it is in the nine findings, and none of it is on this platform. A complete reading is a description, and a description is not a judgement about whether to be in something.

The gap is not a weakness in the routine. The routine is working exactly as intended. A reader who finishes knowing exactly what they still have to look up is in a far better position than a reader who finishes with an impression, and the whole purpose of hunt nine was to make sure the first kind of reader is the one who walks away.

Chitrakoot Cements Limited, Saranga Capital Limited and the agreement between them are invented.
Educational material. Not advice on any investment, tax, budget or market position.