Market Surveillance: What an Exchange Actually Sees
Market surveillance is the continuous examination of orders, trades and positions against tests written in advance, looking for shapes in the data worth a second question. Surveillance sees what was entered and never why. The watching sits inside this machinery for one reason: the determined price it watches decides what is collected that evening from every position in the contract, and a pushed input would collect the wrong amounts from parties who did nothing.
A position is marked every evening and money moves on the result. Where the number behind that movement comes from, and what would happen if somebody could nudge it, is less often stated. The answer is uncomfortable in a useful way: the machinery has no opinion about its own inputs. Fed a wrong figure, it will collect wrong amounts with perfect discipline, from parties who did nothing at all.
Everything here runs on one invented set of figures, so those come first. There is a reference asset, invented for this illustration and standing in for whatever a contract might reference. The spot price of that asset is a PRICE of Rs 2,000.00/-, and the same Rs 2,000.00/- is the exposureThe value of the referenced thing that a position stands on. Exposure is the base a margin percentage is struck on, and it is not an amount that changes hands. carried by one unit. The reference asset pays nothing at all while it is held. A payout during the holding period would change every carried figure that follows, so the absence of one matters.
Financing costs 6.50 per cent a year. Carry one unit for one year and the cost is Rs 2,000.00/- multiplied by 1.065, or Rs 2,130.00/-. That figure is the agreed price on the contract used throughout, and it is a PRICE rather than anybody's view of where the reference asset is going. Against the position sits collateral of Rs 160.00/-, being 8.0 per cent of the Rs 2,000.00/- exposure, and that 8.0 per cent is a teaching figure rather than a requirement anybody has set. The Securities and Exchange Board of India (SEBI) at sebi.gov.in sets the framework, clearing corporations set the actual requirement under it, and the requirement differs by contract and by day. Either limb of that collateral misleads on its own, so the two belong together: Rs 2,000.00/- of exposure standing on Rs 160.00/- posted is 12.50 times, and a gap of Rs 80.00/- between two stated prices, being 4.0 per cent of the Rs 2,000.00/- exposure, is 50.0 per cent of the Rs 160.00/- posted.
What is market surveillance actually watching?
SurveillanceThe continuous examination of orders, trades and positions against tests that were written down before anybody needed them. examines three things and the relations between them. Surveillance examines every orderAn instruction entered into the market. An order may trade, it may be amended, and it may be withdrawn again without ever having traded. entered, whether or not the order ever became a trade. The examination covers the trades that did happen, in what size and at what moment. And it examines the positions carried afterwards, in whatever terms the record keeps them. Then it looks at how those things sit against each other: sizes against sizes, moments against moments, and the same thing appearing again tomorrow.
Surveillance watches for shapes against tests stated in advance rather than for wrongdoing, and that difference is not a technicality. A test is a description of a shape. A shape is a fact about data. Whether anything improper produced that shape is a separate question that data alone was never able to answer, and letting those two run together teaches the reader to convict on arithmetic.
The everyday version does the work of three paragraphs. The office that runs a market building does not watch the shopkeepers. The office could not, and it would learn very little if it tried. The till records are what it watches, and they arrive every evening in the same shape from every stall. When one till shows the same unusual entry at the same minute five evenings running, somebody walks over and asks about it. Nobody has decided anything. Somebody has noticed a shape and gone to find out what made it, and nine times in ten the answer is dull and arrives in one sentence.
Surveillance examines orders, trades and positions. What is it examining them for?
Why does this belong in the same machinery as the collateral?
The link between the watching and the collateral is almost never drawn, and once drawn it never comes apart. The whole arrangement runs on numbers that nobody inside it chose. A settlement priceThe price determined at the close of a day, by a stated method, that every open position in a contract is revalued against. is determined at the close and every open position in the contract is revalued against it. A count of positions decides who is looked at and what is reported. Both of those are inputs. The payments that follow are outputs, and outputs cannot be argued with: they are just what the arithmetic produces from what it was fed.
Work it on the invented contract. The agreed price is Rs 2,130.00/-, a PRICE. The settlement price determined for the day is Rs 2,050.00/-, also a PRICE. The gap between those two stated prices is Rs 80.00/-, or 4.0 per cent of the Rs 2,000.00/- exposure. A long position pays that gap: a PAYMENT of Rs 80.00/- leaves the balance, so Rs 80.00/- of the Rs 160.00/- posted is left, being 50.0 per cent of what was put up. Nothing has gone wrong anywhere in that sequence. The whole of it is an ordinary evening.
Now suppose the determined price had come out Rs 20.00/- lower, at Rs 2,030.00/-. Rs 20.00/- is 1.0 per cent of the Rs 2,000.00/- exposure. The gap between the agreed price of Rs 2,130.00/- and that price is Rs 100.00/-, so the PAYMENT called from every long position in the contract is Rs 100.00/- a unit instead of Rs 80.00/-, and every short position receives Rs 20.00/- a unit that it should not have. The balance left of the Rs 160.00/- posted, at an invented 8.0 per cent of exposure, falls from Rs 80.00/- to Rs 60.00/-. The extra Rs 20.00/- is one eighth of everything that was put up, taken from parties who did nothing except hold a position through an ordinary day.
The machinery would work exactly as designed and would collect the wrong amounts from the right people. No system breaks in that story. No control fails. The arithmetic was never asked to have a view about the number it was handed, so every calculation is correct, every payment is made on time, and every figure reconciles. The inputs are watched for that reason, and not only the outputs, and distortionA movement in a determined price produced by something other than what the price is meant to measure. in a determined price is a different order of problem from an error in a payment. An error in a payment is found by adding up. Everything downstream of a pushed input agrees with it, so adding up finds nothing.
The everyday version is a vegetable market where every stall settles up at the gate against one weighing machine. If the machine reads light by a little, every stall settles perfectly and every stall settles wrongly, and no amount of recounting the money at any stall will ever find it. Somebody has to go and look at the machine. Surveillance is the part that goes and looks at the machine.
One more turn of that arithmetic shows why nobody can quietly absorb a distortion. Take six positions in the same contract, three long and three short, one unit each. On the determined price of Rs 2,050.00/-, the three long positions are called for Rs 80.00/- a unit. The total collected is Rs 240.00/-, and the three short positions receive the same Rs 240.00/-. On a price pushed to Rs 2,030.00/-, the three long positions are called for Rs 100.00/- a unit. The total collected is Rs 300.00/-, and the three short positions receive Rs 300.00/-. The extra Rs 60.00/- taken from one side is the extra Rs 60.00/- credited to the other. Nothing has been created and nothing has been destroyed; something has simply moved from parties who did nothing to parties who did nothing either.
Suppose the settlement price for the day came out Rs 20.00/- lower than it should have, at Rs 2,030.00/- rather than Rs 2,050.00/-, against an agreed price of Rs 2,130.00/-. Who pays for that, and how much?
Which of these figures moves, and which is only determined?
Every figure in this machinery is one of three things, and the label belongs inside the sentence rather than in a note underneath it. A PRICE is what is agreed, quoted or determined. A PAYMENT is what actually leaves or reaches an account. A NET is what is left once obligations in both directions have been set against each other. A settlement price of Rs 2,050.00/- and a payment of Rs 80.00/- sit comfortably in one sentence, and only one of them moves, so mixing the three is the commonest error in this machinery.
Set the ordinary evening beside the pushed one and watch which column changes. The right hand column is not something that happened, but what would follow if a determined price were pushed by Rs 20.00/-, being 1.0 per cent of the Rs 2,000.00/- exposure.
| The figure | Which of the three it is | On the determined price | If pushed by Rs 20.00/- |
|---|---|---|---|
| Agreed price on the contract | a PRICE | Rs 2,130.00/- | Rs 2,130.00/- |
| Settlement price for the day | a PRICE | Rs 2,050.00/- | Rs 2,030.00/- |
| Gap between those two prices | a difference between two PRICES | Rs 80.00/- | Rs 100.00/- |
| Called from one long position | a PAYMENT | Rs 80.00/- out | Rs 100.00/- out |
| Left of the Rs 160.00/- posted, at an invented 8.0 per cent of exposure | a balance, and not a payment | Rs 80.00/- | Rs 60.00/- |
| Taken from every long position that should not have been | a PAYMENT | nil | Rs 20.00/- |
The figure that gets watched sits in the price column and the figure that hurts sits in the payment column, and they are two different objects that a single sentence can make look like one. Keeping the two apart is the whole reason for the labels. Somebody who thinks the settlement price is money will look for the loss in the wrong place, and somebody who thinks the payment is a price will imagine a market moved when what moved was a balance.
A quantity is a different kind of thing again, so the table carries none: every quantity says whether it is NOTIONAL or EXPOSURE, and the exposure the 8.0 per cent is struck on is Rs 2,000.00/- a unit. The exposure did not move in either column of that table. The collateral that backs it moved instead.
What can it see, and what can it never see?
Most people credit the watching with far more reach than it has, and the assumption is worth catching before the limits are set out.
Of the reasons somebody entered a particular order, how much of them can surveillance see?
The detail available to surveillance is genuine. It sees what was entered and at what moment. The size of that entry is there, against the book the order met. So is whether the order traded, was amended or was withdrawn. So is the position that stood at the close of the day. And because the recordThe account of what was seen, what was asked and what was answered, kept so that somebody who was not in the room can check it later. goes back further than any one day, surveillance also sees whether the same shape has appeared before.
Intention, anything that happened outside the market, and the reason anybody did anything at all are exactly what surveillance cannot see. There is no field in any record for why. The reason an order was entered lives in a conversation, a cash need, a mandate, an instruction from somewhere else, or nothing more than somebody changing their mind, and not one of those things is data that any watching arrangement holds.
So everything surveillance produces is a question about a shape, and the answer to that question lives outside the data. This is the honest half that gets left out of every popular account, and leaving it out is what turns a careful arrangement into something readers imagine as a machine that catches people. Surveillance catches nobody. It notices, and then somebody asks.
What makes one pattern worth a second look?
Three relations do most of the work, and it is worth seeing that none of them requires reading anybody's mind. The first is size, taken against the thing the size could move rather than against any absolute figure. An order for one unit and an order that sits against the whole of what a book could absorb are different objects even where the instruction on the screen looks identical.
The second is timing, taken against a moment that matters, and in this machinery the moment that matters most is the one at which a price is determined. The same entry at noon relates to nothing in particular. Placed in the minutes from which a settlement price is drawn, it relates to a number that decides what is collected from every open position in the contract that evening.
The third is repetition. One occurrence is an event and has no shape at all. The same occurrence at the same minute for five days running is a patternA relation between entries rather than a fact about any single entry. Size against size, timing against timing, or the same thing appearing again., and a pattern is the only thing in this subject that can be examined at all, because a single entry contains nothing to compare.
Every one of those three is a relation rather than a threshold. The actual tests an exchange applies, and every value inside them, are set under the framework of SEBI at sebi.gov.in. The tests differ by contract, they differ by day, and they move. The relations hold across contracts, while the values sit with the regulator.
What size of order counts as large enough to be worth a second look in a given contract? Where does that figure come from?
The shape of an alert record carries the argument, and the sheet is worth seeing as an object. Most rows on it are filled from the data, and they are filled in the ordinary way: this was entered, it was this size, it did this to the book, this was asked, this was answered. Two rows are of a different kind. The test that matched, and the threshold sitting inside that test, are set by the regulator rather than produced by the data, and they are the two rows that decide whether there is anything to look at.
What is an alert, and what is it not?
An alertA record that a stated test matched. An alert is a question raised, not a finding made. says that a stated test matched. The whole of what it says stops there. An alert does not say that anything improper happened, it does not say that anybody has been accused of anything, and it does not carry any weight of its own into what follows. Most of what matches has an ordinary explanation, and that explanation usually arrives complete the moment somebody asks for it.
Two errors follow from confusing a match with a finding, and they run in opposite directions. The first is reading an alert as a verdict, which is the commoner of the two and the one that does real damage to people. The second is reading the absence of an alert as a statement that everything was in order. The absence never was that: a test that did not match says only that this test did not match. A conclusion is not the kind of thing this arrangement produces, so neither a match nor a silence is one.
An account is flagged by a stated test. What has been established?
How Derivatives-Market Surveillance Supports Market Integrity
The link between what is watched and what is collected is worth stating carefully rather than as a slogan. Market integrityThe condition of an arrangement's inputs being reliable for everybody inside it, rather than favourable to any one participant. is not a feeling about a market and it is not a claim that a market is fair. Market integrity is a much narrower and much more useful condition: the figures the arrangement runs on can be relied on by everybody in it, including by the people who never see how they are produced.
Follow the chain and the support becomes structural rather than moral. A determined price is an input. The collateral called that evening is computed from it. What each position holderThe party carrying the position, standing in for anybody on either side of a contract. has left of what was put up follows from the collateral called. So the reliability of one number decides whether the collateral arrangement is doing what it exists to do or is quietly redistributing money between parties on the strength of something that was pushed. Watching the inputs is what allows anybody to rely on the outputs without having to trust any particular participant.
There is a second contribution and it is easy to miss. The watching creates a record, and a shape that is examined and explained leaves behind an account of the explanation, so the record is itself part of the integrity. The account is what lets somebody who was not in the room see what happened and why the matter closed. An arrangement in which questions are asked and nothing is written down supports nothing at all.
Now the limit, and it is the sharpest thing in the whole subject. A shape consistent with something improper is also consistent with entirely ordinary trading, and the data holds nothing that separates them. The carry arithmetic already worked above shows why. A one for one offset cancels rupee for rupee only at the final date. With a year still to run, the carry applies to the new spot as well, so the gap between the two contract prices moves by more than the gap between the two spot prices: a gap of Rs 80.00/- on a base of Rs 2,000.00/- corresponds to a gap of Rs 85.20/- on a base of Rs 2,130.00/-, since Rs 1,920.00/- carried for a year at 6.50 per cent is Rs 2,044.80/-. A position holder correcting for that difference adjusts a position a little every day, at whatever minute their own arrangements run. On a screen that is a repeating shape at a repeating minute. In the arithmetic it is bookkeeping.
Why are the tests settled before they are needed?
Because a test written after the shape has been seen is a test somebody can argue was written for them, and the argument arrives at exactly the moment there is least room for one. The same argument governs the method behind a settlement price and what follows a call that is not met. Noticing that it is the same argument in all three places is the strongest evidence it is right.
An arrangement that has to work under pressure settles its rules while nobody yet knows who they will catch. Fix a test in advance and it applies to whoever it happens to match, including the people who wrote it. Fix it afterwards and every party can see, before agreeing to anything, whether the choice helps them. The value of a rule written in ignorance of who it will land on is precisely that ignorance, and it cannot be recovered later by good intentions.
The household version is a chit arrangement among neighbours that decides how a missed payment is treated. Decide it when everybody is paying, and it is a rule. Decide it on the evening the first person misses, and it is a judgement about that person, whatever it says on the paper. Everyone in the room knows the difference. The arrangements that survive settle these things while they are still boring.
Why are the tests settled before the shapes they catch have appeared?
What is written down, and who ends up reading it?
The watching is only half of it, and the record of what was seen, what was asked and what was answered is the other half. That record is what makes the arrangement checkable by somebody who was not in the room, which is a different and much stronger property than any individual judgement being correct. A shape noticed and never written down leaves nothing behind. A shape noticed, asked about, explained and recorded leaves an account that can be examined years later by somebody with no stake in it.
SEBI at sebi.gov.in sets what has to be recorded, what has to be reported onward, to whom and by when. The international origin is worth naming once: the International Organization of Securities Commissions (IOSCO) at iosco.org has set out cross-border principles under which watched markets keep records of what they see and what they do about it, and what applies in India is SEBI's version of that principle rather than the principle itself. Where the arrangement crosses into rates or currencies agreed between two parties rather than on an exchange, what is reported and to whom is a matter for the Reserve Bank of India at rbi.org.in.
What is set by an authority rather than by the arithmetic?
Every row below is set by the authority printed inside the row, each differs by contract and by day, and each moves. The 8.0 per cent used in the arithmetic above is a teaching figure and belongs to none of these rows.
| What it is | Who sets it | Stated here |
|---|---|---|
| What an exchange watches for, what it records and what it passes on | SEBI, sebi.gov.in | nothing |
| What positions are reported, by whom, and by when | SEBI, sebi.gov.in | nothing |
| How much of one contract a single participant may carry | SEBI, sebi.gov.in | nothing |
| What counts as manipulation, and what follows a finding of it | SEBI, sebi.gov.in | nothing |
| Who may carry a position in one of these contracts at all | SEBI, sebi.gov.in | nothing |
The machinery above extends to a second market as one more addition to the table.
What does surveillance not do for one position holder?
Three negatives, and none of them should be softened. A comfortable version of any one of them disappoints at the worst possible moment.
Surveillance does not decide whether a price is right. A determined price is arrived at by a stated method, and watching the inputs to that method is not the same as certifying the number that came out. Nobody in this arrangement issues a certificate that a settlement price of Rs 2,050.00/- was correct.
Surveillance compensates nobody for anything. A payment made on a distorted input is not reversed because a shape was noticed. SEBI at sebi.gov.in sets what follows a finding, and that is covered separately.
And it is not a service provided to the position holder at all. This is the one that surprises people. Surveillance exists so that the arrangement's own inputs can be relied on by everybody inside it. The benefit reaches an individual only through the arrangement working, and anybody who takes it as a protection promised to them personally has misunderstood who it is for.
A position holder asks whether surveillance will get their money back if a price was pushed against them. What is the honest answer?
How does anybody actually use this, on the day it touches them?
Three readers meet this arrangement from three directions and none of them reads it the way a headline does. A position holder meets it as a query about their own activity. A query is a request for the one thing the data cannot supply: the reason. The person asking already has the shape. The sentence that explains it is what they lack, and only one party can produce it.
Somebody assessing an arrangement from outside, whether they are lending against a position or deciding whether to face this market at all, uses it differently again. The useful question is never how many shapes were noticed. A large count reveals almost nothing and can just as easily mean the tests are loose. Three questions decide whether anybody outside the room can check anything: what is written down, who is able to read it, and what happens when a question goes unanswered.
And a household weighing whether it wants to be exposed to a market at all can use the same three questions without knowing any of the machinery. The same instinct asks a committee how the accounts are kept rather than asking whether the treasurer is honest. The first question has an answer that can be checked. The second one does not.
The query that was not answered, and what the silence bought
A position holder receives a query about their own activity. A letter from an institution feels like an accusation to somebody who has never had one before, and that is how they read it. So they answer it defensively, or they answer it late, or they decide that anything they say can only make matters worse and say nothing at all.
Who makes this mistake: people who have never been asked anything by an institution in their life, and that is most people. The reaction is entirely understandable and it is not carelessness. Silence feels like the cautious choice, and in a great many parts of ordinary life it is.
The cost of the silence is the whole of the point. An alert is a question about a shape, the ordinary explanation usually exists, and it is usually mundane enough to fit in three sentences. The explanation is the only part of the record that could not be produced from the data, so the one thing that turns a mundane shape into a serious matter is that it was never explained. The silence does not leave the record neutral. It leaves the shape standing as the whole of the record.
Kill it with the distinction rather than with a warning. An alert says a stated test matched. The answer lives outside the data. Only the person who was asked can produce it, and not answering removes the only thing that was ever going to close the matter.
A position holder is asked to explain a shape in their own activity and decides to say nothing. What happens to the shape?
What can never be settled about being in a watched market?
All of this arrives at one question. Does the watching make a market trustworthy enough to be in? The question is completely reasonable, and it is the one anybody actually wants answered.
The question has no general answer. Not because the answer would be uncomfortable, but because it turns on four things that differ by contract and by day: what is actually tested, what is actually recorded, who is able to read that record, and what follows when a question is not answered. SEBI at sebi.gov.in sets every one of those, and each of them moves.
There is a second reason, and it is the sturdier one. Understanding how a call is struck is not a reason to be on the receiving end of one. The machinery explains itself completely and still says nothing whatever about whether any particular person should be inside it.
References
| Source | What is set out there | Where |
|---|---|---|
| Securities and Exchange Board of India | What an exchange watches for, what it records and what it passes on; what positions are reported, by whom and by when; how much of one contract a single participant may carry; what counts as manipulation and what follows a finding of it; and who may carry a position in one of these contracts at all | sebi.gov.in |
| International Organization of Securities Commissions | Cross-border principles under which watched markets keep records of what they see and what they do about it, with the version that applies in India being the one SEBI has set | iosco.org |
| Reserve Bank of India | Arrangements on rates and currencies agreed between two parties rather than on an exchange, and what such an arrangement is reported as | rbi.org.in |
| Research Papers in Economics | Academic work on cleared markets and the examination of trading records | ideas.repec.org |
| arXiv Quantitative Finance | Preprint repository for work on cleared exposures and market microstructure | arxiv.org |
The reference asset, the contract, the clearing corporation, the clearing member, every position holder and every figure drawn here are invented.
Educational material. Not advice on any investment, tax, budget or market position.
