How Option Volatility Surfaces Describe Market Pricing

A volatility surface records what a set of contracts was being quoted at during one instant, with every premium rewritten into a single unit so that contracts at unlike strikes and unlike lengths can finally be set against each other. A surface describes the pricing of those contracts. A surface does not describe the reference asset, and a surface stops being current the moment quoting moves on.

Underneath that answer sits one awkward fact about premiums. A premium has the strike and the length of the contract baked into it along with everything else. Two premiums quoted on the very same reference asset are therefore not comparable numbers. Rewrite each premium into a unit that has had those two things removed, and what is left is a set of numbers that can honestly be read against one another. A surface is that set, laid out so the reading takes a second rather than an afternoon.

One warning belongs at the top rather than halfway down. The working record standing behind these two contracts carries no figure for the quantity a surface cell would hold, and no run of prices that already happened. The absence turns out to be the spine of the argument. Almost everything worth saying about what a surface describes can be said, and checked, with an empty grid.

Why can two premiums on one reference asset not simply be set side by side?

Instinct says they can. The instinct is the reason this mistake is so common, and it is worth taking seriously before it is dismantled. Both numbers are in rupees. Both are quoted on the same underlying thing. Both are called premiums. Set side by side, one is bigger. Surely the bigger one is, in some sense, dearer.

Larger for what reason, though? The question dissolves the instinct completely. A premium is not a single measurement. A premium is the price of a specific promise, and the promise is described by several things at once. The contract names a level at which somebody may deal. The contract runs for some length of time. The contract points one way or the other: a right to buy, or a right to sell. A change in any one of those changes the premium, without anything the comparison was meant to isolate having moved at all.

Take the level first. A contract struck at a level the reference asset is already close to is a different animal from one struck at a level it is nowhere near, and the premiums will say so loudly. Take the length next. A contract with a year left to run has more chances to become worth something than one with a fortnight left, so it costs more, and that difference is bought and paid for in time rather than in anything else. Take the direction last. A right to buy and a right to sell are not two flavours of the same thing; they face opposite ways, and their premiums sit at different levels for that reason alone.

Now the version from ordinary life, one most households have already lived through. Two stalls stand outside the same office building. One is let for a whole season, the other for a single month, and the season letting costs a great deal more. Does that make the season stall a better spot, or give it better footfall, or make it worth more per day? The higher rent does nothing of the sort. The higher rent says that one letting covers more days than the other. Until somebody puts both on the same footing, one rate per day for the same size of pitch, the two rents are simply not answering the same question, and putting them in a column beside each other only makes the confusion look tidy.

A premium carries the strike and the time left to run inside it, so two premiums differ for reasons that have nothing to do with whatever the comparison was actually meant to isolate. The difference is not a subtlety. The difference is the whole reason the machinery set out below exists at all.

What comes back when a premium is restated into a common unit?

Comparability comes back, and comparability is worth more than it sounds. The manoeuvre is set out here without any of the machinery. The machinery itself is covered separately.

A pricing model is a device that takes several already-known quantities and hands back a premium. Among the quantities handed to it are the strike and the time left to run. The device can be run backwards. Holding the premium somebody has actually quoted, holding the strike and holding the length, the question becomes what the remaining inputA quantity handed to a calculation before it runs, as opposed to one the calculation gives back when it finishes. would have to be for the device to produce exactly that premium. The number that answers the question is the restated premium.

Now look at what happened to the two nuisances. The strike went in. The length went in. Anything handed to a calculation as an input has stopped being part of what the calculation gives back. An answer stripped that way can be laid alongside another answer built the same way. The distance between strike and price was consumed on the way in, so the restated number for a contract struck near the reference asset price and the restated number for one struck far from it are in the same unit. The same goes for length. The fortnight and the year are both fed to the device rather than left sitting in the answer, so a contract with a fortnight to run and a contract with a year to run come back in the same unit.

Once that is true of every contract in a set, the answers can be arranged. Strikes across, lengths down, one restated number to a box. A grid built that way is the surface. A surface exists for exactly one reason: so that prices which could not be compared with one another can be. Everything else people say about surfaces is downstream of that single purpose, and a reader who holds on to it will never mistake the object for something grander.

The trick is an ordinary one. The same trick is performed whenever two rents are converted to a rate per square foot per month, or two loan offers to a cost per rupee borrowed per year. In each case the things that were getting in the way are fed into the conversion, and what survives is read off. Nobody thinks a rate per square foot is a mystical property of a building. The rate is a device for setting two buildings side by side. The restated premium deserves exactly the same modesty.

What does that restatement cost?

The cost is the half of the story that most explanations skip, and skipping it is how readers end up trusting a surface far more than it deserves.

The manoeuvre is worth returning to. Taking the strike and the length out of the answer required putting something in: a device that turns inputs into premiums. There is no version of the trick that works without one. And a device of that sort is built on assumptions. The device assumes something about how a price behaves through time. The device assumes something about how money is carried forward. The device assumes something about what happens between now and the last day. Such assumptions are not a footnote to the restated number. Assumptions are ingredients of the restated number.

Every number sitting on a surface is conditional onTrue only while some stated assumption holds, and quietly untrue the moment that assumption stops holding. the device that produced it, so what a surface actually buys is comparability inside one device's language, not comparability in general. Read the sentence twice. A great deal follows from it.

One consequence follows immediately. Two surfaces built on the same contracts, on the same reference asset, at the same instant of the same day, using different devices, are not comparable with each other. The two surfaces look comparable. Both are grids of numbers with the same axes and similar shapes, and everything about their presentation invites laying one on the other. But a number on the first grid answers the question what would this device need in order to produce that quoted premium. A number on the second grid answers a different question, put by a different device. Same premium, different question, different answer.

The practical failure this produces is quiet and easy to miss. Somebody in a meeting says a figure aloud. Somebody else writes it down beside a figure of their own. Neither person did anything wrong at any single step. A translation happened in the middle of the comparison and nobody announced it, so the comparison that results is meaningless. A number carried across from one device's grid to another's has been translated in silence, and silence is exactly what makes it dangerous rather than merely wrong.

None of this means a surface is useless. A surface has a passport instead. A surface travels freely inside the country that issued it and needs a stamp at every border, and the stamp is a conversation about which device produced which number, held before anyone starts drawing conclusions rather than after.

So what is a surface actually a picture of?

A surface is a picture of what people were willing to pay for a set of contracts at one moment. The picture is the whole of it, and the two words carrying the sentence are worth pulling out and holding up on their own.

WILLING. Every cell on a surface was made out of a premium that somebody quoted. Not a premium somebody calculated in the abstract, not a level somebody thinks is fair, but a level at which a real participant was prepared to deal. Quoting makes each cell a fact about behaviour. Each cell is the recorded willingness of a market, restated into a tidier unit. A shop's sticker price is not a property of the goods, and a cell is not a property of the reference asset either. Both are facts about a seller.

MOMENT. Willingness is a thing that exists at a time and not otherwise. There is no such quantity as what people are generally willing to pay. There is only what they were willing to pay when somebody looked. A surface is the record of a looking.

Put together, the two yield the sentence a reader should carry away: a surface is evidence about a market, and evidence about a market is not evidence about the asset that market happens to be quoting. A market and an asset are two different subjects. The two are related, in the sense that participants presumably think about the reference asset when they decide what to quote, but that relation runs through people's opinions, and opinions are not the asset.

A homely version makes the difference easy to feel. A walk down a street of vegetable sellers, writing down what every one of them is asking for the same vegetable, produces a real, checkable, useful record. The record says what sellers on that street were asking at that hour. The record says nothing about the worth of the vegetable, nothing about the harvest to come, and nothing about what anybody will ask tomorrow morning. The moment the sentence becomes the vegetable is expensive rather than the sellers were asking a lot, the subject has been swapped without anybody noticing, and swapping subjects is the error this whole distinction exists to prevent.

Why is a surface out of date almost as soon as it is drawn?

Because quoting does not stop while the grid is being drawn.

The answer sounds glib and is not meant to be. Continuous quoting is the entire mechanism. A market is a running process, not a state. Participants are revising what they are prepared to deal at continuously, for reasons ranging from the considered to the entirely administrative, and the process has no natural pauses that a picture could be taken during. So when somebody builds a surface, what they are doing is stopping the clock artificially. The builder takes the quoting that was going on at some instant, freezes it, restates it, and lays it out. By the time the grid renders, the process it froze has already moved.

Movement of that sort is why the instant is part of the object rather than a label stuck on the outside. A surface without its instant is not a slightly worse surface. A surface without its instant is a different and much poorer sort of thing: a set of numbers describing a state of affairs that existed at an unspecified time. Nothing in such a description can be acted on and nothing in it can be checked. A surface with no stated instant deserves the same treatment as any figure carrying no as-of date. Neither can be used.

The habit is already applied elsewhere without much thought. A balance figure means nothing until the date it stood at is known. A rate means nothing until its period is known. A vacancy count in a hospital means nothing at all unless somebody knows whether the counting happened this morning or last winter. Options quoting is a faster-moving version of the same problem, and the fast movement is why the habit has to be more disciplined here rather than less.

There is a second consequence, and it is the one that separates careful readers from confident ones. Because the surface is frozen and the market is not, two surfaces of the same contracts taken minutes apart can genuinely disagree, and neither is wrong. The two surfaces are answering the question at different instants. A reader who treats a difference between them as new information about the reference asset has read a clock as though it were a thermometer.

If the numbers across one row differ, what is that saying?

The strongest claim in this sequence needs no figure at all to make. The claim can be followed to the end with an empty grid in front of the reader, and the grid below is exactly that.

Start from something that cannot be argued with. There is one reference asset. Not five of them, not one per contract, one. The asset goes about its business without the faintest awareness of how many contracts have been written against it or at what levels, and the contracts do not reach back and alter it. Whatever it is doing, it is doing one thing.

Now consider a row of restated numbers, all for that one reference asset, all for contracts of the same length, differing only in the level they are struck at. And the numbers differ across the row. The row now needs reading.

The tempting reading is that the reference asset is somehow more prone to travel at one strike than at another. Pause on the reading for a second and it collapses into nothing. Strikes are written on contracts. The reference asset has no idea which strikes exist. The asset cannot be more prone to anything at one of them.

So the differing row is not the asset behaving in several ways at once. The row is a plain statement that the device being used to produce those numbers cannot reproduce every one of those quoted premiums at the same time with a single value. Fed one value, the device prices some of those contracts at the levels being quoted and misses on the others. The row that is not flat is the record of that miss, spread out across the strikes so that the worst of it is visible.

Take the reframing next: it changes the job the grid is doing. A grid that is not flat is evidence about how well one device fits a set of quoted premiums, and it is read as a statement about the pricing of those contracts rather than as a property of the thing they reference. The reframed claim is not smaller than the tempting one. Because the reframed claim points at the device rather than at the asset, it is more useful: a device can be changed, where an asset cannot.

A flat grid would be the special case, not the normal one. A grid on which every box carried the same number would mean the device reproduces every quoted premium in the set with one value. Such a device fits perfectly. Nobody expects that, and the interesting question was never whether the grid is flat. The interesting question is where the grid is not flat and by how much: a map of that is a map of where the device and the quoting disagree.

What happens when all this is tried on the two contracts worked through here?

Time to work it. An argument that has had no number put through it is an argument that has not been tested. The pair below runs through everything that follows, and every figure in it was composed for teaching.

The reference asset is priced at Rs 2,000.00/-. Read that as exposureThe size of the position somebody is on the hook for, as opposed to money that has actually changed hands. A quantity is either notional or exposure, and this one is exposure.: it is the size of the thing being referenced, not cash anybody has paid over. Financing is charged at 6.50 per cent a year. One year is what each of the two contracts has in front of it. Nothing reaches the holder of the reference asset across those twelve months, not one paisa of it. The absence matters: a payment arriving mid-year would move every figure below.

Both contracts are written at a strike of Rs 2,000.00/-. Yes, that is the same number as the price, and no, one has not been copied into the other by accident. The pair is struck at the moneyA contract gets described this way when its agreed level and the price of the thing it references happen to be one and the same figure on the day in question., and that is precisely what at the money means: the agreed level and the current price coincide. Two quantities, two different jobs, one shared figure, and the sharing is deliberate.

The call premium stands at Rs 180.00/-, the put premium at Rs 57.93/-. Neither was worked out here; both arrive already set. The reason is worth being blunt about: producing either one from first principles would call for the very quantity that is absent throughout, the one a surface cell would hold.

Now the demonstration, and it is one that can be carried out in the head. The two premiums sit on one reference asset, at one strike, over one length of time. Every difference listed in the first block above has been held still. The two are still not comparable: one is a right to buy and the other a right to sell. So how much of the gap between them is direction, and how much is anything worth comparing?

The parity relationshipA fixed connection tying together what a right to buy and a right to sell must cost when both carry one strike and one length. Settled in full elsewhere on this platform and drawn on here rather than rebuilt. answers it exactly, and it answers it in financing.

Step one gives the first of the two quantities. The strike, stated as a present valueWhat an amount due on a future date is worth if it is wanted today instead, once the cost of waiting has been taken off it. rather than as an amount due at the end, is Rs 1,877.9343/-. The gap between the strike and its own present value lands on Rs 122.0657/-, and that gap is nothing but the cost of waiting a year at 6.50 per cent a year.

Work the right side. Take Rs 1,877.9343/- away from Rs 2,000.00/- and the answer settles on Rs 122.0657/-. Now come at it from the premiums instead. Rs 180.00/- with Rs 57.93/- taken away leaves Rs 122.07/-. The two routes land 0.43 paise apart, and the reason is dull rather than deep: the record carries the put rounded to two places, where the relationship itself implies Rs 57.9343/-. So the link between the two premiums closes to the paisa, and not exactly: the two printed figures, checked against each other, refuse to. A text asserting an equality that its own numbers decline to deliver quietly trains a reader out of checking anything, and that does more damage than a rounding note ever could.

The demonstration comes to this. The only difference between these two contracts was direction, and the price of that difference turned out to be financing. Not a view. Not an opinion about the reference asset. Financing, worked out from a rate and a period. The argument of the first block above has now been run end to end on figures, and it lands harder that way than it did in prose.

And now the stop, in the same breath. The stop is as much a part of the lesson as the working. Restating either of those premiums into a common unit needs a device, and running such a device is a separate subject. So this is exactly as far as the worked instance goes: two premiums, one gap, one financing explanation, and then a grid drawn on labelled axes with one box ringed for these contracts and that box left as an outline.

Why is the grid drawn empty rather than filled?

A surface is a photograph, and there is no photograph to hand here. The frame stands in its place: two axes carrying their labels, boxes ruled out in the right places, and every last one of them left as an outline. A single value dropped into one of those outlines would be a manufactured figure, and it would be the most convincing figure anywhere above precisely because it sat inside a grid.

The emptiness is worth sitting with. The temptation to fill the boxes is real and it is not lazy. A worked example with numbers in it feels more generous than a frame. A filled grid looks like teaching. But the numbers would have been made up, and made-up numbers in a grid are the most persuasive kind of invention there is: a grid carries an air of having been measured. A reader who took one of those cells away and repeated it to a colleague would be repeating a figure with nothing at all behind it, and neither of them would have any way of knowing.

There is a second reason. The claims established so far are these. Two premiums are not comparable with each other. Restating them buys comparability. Restating them costs a dependence on a device. A surface is a record of willingness at an instant. A surface goes stale at once. A row that is not flat is a remark about fit. Not one of those claims needs a single populated box in order to be true, checkable, or usable. The frame therefore teaches everything a filled grid would have taught, and makes nothing up.

What may be said by somebody holding a real surface, and what may not?

Suppose a genuine grid arrives, populated, with an instant attached and the device that built it named. Ask what is in hand then.

There are three sorts of statement that can be made, and three sorts that cannot, and the boundary between them is the practical output of everything above.

On the first list: that these contracts were being quoted at these levels at that stated instant. Next, that the quoting across the set is not consistent with a single value under this device, a fact read off directly from the grid not being flat. And that this contract stood higher than that one in a unit that has had the strike and the length taken out of it, the very comparison that was impossible before the restatement and is possible now.

On the second list: that the reference asset is going to travel a long way from where it is. Next, that the market expects anything at all. A level at which somebody is willing to deal is not a hope, a forecast or an expectation, and stacking many such levels into a tidy grid does not turn them into one. And that a figure taken from a grid built with a different device can be laid down beside these, a claim the third block dealt with in full.

Everything on the first list is a statement about quoting, and everything on the second is a statement somebody wished were true. The two lists give the cleanest test available. Before any sentence about a surface is said out loud, it is worth asking which of the two it is, and if it is about what anybody is going to do rather than about what anybody quoted, it did not come from the grid.

Who actually opens one of these, and what do they do in the next ten minutes?

Abstractions are easier to hold when somebody can be pictured holding them. So: a surface arrives as an attachment, and three different people open it with ten minutes before the next thing on their calendar.

The first is somebody who has to mark a book of contracts at the end of the day. Their question is narrow and entirely mechanical: does this grid cover the contracts I hold, was it taken at the instant my book is being marked at, and was it built with the device my working papers say I use? Three checks, all of them about provenance rather than about the numbers. If any one fails, the grid does not get used, and the reason it does not get used is everything in the third and fifth blocks above. Notice that this person never asks whether the numbers look sensible. Sensible-looking numbers are not the test, and treating them as one is how a grid from the wrong instant gets quietly adopted.

The second is somebody comparing today's grid with yesterday's for the same set of contracts, and no honest use is commoner. The comparer is looking for where the two grids disagree, and the discipline is to name the subject of the disagreement. The subject is what participants were willing to quote, and the quoting changed. A change in quoting is a real, reportable observation. The failure lurking here is the slip from the quoting has moved to the asset has become riskier. One short sentence separates them, and they are different claims altogether.

The third is somebody in a room full of people who all half-remember what a surface is, and their ten minutes are spent stopping a bad sentence. Somebody will say the market is pricing in a big move. The sentence sounds like a reading of the grid and is not one. The grid says that these contracts were quoted at these levels, and that one device cannot fit them all with a single value. The person who can say that calmly, and can say why the two are not the same sentence, has done more good in that room than anybody with a better grid.

And then a household version. The same skill has nothing to do with contracts. A list arrives of what four shops were asking for the same appliance last Tuesday. The list honestly supports two statements: those were the asking prices last Tuesday, and the shops did not agree. The list cannot honestly support a statement about the worth of the appliance or its price next week, and the entire discipline of reading a surface is that same restraint applied to a much faster market and a much more impressive-looking grid.

Right, so is any of this a reason to do something?

By this point the honest question has arrived: is any of this to be acted on? The question goes unanswered above, and the reason is structural rather than nervous.

A decision would need something underneath it. Somebody would have to hold a view about how far the reference asset could travel between now and the last day, and about how much weight to put on each of those distances. No such view appears above, and a surface has never contained one. Boxes hold restated levels people quoted. Levels people quoted are not a map of where anything is going, however many of them are arranged into a rectangle.

Then there is the reader, whom no written argument can see. A written argument does not know what else the reader is holding, what would have to be sold if the arrangement went against them, what tax would fall due on the way out, or where the money has to be in March.

Last, an arrangement of this sort costs something to carry all the way to the end and something quite different to get out of early. Neither of those two figures appears anywhere above, and neither sits in the record these contracts were drawn from.

Reading a surface and using one are separate skills, and only the first has been taught above. Understanding a mechanism thoroughly is not a reason to use it, and a mechanism that has been understood is neither cheap nor attractive for having been understood.

There is one more thing to say about the contracts compared above, and it is about who decides they exist at all. Four questions sit outside everything argued above, and all four have the same address: the rungs a contract is offered at and how far apart those rungs sit; the calendar that decides when a contract starts being offered and the day it stops running; the disclosure a holder owes upward about what sits on its books; and the registration gate somebody passes through before dealing at all. Between them, the first two decide which contracts exist to be quoted in the first place. The same two therefore decide what a surface could even have boxes for.

The reference asset above, its price, its financing rate and both premiums quoted against it are invented.
Educational material. Not advice on any investment, tax, budget or market position.