Guide · Derivatives and risk · Updated 2026-09-19

The offset is a property of your account, not of your position

The margin arithmetic on this page is computed from inputs stated in the text, so a reader can redo it. The basis figures are measured from cached exchange data covering 528 sessions from 2024-08-01 to 2026-09-18. Neither set is quoted from anywhere.

Two full margins for one risk

Margin is an estimate of what a position could plausibly lose over the time it would take to close it out. That framing is what makes the double charge indefensible. Hold a basket of stocks in the cash segment and a short futures position on the index those stocks make up, and the combined position does not have two exposures to a market fall. It has almost none. What it has instead is an exposure to the gap between the two legs, which is a much smaller and quite different quantity.

Charging the cash segment margin on the long leg and the derivatives margin on the short leg adds two loss estimates that were each computed as though the other leg did not exist. The sum describes a position nobody holds. On the illustrative inputs below the sum is ₹20,50,000 against a pair whose worst measured single day across two years of data moved by ₹79,521.

Cross margining is the risk model that says so. It is worth being precise about what kind of thing it is, because the language around it invites the wrong reading. It is not a facility extended to good customers, not a credit line, and not a discount. It is the clearing corporation computing a more accurate number. The relief is a consequence of the better estimate, not the purpose of it, and that distinction explains most of the framework's shape: every restriction in it exists to make sure the better estimate is actually better.

The list is shorter than the economic logic

If offsetting risk were the only test, anything reliably negatively correlated with your book would qualify. The framework does not go there, and the reason it does not is the most instructive thing about it.

The recognised pairs are ones where the two instruments share an underlying by construction. Index futures against the constituent stock futures of the same index. Index futures against a basket of those constituents held in the cash segment. Single stock futures against the same stock held in the cash segment. In each case the relationship between the legs is written into the definition of the instruments, not estimated from their price history.

That is the boundary, and it is drawn there because correlation estimated from history is at its least trustworthy precisely when margin matters. Two series that moved together for three years can decouple in a week, and the week they decouple is the week the clearing corporation is exposed. A relationship that comes from index construction rules does not decouple, because it is not a relationship between two things at all. It is the same thing counted twice.

The framework did step across that line once, and the way it did is revealing. A facility for offsetting positions in highly correlated equity indices was introduced by circular in November 2019, and it levied a spread margin of 30 per cent of the total applicable margin on the eligible offsetting positions, to begin with, computed at client level on an online real time basis. A higher retained percentage for a statistical relationship than for a structural one is the model pricing the difference between the two, which is exactly what it should do.

The last row is a small rule with a large lesson in it. Margin relief is not additive. A position that has already had its requirement reduced on one ground is not available to reduce it again on another, and a clearing corporation that allowed it would be estimating the same risk as absent twice.

The computation, line by line

Take a long basket in the cash segment replicating a broad market index, worth ₹1,00,00,000, against a short position in futures on that index with a notional of ₹1,00,00,000. The margin rates below are stated as inputs and are illustrative: a real value at risk rate is computed per security and moves every day, and a real initial margin rate is an output of a scenario grid rather than a percentage anybody publishes. The arithmetic built on them is not illustrative.

The spread margin is levied as a percentage of that total applicable margin, which is worth pausing on. The relief is a percentage of a number that is itself computed daily from volatility, so the rupee benefit moves even when the position and the rate are both unchanged. On a quiet week the offset is worth less in rupees than on a violent one, and a treasury function that budgets it as a fixed figure has mis-modelled it.

The last column is the one a desk actually cares about. At a rate of 25 per cent the same margin supports 4.0 times the exposure it would otherwise support, which is why the framework matters far more to a participant running at its capital limit than to one with idle collateral. It also explains why the operational requirements below are worth an institution's effort and are almost never worth a small account's, a point the closing section returns to.

A partial match is matched partially

Hedges are rarely exact. Lot sizes are what they are, and a view is often expressed on part of a book rather than all of it. The framework handles this in the obvious way, and the obvious way has a consequence people miss.

Only the overlap is offset. Take the same ₹1,00,00,000 cash basket against a short futures position of ₹60,00,000 instead. ₹60,00,000 of each leg pairs off. The remaining ₹40,00,000 of cash is an ordinary long position and carries ordinary margin.

A hedge covering 60 per cent of the position delivers 53.3 per cent relief, not 75 per cent. The headline rate applies to the matched portion and the matched portion only, so the effective relief on a book is a weighted figure that has to be computed rather than assumed. Anyone sizing a hedge partly for its margin effect should be working from the second column of that table and not from the rate.

The benefit is a property of the account, not of the position

This is the practical spine of the subject and the part generic explanations skip, because it is administrative rather than conceptual. The concept is simple. The machinery is where the benefit is actually lost.

A clearing corporation's risk system works from position records. Those records carry a client identity within each segment. Nothing in the record of a long cash position says anything about a futures position held somewhere else, and nothing in the futures record points back. For the two to be treated as one pair, the link between the identities has to be established in advance and in a form the system can act on.

Where both legs clear through the same clearing member, that member files the client details with the clearing corporation, and the link is made. Where the two legs clear through different clearing members, the necessary agreements have to be in place between them first, which is a legal step rather than a data one. There is also a narrower route for the case where both legs sit inside the derivatives segment, being index futures against constituent stock futures, where the derivatives clearing member supplies client details and the agreements are not required, precisely because no second segment and no second intermediary are involved.

The institutional shape of the market is why the second route exists at all. A large participant commonly settles its cash segment business through a custodian and clears its derivatives through a clearing member, and those are different entities with different contractual relationships to the client. The clearing member in the derivatives segment has to intimate the client's details to the clearing corporation together with a letter from the trading member or custodian identifying the same client on the cash side. Two firms, one client, one piece of paper that makes the risk system treat them as connected.

There is a further wrinkle that tells you how seriously the identity question is taken. A client may maintain two accounts with their respective members for the purpose of availing the cross margin benefit, and for that purpose only. The framework would rather create a second, restricted account than allow the link between identities to be loose.

The consequence is the sentence this whole guide exists for. A trader who is economically hedged but operationally unrecognised gets nothing. Not a reduced benefit, not a delayed one. The full margin on both legs, indefinitely, with no error message, no flag on a statement and no line item explaining the absence. The margin is simply what it is, and it looks exactly like a correct number, because from the risk system's point of view it is one.

The bargain has a second side that is easy to overlook while the benefit is being enjoyed. Positions that are linked for margin purposes are linked for default purposes too. A framework that computes one requirement across two segments is a framework that can act across both when something goes wrong. That is not a hidden penalty, it is the necessary counterpart of the relief, but it is a reason the agreements are substantive documents rather than a form.

The residual the model is pricing

Why is the offset partial? Not caution for its own sake. The pair genuinely moves, and how much it moves is measurable rather than a matter of opinion.

For a one to one long cash and short futures pair held on the same contract, the day's profit and loss is the change in the index less the change in the futures price, which is exactly the negative of the change in the basis. That quantity is directly observable from daily settlement data, so the residual risk of a perfectly hedged pair can be measured rather than assumed.

Three things in that table are worth reading carefully. The hedge works: the pair's variability is a small fraction of the outright leg's. The fraction is stable across five very different indices, from 14.1 to 15.6 per cent, which suggests it is a property of the futures and cash relationship rather than of any one market. And it is nowhere near zero. The 99th percentile one day move ran from 0.35 to 0.57 per cent of notional depending on the index.

The roll rows are the interesting ones. A session on which the near contract changed moved the pair by 0.55 per cent on average, against 0.10 per cent on an ordinary session, a factor of about 5.6. The residual in this relationship is not spread evenly through time. It clusters at the moment the derivative leg is replaced, which is also the moment the offset itself is at risk.

Set the measurement against the charge. At a 25 per cent spread margin the pair retains ₹5,12,500 of margin, which is about 15 times the ₹35,165 that the measured 99th percentile one day move implies on this leg. The model is conservative relative to the cleanest possible version of the pair, and there are four good reasons it should be. A close out takes more than one day. A real basket is not the index, so it carries tracking error on top of the basis. The roll is a discontinuity rather than a draw from the same distribution. And the measurement above covers a particular two year window, which contained what it contained and not the events it did not. The point of the measurement is not that the charge is too high. It is that the residual is real, positive and concentrated, which is why the offset is a percentage and not a cancellation.

What withdraws the offset, and when

The benefit is computed at client level on an online real time basis. That phrasing, which comes from the framework itself, has a consequence worth stating plainly: it is not a status conferred for a period. It is an output recomputed from the current set of positions, so it appears when the pair appears and leaves when the pair leaves, within the session.

Closing one leg is the obvious case and the least troublesome, because the trader is present when it happens. The expiry of the derivative leg is the one that causes damage, because it is a scheduled event that removes the offset from a cash position nobody touched, on a date set months earlier. A participant who has sized a book against the reduced requirement finds the requirement back at full size on a leg that has not moved. The guide to the spread margin benefit and the expiry day it disappears works through the same mechanism inside the derivatives segment, where it bites harder still.

The third case is administrative and is the one that catches institutions rather than individuals. If the arrangement between two clearing members lapses, or a client is moved between accounts, or the filed details stop matching the positions, the legs stop being recognised as related and the offset stops. The position is unchanged. The paperwork is not.

The fourth is the early pay-in exclusion covered above: a cash leg on which margin benefit has already been given for early pay-in of securities or funds is not counted among the offsetting positions.

What changed, and what most published explanations still say

The framework has a history, and the current shape of it is recent enough that a good deal of what is written about it describes an earlier version.

The April 2024 step is the one that dates the most competing material. Before it, an explanation of this framework that said the two legs must share an expiry month was correct. After it, that sentence is wrong as a general statement, and it appears in a great deal of currently published material because nobody revisits an explainer once the rule under it moves. If you are reading any description of cross margining, the quickest test of whether it has been maintained is to look for what it says about expiry dates.

There is a second confusion worth heading off, because two reforms with similar vocabulary have been running in parallel. Position limits and margin are different frameworks answering different questions: one governs how large a position may be, the other what it costs to hold. Measuring positions against a limit in delta adjusted, futures equivalent terms is a change to the first. Cross margining sits in the second, and its offset is a spread margin levied on positions the clearing corporation identifies as eligible and offsetting. A book that happens to be delta neutral across instruments is not thereby a cross margin pair, and a trader who assumes otherwise will be surprised by the margin file rather than by the position report.

Who this framework is actually for

An honest scope note, because the gap between what the rules permit and what an ordinary account can use is wide here.

The facility is stated to be available to all categories of market participants, so nothing excludes an individual by rule. In practice it is used overwhelmingly by institutional participants, and the reasons are structural rather than discriminatory. The relief matters in proportion to how binding your capital constraint is, and the operational cost of claiming it is broadly fixed. A participant whose whole book is constrained by margin will spend the effort. A participant with collateral to spare will not notice the benefit and will certainly not notice its absence.

The eligible pairs also fit an institutional book more naturally than a retail one. Holding a basket that replicates an index in the cash segment, against index futures, is a routine institutional position and an unusual individual one. The single stock version is the case an individual might plausibly run, and it runs into lot sizes, into whether the member has enabled the facility for client accounts, and into how the cash leg is recorded.

So the correct posture for an individual trader is to understand what the framework is and then to ask, rather than to assume. Put four questions to your own member. Is the facility enabled for client accounts. Which of my cash segment holdings count as offsetting positions. What has to be filed, by whom, and when. And what happens to my margin on the day the derivative leg expires. If the answers do not come back clearly, the working assumption should be that the benefit does not apply to you, because that is what the margin file will assume.

The larger lesson generalises past this one facility. Margin is not a fact about a position, it is an output of a model applied to records, and the records are the product of arrangements somebody had to make. Reading the rule and reading the plumbing are two different jobs, and only one of them decides what you are charged. Our curriculum treats that gap between the rule and the machinery as a subject in its own right, because it is where most of the costly surprises in a real book come from. Work through how derivatives margins are maintained through the day and what pledged collateral does and does not buy you alongside this, and the shape of the whole margin system becomes considerably easier to hold in your head.