The Reset Date: When the Floating Leg Is Fixed
A reset date is the day the floating leg of a swap reads a published rate and adopts it for the period about to begin. The rate is settled at the start of that period and the money is handed over at the end of it. Every period carries its own reset, and only a reset that has already happened can produce an amount anybody is able to compute.
A floating leg has to follow something, and following something means going out on a stated day and reading it. The reset date is that day. The reset date is the single moment in a swap where two parties hand the answer to somebody outside the room, and whatever that outside number turns out to be decides what one of the two legs does for the whole of the next period. Everything else in the document was argued over and agreed. The benchmark reading alone is simply collected.
Two invented businesses stand either side of one agreement throughout this guide. Chitrakoot Cements Limited is the side whose rate was settled in the document and then stopped moving: 7.20 per cent a year, outward, in every period the arrangement runs. The rate travelling back to Chitrakoot Cements is whatever the floating benchmarkA rate put together and published by somebody standing outside the two parties, so that both of them read the same number from the same place. happens to read at the time. Saranga Capital Limited occupies the mirror of that sentence, sending the benchmark reading outward and drawing the 7.20 per cent a year inward. Written across the top of both obligations is a notionalThe headline figure a rate gets applied to. It sets the size of the arithmetic on each leg and is never itself handed from one side to the other. of Rs 1,000 crore. The notional is what the two rates get applied to. The notional is not a sum either side ever hands over.
Resets come as a series, and exactly one member of that series has happened. The opening reset produced 6.00 per cent a year. Nobody publishes what a benchmark will read on a date that has not arrived, so the worth of the arrangement on a given morning, the amount for the second period and the rate covering some later stretch of time are all unknown, and unknown in the same way. Where an answer of that kind would normally sit, the question stands and the space beneath it stays empty. An empty box teaches more than a filled in guess, because a guess cannot be told apart from a reading once somebody has written it down.
A swap has a fixed leg and a floating leg. How many of the two carry reset dates?
What is a reset date, and which of the two legs does it touch?
A reset date is the day on which the floating leg goes out to a published rate, reads it, and writes that reading into the agreement as the rate for the period about to begin. A reset is a lookup with a date on it. Nothing is negotiated on that day, nothing is signed, and no money changes hands. Somebody checks a number and records it, and the recording is what makes the number binding.
The fixed leg has no resets at all, and that asymmetry is what separates the two legs: one leg is settled for its entire life on the day of signing and the other is settled one period at a time. Chitrakoot Cements pays 7.20 per cent a year. Chitrakoot Cements paid 7.20 per cent a year in the first period, it will pay 7.20 per cent a year in the last one, and there is no day anywhere in the calendar on which anybody goes and looks that figure up. The fixed rate was argued over once, written down once, and then left alone. Saranga Capital sits on the other side of an obligation that has to be refreshed before every single period, and refreshing it means going and reading something.
The household version runs alongside it, and the shape is a familiar one. A household has a rent agreement fixed for three years and an electricity supply billed on a meter. The rent needs no diary entry. The rent is written on a sheet of paper in a drawer, and the sheet of paper is the answer, so nobody in that household ever goes and looks the rent up. The electricity is different. Somebody comes on a stated day, reads the meter, and that reading decides what the next bill is built from. The meter reading day is a reset date. The meter reading day is not the day the bill is paid, and it is not a negotiation about the tariff. A number simply gets collected.
Notice what follows from that for anyone trying to keep track of the arrangement. The fixed leg needs watching once, at the start, and then never again. The floating leg needs a calendar. Every reset date is a day somebody has to do something, and the something is small and completely unforgiving: read the right series, on the right day, and write it down with the date attached. A reader who thinks both legs reset has not misread a detail. The mistake is to imagine an instrument in which the whole document reopens every few months, and no such instrument exists.
Why is the rate fixed at the start of a period and paid at the end of it?
Two moments matter in every period of a floating leg, and they sit at opposite ends of it. The reading is taken at the start of the period, on the reset date. The money changes hands at the end of the same period, on the payment date. The rate is known before the period runs and the amount is settled after it has run, and a reader who holds those two moments apart has most of the mechanism.
Ask the obvious question. If both events concern the same period, why not do them together and save everybody a diary entry? Because the two sides want different things at different times. Before the period starts, both of them want to know what basis they are on. The basis shapes what a business does with everything around the arrangement, so a business planning its year needs to know today whether the coming stretch of time is built on 6.00 per cent a year or something else. Waiting until the period is over to find out what basis it was on would leave both sides operating blind through the whole of it.
And the other half of the answer is arithmetic rather than preference. A rate on its own is not an amount. Turning 6.00 per cent a year into rupees takes the length of the period in days, counted by whichever method the document names, and until the period has actually elapsed its length is a plan rather than a fact. Holidays move. Business day conventions shift a date forward or back. The day count fractionThe slice of a year a period is treated as covering, once its days have been counted by whichever method the document names. Different methods count the same stretch of time differently. is settled only once the calendar has finished happening. So the sequence is forced: fix the basis at the front, count the days as they pass, and settle the amount at the back.
The electricity meter has the same two moments and nobody finds it confusing. The meter is read on a stated day and the reading is the basis. Until the month has finished nobody knows how many units ran through, so the bill arrives afterwards. A household that treated the meter reading day as the day the money left the account would be doing something strange. Nothing left the account. Somebody wrote a number on a card.
The reset for a period has just happened. Which is settled now, the rate or the amount?
Where does the number a reset produces actually come from?
Here is the part of a swap that is unlike every other part of it. Almost everything in the document was settled by the two sides talking to each other. The notional, the fixed rate, the dates, the counting method, the wording about what happens if something goes wrong: all of it came out of a negotiation between Chitrakoot Cements and Saranga Capital. The reading the floating leg takes did not. The reading comes from outside, compiled and published by an administratorThe body that gathers the inputs to a benchmark, works out the published figure and puts it out on a stated day. who has never met either of them and has no interest in which of them is better off.
Neither party produces the number and neither party can adjust it, and that is the entire reason for using a published rate instead of one the two of them agree between themselves each period. If Chitrakoot Cements could nudge the reading, the floating leg would be worthless as an obligation. If Saranga Capital could, the same. The arrangement works because both of them look at the same place on the same morning and find the same figure sitting there, whether they like it or not.
The agreement does not merely say which published rate. The agreement says which one, read on which day, at which time, and where it is to be found. All four are boring to negotiate and they are the difference between an arrangement that settles itself and one that produces an argument every period. Two sides looking at the same series but at different times of day can genuinely see different numbers, and each of them will be right about what they saw.
One habit follows from all four: a benchmark reading without its date is not a reading at all. The same published series gives one figure on Tuesday and a different figure on Wednesday. A number written into a note, an email or a spreadsheet without the day it was read on cannot be checked by anybody, including the person who wrote it down. Without the day, the morning to look at is gone and nobody can go back to the source to confirm what was seen. So the reading and the date travel together or the reading is not worth carrying.
A street vendor buying vegetables at a wholesale market knows this without being told. The price posted at the market this morning is the price. The posted price is not the vendor's price and not the buyer's price, and neither of them can talk it up or down. The whole point of the posted price is that it belongs to a particular day, so a vendor who reports the rate without saying which morning has reported almost nothing.
Somebody supplies a figure of 6.00 per cent a year for the floating benchmark. What should be asked before it is written down?
What is true in the gap between a reset date and the payment date it feeds?
The moment the reset has happened, the rate for the period is a settled fact. Both sides can now work out what the floating leg will come to, subject only to counting the days when the period has finished running. And nothing has been paid. The gap between the two dates can be a whole period long, and it is where the commonest error lives. Two different states of one number look like a single state to anybody moving at speed.
Two words hold the two states apart: the amount is knowable, and it is not due. Knowable means the arithmetic can be done today and nobody has to guess at anything. Due means the date on which the transfer happens has arrived. A reset moves an amount into the first state and does not touch the second. There is no partial payment, no accrual that leaves the account, and nothing that a bank statement would show.
A party does two things with a knowable amount, and they are different things. The party plans for the transfer. Knowing the basis, it can put that transfer in the right week of the right month with confidence rather than as an estimate. And the party does not book the cash. The cash has not moved and will not move until the payment date. Planning for a payment and recording a payment are separate acts, and the reset date licenses only the first of them.
An everyday version. A wedding caterer confirms the menu and the headcount six weeks before the event, and from that moment the bill is knowable to the rupee. Nobody in the household thinks the money has left. The confirmation removed the uncertainty and moved nothing. Six weeks later the payment happens, and it happens on the day it was always going to happen.
It is the morning after the reset. Rs 12.00 crore of net difference is now computable for the period. May a party book that cash today?
What does one reset actually settle, worked all the way through?
Take the first reset on the agreement between Chitrakoot Cements Limited and Saranga Capital Limited, and follow it through from the reading to the rupees. The opening reset date falls where the first period begins, and on that morning the floating benchmark shows 6.00 per cent a year. The reading is entered into the agreement as the floating rate for that period, and from that moment it is a fact both sides are bound by. Nothing is paid on the day it is entered.
Now watch what became computable the instant the reading landed. The notional is Rs 1,000 crore, or Rs 10,00,00,00,000/-. Apply 7.20 per cent a year, the rate written in at signing, across one whole period counted at a day count fraction of 1.0000, and the fixed leg lands on Rs 72,00,00,000/-, or Rs 72.00 crore. Apply the reading of 6.00 per cent a year over that same stretch on that same 1.0000 fraction, and the floating leg lands on Rs 60,00,00,000/-, or Rs 60.00 crore. Take the smaller away from the larger and Rs 12,00,00,000/- is left standing, or Rs 12.00 crore of net difference. Chitrakoot Cements sits on the larger of the two legs, so that difference travels to Saranga Capital.
A shorter route reaches the same rupees, and it shows what the reset really produced. The distance between 7.20 per cent a year and 6.00 per cent a year is 1.20 percentage pointsHow the gap between two rates gets measured. A point of separation works out at one in every hundred of whatever the two rates bite on.. Applied straight to the notional of Rs 1,000 crore, that distance gives Rs 12,00,00,000/- in one step, without either leg being computed separately. The two routes agree because the same Rs 1,000 crore sits under each leg and cancels straight out of the subtraction, so this is forced arithmetic rather than a check of one method against another. Taken back the other way, Rs 12.00 crore measured against a notional of Rs 1,000 crore is 1.2 per cent of it.
| Period one, after the reset | Rate | Amount |
|---|---|---|
| The fixed leg, settled at signing | 7.20 per cent a year | Rs 72,00,00,000/- |
| The floating leg, read on the reset date | 6.00 per cent a year | Rs 60,00,00,000/- |
| The net difference, Chitrakoot Cements to Saranga Capital | 1.20 percentage points | Rs 12,00,00,000/- |
Three of those four figures were unknowable the day before the reset. The fixed leg was never in question. All of them were settled the moment after the reading landed. A reset does that much and no more, and it is worth noticing how small the act was that did it. Somebody read a number and wrote it down with a date.
Now the same table for the second period, with the same headings and the same row labels.
| Period two, before any reset | Rate | Amount |
|---|---|---|
| The fixed leg, settled at signing | 7.20 per cent a year | Rs 72,00,00,000/- |
| The floating leg, read on the reset date | no reading taken | no reading taken |
| The net difference, direction not yet known | no reading taken | no reading taken |
The second table is empty in three of its six cells because the reset for that period has not happened, nobody has read the floating benchmark on that day, and no reading for it exists anywhere here. Notice which cell is still filled. The fixed leg for period two is Rs 72,00,00,000/- and always was. A rate of 7.20 per cent a year never needed a reading. The emptiness is on one side only, and it is on the side that has to go out and look something up.
Does a reset change the notional, or anything else in the agreement?
Almost nothing, and the shortest way to see it is to look at what a reset produces. A reset produces a percentage. That is all. A percentage is not a sum of money and it does not become one until somebody applies it to something and counts some days.
A reset changes one number in one leg for one period, and everything else in the document is exactly where it was the moment before. The list of what a reset leaves untouched is long, and every item on it is something a reader might reasonably have expected to move. Work through it. The notional of Rs 1,000 crore is unchanged, in this period and in every other period the arrangement will ever have. The fixed rate, 7.20 per cent a year, is unchanged. The counting method is unchanged. The calendar of reset dates and payment dates is unchanged. The two parties are the same two parties. The end date of the arrangement has not moved by a day.
The notional is the item most often misread, so it deserves a sentence of its own. A reset decides which percentage gets applied to Rs 1,000 crore for the coming period. The reset does not touch the Rs 1,000 crore. The notional was written into the agreement to give the rates something to bite on, and it stays exactly where it was written whether the reading comes in high, low or unchanged. Nothing about a reset makes the arrangement bigger or smaller.
A reset is not a renegotiation, and expecting it to reopen the agreement mistakes one for the other. There is a lookup. Two sides agreed in advance to be bound by whatever a particular published figure says on a particular morning, and the reset date is simply the morning arriving. Nobody is at a table. Nobody has a position. Somebody reads and records.
How many resets does an arrangement have, and what does the schedule look like?
One per period, running in a line for as long as the arrangement does. Each period begins with a reset and ends with a payment, and the next period begins with the next reset. Written out that way it sounds trivial, and drawn out in full it teaches something a sentence does not.
Draw the whole line at once and the picture is a row of identical boxes on the fixed leg and a row of boxes on the floating leg of which exactly one is filled. The empty boxes are the honest shape of a floating leg at any moment in its life: one period settled, and a queue of dates that are already agreed with nothing in them.
Which gives a useful measure for anybody trying to understand how much of an arrangement is still undecided. Count the resets ahead. The count of resets ahead is the number of times somebody will have to go and read something before the arrangement is finished, and it is the number of separate figures still outside anybody's knowledge. The count falls by exactly one each period, whatever the readings turn out to be, and nothing else about a floating leg behaves that predictably. Uncertainty of this kind is measured in unread dates rather than in the level of anything, so a high reading and a low reading remove the same amount of it.
The timeline below is about to stretch from one period to eight. What should the floating leg row do?
Stretch the timeline and watch which row fills
One control: how many periods of the arrangement are drawn. Held constant throughout: a notional of Rs 1,000 crore, a fixed leg set at 7.20 per cent a year, the single reading of 6.00 per cent a year taken on the opening reset date, and every period counted at 1.0000. The emphasis selector changes what is highlighted and touches no arithmetic at all.
Emphasis:
Four periods are drawn.
The default setting reproduces the worked instance above exactly. At four periods drawn, period one reads a fixed leg of Rs 72.00 crore, a floating leg of Rs 60.00 crore and Rs 12.00 crore of net difference against a notional of Rs 1,000 crore, handed by Chitrakoot Cements Limited to Saranga Capital Limited. The three periods after it read fixed legs of Rs 72.00 crore with empty floating boxes and empty net boxes, and three resets are still ahead. Drag the control down to a single period and the two rows look identical. Widening the range takes that impression away.
Why is only one reset in this record filled in?
A reset takes its value from a published reading on a date that has not arrived, and nobody has published those readings, so none of them is guessed at. The alternative is seven further made up numbers that would look more complete and settle nothing.
The missing readings are the part that matters. Notice that the empty boxes are drawn rather than the timeline stopping after the first one. The dates are already fixed and only the readings are missing. Those are two very different kinds of not knowing, and merging them makes the arrangement look far vaguer than it actually is.
Separate the two. The dates are settled. Every reset date and every payment date, from the opening period through to the closing one, went into the document at signing, and nobody has to wait to find out when they fall. The calendar is completely determined. Nobody knows what a published series will say on each of those mornings. Stop the timeline at period one and a reader sees an arrangement that fades into fog after the first period. Draw the empty boxes and the same reader sees an arrangement with a fully known skeleton and one unknown number hanging on each rib of it. The second picture is the true one and it is much less alarming.
Four parts fill one of those empty boxes, and the list is worth being able to recite: a reading of the named benchmark, taken on that period's reset date, published by its administrator, with the date attached to it. All four parts, and nothing less. A figure that arrives without a date is not a reading. A figure taken on the wrong day is a reading of something else. A figure somebody estimated from where things seem to be heading is not a reading at all, and it is the most dangerous of the three. Once it is typed into a box it looks exactly like the real thing, and nobody downstream can tell them apart.
What exactly would fill one empty box on the floating leg row?
What goes wrong at a reset, and what does the agreement provide for?
Four things, and the agreement has an answer to each of them because each of them has happened to somebody. Very little of the list is about the level of the rate.
The argument at a reset is almost never about the number itself. The argument is about the day, the source, or what happens when there is nothing to read. Nobody disputes that a published figure said what it said. The dispute is whether that was the figure the document pointed at.
The first failure is a reading taken from the wrong day, usually a day early or a day late, and it is the commonest of the four by a distance. Two sides look at the same published series, take it on adjacent mornings, and end up disagreeing about the rate for an entire period. Nobody is being difficult and nobody has been careless in any way that would show up at the time. The document names the day, and the day is the only thing that settles it.
The second is that the reading is not published on the reset date at all. Something interrupts the publication and there is simply nothing to look at. The document has to say what gets used instead, and that wording is called the fallbackThe wording sitting inside an agreement that says which figure is used when the one it normally reads is not there.. The fallback was negotiated rather than standardised, so it differs between documents. The answer sits in the reader's own paperwork rather than in any general account of the mechanism.
The third is that the reading stops being published permanently. A series can be retired, and when one is, every agreement pointing at it has to know what replaces it. A document written without that clause leaves a floating leg pointing at nothing, and two parties with no agreed way of deciding which rate the floating leg takes next.
The fourth has no clause behind it: the reading was taken correctly and entered into a system incorrectly. The right figure, the right day, the wrong keystroke. A keystroke is the only one of the four that no wording prevents, and the only defence against it is two people looking at the same figure against the same source.
All four answers live in the agreement in front of the reader rather than in any general account, and somebody who has never looked up the fallback wording in their own document does not know what their floating leg does in the one case where it matters most. The fallback wording is short, it usually sits near the definitions, and reading it once costs a few minutes on a quiet afternoon rather than an argument on a morning when something has gone wrong.
The floating benchmark is not published on a reset date. Where does the answer to what happens next sit?
The error that gets made, and what it costs
Somebody sees that the rate for the period has been fixed and treats that as the event. The expected transfer goes into the cash forecast on the reset date rather than on the payment date. The reset genuinely is the moment the uncertainty disappears, so it feels like the moment something happened, and the mistake is an easy one to make.
Watch what it does to the forecast. Every expected movement now sits one full period early. In a steady arrangement the errors overlap each other, one period's misplaced entry landing where the previous period's belonged, so the total for the year comes out right and nothing looks wrong on any summary anybody reads. The error survives for years precisely because it reconciles annually, and it becomes visible only in the one week where it matters: the week the forecast said money was moving and none did, or the week it said nothing was moving and Rs 12,00,00,000/- of net difference left the account.
The cost is a liquidity bufferCash held back deliberately so that a known outflow can be met on the day it lands, rather than raised at short notice. sized for the wrong week, held by a team with every reason to believe its forecast is accurate. The annual total has reconciled every year since the arrangement was signed. Nobody involved has been careless. The forecast is wrong in its timing and right in its total, and the total is the thing everybody checks.
How does a treasury team actually use a reset calendar?
Everything above is mechanism. The work somebody does with the mechanism on a Monday is short and unglamorous.
A borrower's treasury team keeps a reset calendar. The calendar is nothing more than a list of dates with a name against each. On the morning of a reset, one person reads the series the document names, at the time the document names, and records the figure with its date beside it. A second person reads the same source and confirms the same figure. The second reading is the whole of the control, and it exists because the fourth failure above is a typing error that no clause can prevent. Two pairs of eyes on one number, once a period.
The step after the confirmation is where a reset calendar earns its keep. The moment the reading is confirmed, the net difference for the period becomes computable, so the cash forecast entry for that period can move from an estimate to a known figure. The entry goes against the payment date, not against today. A team that gets this right ends up with a forecast where the near periods are hard numbers and the far ones are marked as depending on readings not yet taken, and the boundary between the two is exactly one reset ahead of today.
An analyst on the outside cannot see the calendar, so the same structure reads differently. The analyst can see how much of an obligation is settled and how much is still waiting on a lookup. A floating leg with many resets ahead of it is not riskier in any judgement sense, it is simply less determined, and the honest way to describe it is by the count of unread dates rather than by a guess about where those readings will land. A counterpartyWhoever stands opposite a party in a privately agreed arrangement, carrying obligations that mirror its own. credit team asks a related question: on the reset dates ahead, does the other side have the operational capacity to read, confirm and record on the day, in every period, from now until the arrangement finishes. The capacity to read, confirm and record is an operations question rather than a market one, and it is asked far less often than it deserves to be.
For a household the equivalent is small and the discipline is identical. Somebody on a floating rate home loan has a reset date on it, usually written into the sanction letter and usually ignored. Knowing the date means knowing the one morning in the period when the number that governs the next several months gets decided, and it means the change in the instalment arrives as something expected rather than as a surprise in a bank message.
Who decides which benchmark readings may be referenced at all?
Set by an authority, and changed by it
| What is set by an authority | Where to confirm it |
|---|---|
| Which benchmark readings may be referenced in an arrangement, and who administers them | Reserve Bank of India, rbi.org.in |
| What has to be published about a benchmark reading, and by whom | Reserve Bank of India, rbi.org.in |
Whatever belongs in these rows is decided by the authority printed inside the row. The decision has been changed before and will be changed again, and a copy of it made anywhere else would not be merely stale on the morning it moved. A stale copy would be wrong, and wrong in a way a reader trusting it could not detect.
Which published series a market uses is set and replaced by the body printed in the rows above. Series have been retired before and others will be. A reader given a name walks away holding a name instead of a mechanism, and keeps holding it long after the name has been withdrawn. A reader holding the mechanism treats the name as a detail to be looked up in the relevant document, where it was always going to be anyway.
What does a reset date not tell anybody?
Knowing the day on which a floating leg takes its number is a reading skill. On its own it is not a reason to be inside an arrangement of this kind, and whether one suits a particular party is not settled here. The question stays unanswerable until somebody has assembled four things about that party in particular: what it already owes and on which basis, over what stretch of time those obligations run, how much of the cash already coming in moves when a published rate moves, and what it would actually do in a period where the reading landed against it. Those four things sit with the party and nowhere else. An answer offered without them would be an answer about a position nobody has looked at.
The same holds for the arrangement itself. A worth for either side would have to be built out of a list of what the benchmark reads on the dates still ahead, and nobody can assemble that list before the readings exist. The mechanism stands on its own, and the empty spaces sit where the missing readings would go.
Name everything in the agreement that a reset changes.
Six neighbouring questions, and where each of them is answered.
The payment date on which the money actually moves is covered separately. Turning the days inside a period into a slice of a year, and why one counting method pays out differently from another on the same rate, is covered separately. The swap curve, and why one rate at one maturity is not a curve, is covered separately. How collateral moves between dates is covered separately. Which published series is used in any market, and how it is administered, is set by the Reserve Bank of India at rbi.org.in and changes.
The sixth is different in kind. The worth of an arrangement at any moment between two resets rests on a schedule of expected benchmark readings, and no such schedule exists until the readings themselves do. A worth produced without one would be a fabrication a reader could not detect.
Where the named authorities publish
| Source | What it covers here | Site |
|---|---|---|
| Reserve Bank of India | Which benchmark readings may be referenced in a rate arrangement, and who administers them. | rbi.org.in |
| Reserve Bank of India | What has to be published about a benchmark reading, and by whom. | rbi.org.in |
| Securities and Exchange Board of India (SEBI) | Exchange traded contracts, and what may be published as research. | sebi.gov.in |
| Bank for International Settlements | Where cross border statistics on privately agreed arrangements are published. | bis.org |
Chitrakoot Cements Limited, Saranga Capital Limited, the floating benchmark they are written against and the agreement between them are invented.
Educational material. Not advice on any investment, tax, budget or market position.
