What a swap is

The line that appears without a trade 

Nothing about the position changed on those nights. No order was sent, no size was altered, and the opening price recorded against the contract is the same number it was at the start. What changed is the date the contract is deemed to settle on. A spot transaction agreed today is due to settle two business days later, and that date is part of what was agreed. A position still open at the end of the trading day would otherwise fall due on a date that has already arrived, so the market rolls it forward: the settlement date moves on, and the interest owed across the day it moved by is settled in cash. The swap is that cash amount. It is the reason a flat position can still move a balance.

The name is inherited rather than descriptive. In the interbank market a foreign exchange swap is a genuine pair of transactions, one agreed now and one agreed back at a later date, and the price difference between the two legs is the interest carried across the gap between them. The retail line does the same job without the machinery: one adjustment, applied once a night, in place of two transactions neither party wants for its own sake. Firms label the entry variously as swap, rollover, overnight financing or an overnight adjustment. Those words describe a single thing.

Why there is anything to settle 

A currency pair is two interest rates carried inside one price. A position in a pair holds one of the two currencies and owes the other, so across a day it generates an amount receivable on the currency held and an amount payable on the currency owed. The adjustment is the difference between those two amounts, modified by the counterparty's markup. That is the whole origin of the line for a currency pair, and it explains a property that surprises readers who meet it on a statement first: the size of the adjustment has nothing to do with whether the position is winning or losing, and everything to do with two central banks that have never heard of the position.

For a contract written on something other than a currency pair the origin is different and the shape is the same. The full notional value of a CFD is never funded, so the portion that was not posted is financed for as long as the contract stays open, conventionally at an interbank reference rate for the currency the contract is denominated in, adjusted by the counterparty. Contracts on shares and share indices carry further adjustments of their own on the days their underlying markets distribute. How the calculation differs between instrument families is a subject in its own right, and the arithmetic below is the currency pair case, which is the one every other case is described against.

Two properties follow from that origin, and both of them separate swap from every other cost taught in this module. It accrues per night rather than per trade, so it is indifferent to how often a position is opened and closed and sensitive only to how long one stays open. And it is recorded as its own line rather than folded into the position's profit and loss, which is why it survives a position closed at exactly its opening price. Spread and commission are paid at the moment a position is opened or closed. Financing is paid by the calendar, on a schedule the position has no influence over.

How the rate is quoted 

The rate is published per instrument in its contract specifications, alongside its size convention and its trading hours, and it is published as two numbers rather than one, because direction determines the sign. Conventions for expressing it differ between firms and between instrument families: points per lot per night, a money amount per lot per night, or an annualised percentage applied to the notional value. The three forms are convertible into one another and none is more authoritative than the others. The specifications for the instruments a CFD can be written on at YAL sit across the markets pages.

A published rate is a current statement rather than a fixed term. Four things sit behind the number, and any of them can move it:

• The two policy rates behind a currency pair, or the reference rate for the currency a contract is denominated in. They move when the central banks setting them move them, and the financing line moves with them.
• The direction held, which is why the specification carries two numbers rather than one.
• The counterparty's markup, applied to each direction separately rather than shared between them.
• The days of interest the settlement calendar assigns to the night in question, which is not always a single day.

Why the number has two signs 

In principle the direction holding the higher yielding currency is credited and the direction holding the lower yielding one is debited, because that is what the difference between two interest rates produces. The qualifier is doing real work. The markup is applied to each direction independently rather than split between them, so the credited amount is the smaller of the two figures for the same pair on the same night, and when two policy rates sit close together the arithmetic commonly produces a debit on both directions at once. A financing line read as though the two sides were mirror images returns the wrong number, and it returns it in the direction that flatters the calculation.

The night that counts three days 

Settlement runs on business days, and a weekend contains none. A position held past the cut off in the middle of the week rolls its settlement date across the weekend in one move, three calendar days rather than one, so three days of interest are applied on that single night. Nothing about the market changes that evening. The calendar changes, and the charge follows the calendar rather than the trading. Across a full week the effect is arithmetic rather than punitive: seven calendar days of interest are applied across the five nights the market is open, which is exactly what the triple night exists to accomplish.

The night it lands on is a convention rather than a rule of the market, and it is not universal. Currency pairs conventionally take it on Wednesday, because a Wednesday transaction settling two business days later settles on the Friday, and the roll moves that to the Monday. Instruments settling on an underlying exchange's calendar commonly take it on Friday instead. A national holiday in either currency of a pair can move the day, or add a further day to the count on a night that would otherwise have carried one. The convention is published per instrument, firms differ, and the specification is the only reliable statement of which night applies to which contract.

Counting the nights 

The calculation is the rate multiplied by the days of interest applied, and the only part of it that takes any care is the count. What is counted is not days held. It is cut offs crossed, with the triple night counted as three. A position opened and closed inside one trading day crosses none of them and carries no financing at all, which is the whole difference between a position closed before the cut off and the same position closed a minute after it.

Two things in that block are worth reading twice. The days of interest exceed the nights crossed, so a count made by counting entries on a statement understates the charge unless the triple night is recognised for what it is. And the two totals are not mirror images, for the reason set out above: the markup is applied to each direction separately, so the same days produce a much larger debit on one side than the credit they produce on the other. Financing is the one cost in this module whose size is set almost entirely outside the transaction, and the only term of it a trader sets is the number of nights.

Where practitioners disagree 

The first disagreement is about the credit. A long standing tradition in currency markets, usually called the carry trade, is built around holding the higher yielding side of a pair and receiving the differential across time. It carries its own well documented limit: the differential that produces the credit is one of the things that moves the price of the pair, so the two quantities are not independent, and a repricing driven by a change in rate expectations can move the price by more in a session than the differential accrues across a long sequence of nights. The markup narrows the credited side further. Practitioners who use the framework and practitioners who reject it are reading the same arithmetic and weighting its two terms differently, which is why the argument has never resolved.

The second concerns the debit, and it is the more practical of the two. One tradition treats financing as a running cost that accrues every night regardless of what the market does, and reads it as a reason to think of holding periods in nights. Another observes that closing a position before the cut off and reopening it afterwards crosses the spread twice, in the window where the spread is commonly at its widest, and that the two crossings can exceed several nights of financing. Which of the two amounts is larger depends on the instrument, on the rate in force at the time and on the width of the spread at the moment of re entry, so neither position generalises into a rule, and this page does not put one forward.