Netting

8.0 What this chapter gives you#

1. You will be able to explain how about £23 billion of daily Bacs payments is discharged by moving about £5.3 billion between settlement accounts, with nobody short-changed.
2. You will be able to compute bilateral and multilateral net positions from a matrix of flows, and say why multilateral almost always wins.
3. You will be able to explain why the sum of all net debit positions must equal the sum of all net credit positions.
4. You will be able to say why Faster Payments nets harder than Bacs, and connect the answer to the direction of the traffic rather than to the technology.
5. You will be able to explain why netting does not reduce risk but concentrates it, turning diversified credit risk into liquidity risk at a single moment.
6. You will be able to work through an unwind and show why a bank with no meaningful exposure to the failed participant can suddenly owe five times more.
7. You will be able to distinguish position netting, netting by novation and close-out netting, and say which of them survives an administrator’s cherry-picking.
8. You will be able to state the Lamfalussy “cover one” standard and name the controls that implement it: caps, prefunding and loss sharing.
9. You will be able to describe what a central counterparty does through novation, and the order in which its default waterfall is consumed.
10. You will be able to give the real reasons Bacs takes three days, and explain why the debit and the credit both fall on day three.

On an average working day in 2024, the Bacs system carried about 26.8 million payments worth roughly £23.0 billion between British banks. At the end of that day, the amount of money that actually moved between those banks in the Bank of England’s settlement accounts averaged £5,265 million.

Nobody was short-changed. Every salary was paid in full, every direct debit was collected in full, every recipient saw the exact amount they were owed. Twenty-three billion pounds of value was transferred and about five billion pounds of money moved.

That is netting. It is arithmetic so simple that a child can do it, and its consequences run through every payment system on earth. It is the reason interbank settlement is affordable at all. It is also the reason a payment system can fail in a way that a pile of individual payments cannot, and the reason central banks spent the 1980s and 1990s writing rules about it.

The previous chapter separated clearing from settlement. This chapter is about what happens in the gap: the operation that turns a very large number of small obligations into a very small number of large ones.

The plain version#

Four friends share a house: Ama, Ben, Cara and Dev. Over the course of a week they keep buying things for each other, because that is what people who live together do. Somebody gets to the front of the queue first and pays for two coffees. Somebody covers a train fare. Somebody buys the bin bags.

By Sunday night the situation looks like this.

Twelve separate debts. If everybody paid everybody exactly what they owed, twelve payments would be made and £102 in cash would change hands.

Nobody does that. It would take all evening and everybody would need to have the right cash on them.

The first improvement: settle up in pairs#

The obvious short cut is to deal with one pair at a time. Ama owes Ben £12 and Ben owes Ama £9, so instead of two payments totalling £21, Ama hands Ben £3 and both debts are gone.

Do this for every pair and you get six payments instead of twelve.

Twelve payments have become six, and £102 of cash movement has become £44. Every debt has been discharged in full. Nobody has been given a discount and nobody has taken a haircut. The debts cancelled each other out because they pointed in opposite directions.

This is called bilateral netting. Bilateral means two-sided: you look at one pair of people at a time.

The second improvement: one pot#

There is a better way, and it takes about a minute with a pencil.

Instead of pairing people up, work out for each person the total they owe everybody and the total everybody owes them. Then take the difference.

Ama puts £7 on the kitchen table. Ben puts £3 on the table. Cara puts £8 on the table. There is now £18 on the table, and Dev picks it up.

Four movements of cash instead of twelve. Eighteen pounds instead of one hundred and two. And, once again, every single one of those twelve original debts has been paid in full. Nobody is out of pocket by a penny more or less than they should be.

Notice that the amounts owed and the amounts owing balance exactly. Ama, Ben and Cara owe £7, £3 and £8, which comes to £18, and Dev is owed £18. That is not a coincidence and it is not luck. Every debt has two ends. If you add up everything everybody owes and subtract everything everybody is owed, the answer must be zero, because you have counted each debt once as a plus and once as a minus. So the money going into the pot always equals the money coming out of it.

This is called multilateral netting. Multilateral means many-sided: everybody at once.

Now do it with banks#

Replace the four friends with four banks, and replace the coffees and train fares with customer payments.

When a Barclays customer pays a Lloyds customer £180, that is not really a payment from a person to a person. Underneath, it is Barclays owing Lloyds £180. The customer’s balance goes down at Barclays, the recipient’s balance goes up at Lloyds, and now Barclays is £180 short with Lloyds and has to make it good.

There are about 27 million of those a day on Bacs alone, and they run in every direction at once, because Barclays customers pay Lloyds customers all day and Lloyds customers pay Barclays customers all day. The great majority of the value cancels out, exactly the way Ama’s £12 and Ben’s £9 cancelled out.

So the banks do precisely what the flatmates did. They add everything up, work out one number for each bank, and move that. Twenty-seven million payments become a few dozen movements in the Bank of England’s books.

Why this is such a spectacular deal#

Money sitting in a bank’s settlement account at the Bank of England is money doing nothing else. It is not lending and it is not investing; it is sitting there so that payments can be made out of it. A bank that had to fund every outgoing payment as it happened would need a very large pile of cash standing still all day.

Netting means it needs a very much smaller pile. In the flatmates example the cash needed fell from £102 to £18. If you had asked all four to have the full £102 available on Sunday evening, somebody would have said no. At £18 nobody even thinks about it. Multiply that by a national payment system and netting stops being convenient and becomes the difference between a payment system a country can afford to run and one it cannot.

And why it is dangerous#

Here is the catch, and it is worth understanding before we go any further.

Look at Dev. He is owed £18 and he is expecting to be handed £18 on Sunday night. But he is not being handed £18 by any one person. He is being handed £18 by an arrangement. If Cara turns out to have nothing in her wallet at the crucial moment, the pot has £10 in it, not £18, and Dev is short.

Worse: if Cara’s debts have to be taken out of the calculation entirely, everybody’s number changes. It is not just that Cara’s £8 goes missing. Ama, who owed £7, might now owe more, or less. Ben, who owed £3, might turn out to be owed money. Nobody can predict which way their position moves until the sum is redone.

That is the trade the whole payments industry has made. Netting makes settlement cheap by making everybody depend on one calculation. When the calculation works, it is enormously efficient. When it does not, the failure lands on everybody at once rather than on the person who was unlucky enough to be owed money by the party that failed.

And why waiting is useful#

One last thing before we take the analogy apart.

The longer you wait before squaring up, the more cancels out. If the flatmates settled every debt the moment it arose, nothing would ever net, because there would never be two debts alive at the same time to cancel against each other. If they wait a day, some of it cancels. If they wait a week, most of it does.

This is not a small effect. It is the reason a payment system that waits three days can move a fifth of the money that a payment system settling instantly would have to move.

That is the thread that runs to the end of this chapter. When you eventually ask why a Bacs payment takes three days, part of the answer is: because three days is how long you have to wait for enough of it to cancel out.

Where the plain version stops being true#

The kitchen table is the right picture, and it is very close to what actually happens. Four things about it are wrong in ways that matter.

Nobody is paid less, and the netting is invisible to customers#

This is the misunderstanding that most often survives a first explanation, and it is worth killing immediately.

Netting operates on the obligations between banks. It does not touch the payments between customers. If your employer pays you £2,400 and your landlord takes £1,100 by direct debit on the same morning, you are credited £2,400 and debited £1,100. You are not credited £1,300. Your bank’s ledger records both movements in full, because your bank’s ledger is a record of what your bank owes you, and it owes you both things separately.

What is netted is the layer underneath: the amount your bank owes the other banks. The customer layer is gross. The interbank layer is net. They are different ledgers with different arithmetic, and the whole efficiency of the system comes from the fact that one can be compressed without touching the other.

The same is true within a bank. If a Lloyds customer pays another Lloyds customer, that payment never reaches an interbank system at all. It is settled by moving a number from one row of Lloyds’ own ledger to another. The Bank of England’s published payment figures explicitly exclude payments internalised within a single institution, for exactly this reason.

Netting is a legal act, not just arithmetic#

Adding up is easy. Making the answer stick when somebody goes bankrupt is not, and this is where most of the industry’s difficulty actually lies.

Suppose Cara is declared insolvent on Sunday afternoon. Her administrator’s job is to maximise the return to her creditors. A sufficiently aggressive administrator will look at the twelve debts and say: the £20 that Cara owes Dev is a claim in the insolvency, to be paid at whatever pence in the pound the estate can manage; the £3 that Dev owes Cara is an asset of the estate and must be paid in full, today. That is called cherry-picking, and it converts a tidy £17 net obligation into a £20 unsecured claim and a £3 immediate demand.

If that were allowed, netting would provide no protection whatever, and every bank in the country would have to hold capital and liquidity against its gross exposures rather than its net ones. So it is not allowed, but only because specific law says so. In the United Kingdom the relevant instrument is the Financial Markets and Insolvency (Settlement Finality) Regulations 1999, SI 1999/2979, which implemented Directive 98/26/EC. Systems designated under those regulations get statutory protection: transfer orders entered into a designated system, and the netting of those orders, take precedence over ordinary insolvency law and cannot be unwound by an administrator.

Designation is not automatic and not universal. There is a real difference between netting that is legally robust and netting that is merely a spreadsheet, and practitioners care about which one they are looking at.

There is usually no pot#

The kitchen table suggests a fund sitting in the middle holding money. In most net settlement systems there is no such thing.

The system operator calculates the net positions and sends the Bank of England a set of instructions. The Bank debits the accounts of the banks in net debit and credits the accounts of the banks in net credit, in a single simultaneous posting. Nothing is held in the middle at any point. The “pot” is a bookkeeping convenience, not a place.

There is an important exception, and it is exactly where the risk is. Some systems do require participants to put money aside in advance, in a segregated account at the Bank of England, so that there is something to fall back on if a participant cannot pay. That is called prefunding, and the Bank of England introduced it for both Bacs and Faster Payments in September 2015. Where prefunding applies there genuinely is a pot, and it exists because the arithmetic on its own was judged not safe enough.

Netting does not reduce risk; it concentrates it#

The flatmates example makes netting look like pure gain. It is not.

Before netting, Dev has three separate exposures: to Ama, to Ben and to Cara. If Cara fails, Dev loses what Cara owed him and keeps what the other two owed him. The damage is proportionate and contained.

After netting, Dev has one exposure: to the arrangement. If any participant fails, the amount Dev receives changes, and by an amount that has nothing to do with his direct dealings with the party that failed. He has traded a set of small, understandable, diversified exposures for a single large exposure to a mechanism.

That is not a criticism of netting. It is the deal. Netting converts credit risk that was spread across many counterparties into liquidity risk concentrated in one moment, and the entire architecture of modern payment system regulation exists to manage that one moment. The technical version works through exactly what happens when it goes wrong.

The technical version#

Vocabulary#

The Committee on Payments and Market Infrastructures, the standard-setting body at the Bank for International Settlements, defines clearing as “the process of transmitting, reconciling and, in some cases, confirming transactions prior to settlement, potentially including the netting of transactions and the establishment of final positions for settlement”. Netting is therefore a sub-process of clearing, not a synonym for it, and the qualifiers matter: “in some cases”, “potentially including”. A system can clear without netting.

The working definitions used throughout the rest of this book are as follows.

Gross settlement means each payment obligation is settled individually, for its full amount. Net settlement means obligations are offset against each other and only the resulting balance is settled.

Bilateral netting offsets obligations between one pair of participants. Each pair produces one net figure and one settlement movement.

Multilateral netting offsets each participant’s obligations against all other participants simultaneously. Each participant produces one net figure: a net debit position if it owes the system, a net credit position if the system owes it. The sum of all net debit positions equals the sum of all net credit positions, always, by construction.

Deferred net settlement (DNS) means the netted figure is settled at a scheduled time after the underlying payments have been exchanged. Real-time gross settlement (RTGS) means each payment is settled individually as it arises. The Bank of England currently supports four settlement models in its RTGS infrastructure: real-time gross settlement, delivery versus payment, prefunded deferred net settlement and unfunded deferred net settlement. Only CHAPS uses the first. The compression achieved is expressed as a netting ratio: the netted value as a proportion of the gross value, or its complement, the percentage of value eliminated.

Four banks, worked in full#

Take four settlement participants and one clearing cycle. The matrix below shows the total value of payment instructions sent by each bank to each other bank during the cycle, in millions of pounds. Rows are payers, columns are payees.

Twelve directional flows. Aggregate gross value: £3,090 million.

If this system settled gross, twelve settlement movements totalling £3,090 million would be required, and each bank would need enough liquidity in its settlement account to cover its full outgoing total: £860 million for Northbank, £840 million each for Eastbank and Southbank, £550 million for Westbank. Aggregate liquidity requirement: £3,090 million.

Bilateral netting#

Offset each pair.

Six settlement movements totalling £290 million. That is 9.4 per cent of the gross value: bilateral netting has eliminated 90.6 per cent of the value that would otherwise have had to move. Note the liquidity consequence. Northbank must fund £40 million to Eastbank and £100 million to Westbank, £140 million in all, while separately receiving £30 million from Southbank, and it cannot use the incoming £30 million to fund the outgoing £140 million, because bilateral netting produces six independent settlements rather than one.

Multilateral netting#

Now offset everything against everything.

Northbank, Eastbank and Southbank are debited £110 million, £10 million and £80 million respectively. Westbank is credited £200 million. Debits sum to £200 million; credits sum to £200 million.

Four postings to settlement accounts, moving £200 million. That is 6.5 per cent of the gross value: multilateral netting has eliminated 93.5 per cent.

The three side by side#

Multilateral netting beats bilateral netting here, as it almost always does, because it allows a claim on one participant to fund an obligation to a different participant. Northbank’s £30 million claim on Southbank reduces its aggregate obligation from £140 million to £110 million, which bilateral netting cannot achieve. The efficiency grows with the number of participants and with how balanced the flows are, and is weakest when one participant is a persistent net sender to everybody.

What netting buys, in real numbers#

The four-bank example is illustrative. The following are not.

Bacs and Faster Payments#

Pay.UK, which operates Bacs, the Faster Payment System and the Image Clearing System, reported that in 2024 Bacs processed 6,811 million payments worth £5,838,567 million, an average of 26.8 million payments and £22,986 million per working day. In 2025 the figures rose to 6,864 million payments worth £6,049,248 million, an average of 27.1 million payments and £23,910 million a day.

The Bank of England, which settles those obligations, reported that daily average Bacs net settlement values in RTGS in 2024 were £5,265 million.

So roughly £23 billion of customer payment value per working day was discharged by moving roughly £5.3 billion between settlement accounts. Netting eliminated of the order of three quarters of the value. The two figures come from different publishers and may be computed over slightly different numbers of days, so the ratio should be read as approximately 77 per cent eliminated rather than to three significant figures. The order of magnitude is not in doubt.

The compression in the number of movements is more dramatic still. At the end of 2025 Bacs had 33 direct customers, over 330 indirect customers and around 117,000 service users submitting payment files. Twenty-seven million payment instructions a day, originating from a hundred and seventeen thousand organisations, resolve to at most thirty-three postings in the Bank of England’s ledger, once per business day.

Faster Payments compresses harder. Pay.UK reported total Faster Payments value of £4,242,380 million in 2024, while the Bank of England reported daily average Faster Payments net settlement values of £1,929 million. Faster Payments settles three times each business day, so the annual settled value is of the order of £490 billion against gross value of £4,242 billion. Netting is eliminating something in the region of 88 per cent.

The reason Faster Payments nets harder than Bacs is instructive. Bacs traffic is dominated by direct credits — payroll and benefits — which flow overwhelmingly in one direction on particular days of the month. Faster Payments traffic is person-to-person and person-to-business, running in both directions continuously. Balanced flows net well. Directional flows do not.

For comparison, the Bank of England’s 2024 daily averages across the whole of RTGS were:

CHAPS settles gross and therefore appears at its full value. Every other retail system in the table has already been compressed before it reaches the Bank of England. That single table is the clearest statement in print of what netting does to a national payment system: seven retail systems carrying the overwhelming majority of the country’s payment volume account for about 1.5 per cent of the value settled.

Securities: the DTCC#

In the United States, the National Securities Clearing Corporation, part of the Depository Trust and Clearing Corporation, clears the great majority of US equity trades. DTCC states that NSCC netting reduces the value of payments that need to be exchanged by an average of 98 to 99 per cent every day.

That is a higher ratio than any payment system achieves, and the reason is structural: securities trading involves the same instruments changing hands many times in a day between the same set of intermediaries, so the offsetting is extraordinarily dense. NSCC’s continuous net settlement service reduces each member’s obligations in each security to a single net long or short position. For scale, DTCC reported that on 9 April 2025 NSCC reached a peak value of $5.55 trillion, against a previous peak of $5.22 trillion on 20 December 2024.

Foreign exchange: CLS#

CLS Bank settles foreign exchange on a payment-versus-payment basis in eighteen currencies, holding accounts with all eighteen central banks. It settles the gross value of each instruction, but funds the settlement on a multilaterally netted basis: each settlement member transfers only the net amount of its combined obligations in each currency.

CLS reports that this, combined with its in/out swap liquidity tool, produces an average funding requirement of less than 1 per cent of the total value of all trades for participating settlement members. In the first half of 2022 CLS settled an average of over $6.5 trillion every day, with over 70 settlement members and over 30,000 third-party participants.

Less than one per cent. That is the number to remember when somebody claims that netting is a marginal optimisation.

The 2025 BIS Triennial Central Bank Survey put a broader figure on the same phenomenon. More than $14 trillion of gross foreign exchange obligations were settled on an average day in April 2025. More than $2 trillion of that, about 15 per cent, was subject to pre-settlement netting, and that $2.2 trillion was reduced to $337 billion of actual settlement — a reduction of roughly 85 per cent on the netted portion alone.

The legal forms of netting#

The distinctions below are not academic. They determine whether a netting arrangement survives an insolvency, and therefore whether a bank may recognise net rather than gross exposures for regulatory capital purposes.

Position netting, sometimes called payment netting or settlement netting, computes a net figure for settlement purposes but does not legally discharge the underlying obligations. The gross obligations continue to exist until they are settled. If a participant fails before settlement, the net calculation has no independent legal force and the gross obligations are what an administrator sees. This is the form that gives rise to unwind risk.

Netting by novation replaces each new obligation between a pair of parties with a single new obligation that discharges and supersedes the previous one. At any moment there is exactly one obligation between the pair, of the net amount. Because the gross obligations have been legally extinguished as they arose, there is nothing for an administrator to cherry-pick.

Close-out netting operates on default rather than on schedule. On a defined termination event, all outstanding contracts between the parties are terminated, valued at market, and reduced to a single net sum payable one way. This is the mechanism in section 6 of the ISDA Master Agreement and it is why the Master Agreement insists that all transactions under it form a single agreement.

Multilateral netting cannot be achieved by novation between pairs alone, because there is no bilateral contract between Northbank and the system. It requires either a central counterparty that becomes the legal counterparty to everybody, or statutory protection of the sort the Settlement Finality Regulations 1999 confer on designated systems, or both.

Where the risk goes: the unwind, worked through#

Return to the four banks and suppose Northbank fails during the cycle, before settlement, and cannot meet its net debit of £110 million.

In a system with position netting and no prefunding, the classic remedy is to remove the failed participant’s payments and recalculate. This is called an unwind, and it is worth seeing what it does.

Delete every flow to and from Northbank. The remaining matrix is:

Recalculate the net positions and compare.

Look carefully at what has happened.

Eastbank had nothing to do with Northbank’s failure in any meaningful sense — its bilateral position with Northbank was a net £40 million payable. Yet its settlement obligation has increased fivefold, from £10 million to £50 million, and it must find the difference at the moment of settlement, with no notice. Southbank’s obligation has fallen. Westbank, which had planned its day around receiving £200 million, receives £100 million.

Two properties of that table are the reason central banks stopped tolerating unwind as a risk control. First, the direction and size of the change bear no reliable relationship to a participant’s dealings with the failed party; Eastbank cannot manage its exposure to an unwind by managing its exposure to Northbank, because the effect propagates through everybody else’s positions too. Second, the change is discovered at exactly the worst moment: a bank learns that it needs £40 million more, in central bank money, at the instant a major counterparty has failed and every other bank in the country is scrambling for liquidity. That is the mechanism by which a single failure becomes a system-wide liquidity event.

The rules written in response#

In November 1990 the central banks of the Group of Ten published the Report of the Committee on Interbank Netting Schemes, universally known as the Lamfalussy Report after its chairman, Alexandre Lamfalussy. It set six minimum standards for netting schemes, which remain the foundation of every rule written since.

In summary, the six standards require that netting systems have a well-founded legal basis in all relevant jurisdictions; that participants clearly understand the effect of the scheme on each financial risk; that multilateral netting systems have clearly defined procedures for managing credit and liquidity risk, specifying who is responsible for what and placing limits on the maximum credit exposure any participant can create; that multilateral netting systems be capable, at a minimum, of ensuring timely completion of daily settlement in the event of an inability to settle by the participant with the largest single net debit position; that admission criteria be objective, publicly disclosed and permit fair and open access; and that the operational reliability of technical systems and the availability of backup facilities be assured.

The fourth is the one that changed the industry. It is the requirement now universally known as “cover one”: the system must be able to complete settlement on time even if its single largest debtor cannot pay. Everything below follows from it.

Net debit caps and net sender caps. A cap limits how large a net debit position a participant may build up between settlements. When a participant reaches its cap it can no longer send payments until incoming payments reduce its position or settlement occurs. In the Faster Payment System these are called Net Sender Caps, and Pay.UK announced in July 2026 a move away from a prescriptive formula, based on a Minimum Net Sender Cap and a Peak Contingency Value, towards participants setting their own caps against quarterly plans reviewed by Pay.UK. The purpose is unchanged: to bound the size of the hole any single participant can leave.

Prefunding. The Bank of England introduced prefunding for Bacs and Faster Payments in September 2015 to address the settlement risk arising from the build-up of obligations in deferred net systems. Under prefunding, each settlement participant holds cash in a segregated prefunding account in RTGS equal to its cap. If it defaults, the cash set aside is used to complete settlement. As the Bank puts it, this eliminates credit risk between settlement participants: the money is already there, in central bank money, before the obligation arises.

Prefunding is the reason the modern British answer to “what if a bank cannot settle its Bacs position” is no longer “unwind the cycle”. It is “use the money it already put aside”.

Loss-sharing arrangements. Where prefunding is not used, systems typically operate survivors-pay arrangements: a defaulter’s shortfall is allocated among the remaining participants according to a pre-agreed formula, usually related to their bilateral exposures to the failed party. These arrangements produce certainty of settlement completion at the cost of mutualised loss, and their existence is one of the reasons access criteria to net settlement systems are strict.

Central counterparties in outline#

A central counterparty is the industrial version of the kitchen table. It is worth understanding in outline because it is where netting is taken to its logical conclusion, and because central counterparties are now the largest single concentration of financial risk in the world.

Novation#

The defining act of a CCP is novation. When two parties agree a trade and submit it for clearing, the original contract between them is legally extinguished and replaced by two new contracts: one between the buyer and the CCP, and one between the CCP and the seller. The CCP becomes buyer to every seller and seller to every buyer.

That single legal move accomplishes two things at once. It makes multilateral netting legally trivial, because every participant now has exactly one counterparty and can net everything it has against that one counterparty. And it removes the need for any participant to assess the creditworthiness of any other participant, because it deals only with the CCP.

Margin and the default waterfall#

Because everybody’s credit risk is now concentrated in one place, that place must be made extremely difficult to break. CCPs do this with a layered stack of financial resources, consumed in a fixed order known as the default waterfall.

Initial margin is collateral posted by each clearing member against the potential future loss on its portfolio, sized to cover the likely movement in that portfolio over the period it would take to close it out. LCH’s SwapClear service calculates initial margin using its PAIRS methodology — Portfolio Approach to Interest Rate Scenarios — drawing on ten years of historical market data.

Variation margin is collected daily, and intraday where required, to settle the actual mark-to-market change in each portfolio. Variation margin is not collateral; it is payment of losses as they accrue, which prevents losses from accumulating.

The default fund is a mutualised pool contributed by all clearing members.

Skin in the game is a tranche of the CCP’s own capital, deliberately placed in the waterfall ahead of the surviving members’ contributions so that the CCP shares the loss before its members do.

At LCH the consumption order is the defaulting member’s posted margin, then the defaulter’s default fund contribution, then LCH’s own capital, then the non-defaulting members’ default fund contributions. If all of that is exhausted, variation margin gains haircutting is applied as a final measure: members who have made gains have those gains cut to fill the hole.

The international standard governing the size of these resources is the CPMI-IOSCO Principles for Financial Market Infrastructures, published in 2012. Principle 4 requires an FMI to maintain financial resources sufficient to cover the default of the participant to which it has the largest exposure in extreme but plausible market conditions — cover one — and requires CCPs involved in activities with a more complex risk profile, or that are systemically important in more than one jurisdiction, to cover the two largest participants and their affiliates. That is the origin of “cover two”.

The default management process, and the one time it was tested at scale#

When a clearing member defaults, a CCP does not simply liquidate. LCH’s SwapClear process runs in three stages: porting non-defaulting clients to surviving clearing members; neutralising the residual risk by hedging, with the assistance of a default management group drawn from member firms; and then auctioning the defaulter’s portfolio, split by product and currency, to the surviving members.

The definitive test was Lehman Brothers. At default, Lehman’s cleared interest rate derivatives portfolio at LCH comprised 66,390 trades with a notional value of $9 trillion, against which it had posted initial margin of the order of $2 billion. LCH completed the default management within about three weeks, using 100 per cent of the variation margin posted by the defaulting member and only 35 per cent of its initial margin. No mutualised default fund resources were consumed and no surviving member lost money.

That outcome is the strongest argument ever made for central clearing, and it is the reason the G20 leaders committed at Pittsburgh in September 2009 to central clearing of standardised over-the-counter derivatives, a commitment implemented in Europe through the European Market Infrastructure Regulation and retained in UK law after withdrawal from the European Union.

The concentration objection#

The obvious criticism follows directly from the plain version of this chapter. Netting concentrates risk, and a CCP is netting taken to its limit. A CCP that fails does not fail one counterparty; it fails an entire market simultaneously, and the markets it clears may have no functioning alternative.

This is not a fringe concern. It is why CCPs are supervised directly by central banks — in the United Kingdom, by the Bank of England, which supervises central counterparties alongside central securities depositories and recognised payment systems — why the PFMI impose cover-two requirements, why CCP recovery and resolution regimes exist, and why the question of who bears loss beyond the waterfall is among the most contested in financial regulation.

One institutional fact captures the point. LCH Ltd is a CHAPS Direct Participant: a clearing house sitting directly on the Bank of England’s real-time gross settlement system, alongside the high street banks, because when a clearing house needs to move money it cannot afford to depend on anybody else’s netting.

Why Bacs takes three days#

We can now answer the question properly.

The Bacs cycle has three named days, and the scheme documentation is precise about them.

Day 1, input day. This is the latest day on which a service user or bureau may submit a payment file for a given processing cycle. Files must be transmitted between 07:00 and 22:30. Submissions may be made up to 30 days in advance of the payment date, and a payment can be recalled after submission provided the submitter’s payment service provider is notified before a specified cut-off. Through the day, submitted data is validated and sorted into bank order by the central infrastructure.

Day 2, processing day. The sorted files are delivered to the recipient payment service providers, which process each payment into their own systems.

Day 3, entry day. Payments are simultaneously credited to recipients’ accounts and debited from the payer’s account. Interbank settlement of the multilateral net positions takes place at the Bank of England once that business day.

Now: what does each day buy?

The netting window. Netting only compresses obligations that coexist. A wider window means more offsetting. The Bacs cycle accumulates a full day of national payroll, benefits, utility collections and supplier payments before it computes anybody’s position, and the result is the compression already quoted: about £23 billion of gross value discharged with about £5.3 billion of settlement.

Simultaneity of debit and credit. This is the point most commentary misses entirely. Bacs is often described as “three days of the money being in limbo”. It is not. Nothing leaves the payer’s account on day 1 or day 2. The debit and the credit both occur on day 3, at the beginning of the operating day, and the interbank settlement occurs on the same day. The three-day figure is the length of a pipeline, not the length of an exposure. That is a materially different risk profile from a system where the credit happens instantly and settlement happens later.

The exception window. Direct debits and direct credits can go wrong in specific, catalogued ways: a closed account, an insufficient balance, a cancelled instruction, a changed account number. Bacs provides a set of automated advice services for exactly these cases — ADDACS for direct debit amendments and cancellations, AUDDIS for mandate lodgement, ARUCS and AWACS for returned and amended credits. Those services require a defined window in which the receiving bank can inspect an item, decide whether it can be applied, and generate a report back to the originator. A cycle with no processing day has nowhere to put that work.

Predictability, for both sides. An originator running payroll for 12,000 people needs to know precisely which day the money lands, weeks in advance; a fixed cycle with a published calendar delivers that, and “as fast as possible” does not, because it is not a date. A bank, equally, that knows its Bacs settlement occurs once a day on a known set of items can forecast its central bank money position, arrange collateral and fund accordingly, which is far cheaper than holding a buffer against a continuous unpredictable drain.

So is three days a design choice or a defect?

The honest answer is: both, and the proportions have changed over time. Bacs began in 1968 as the Inter-Bank Computer Bureau, in an era when payment files travelled between banks as reels of magnetic tape moved by van. Three days had a physical meaning then that it does not have now, and it would be dishonest to pretend the current cycle length was derived from first principles in the present decade.

But the evidence for it being load-bearing rather than merely vestigial is strong, and it is this: when the United Kingdom decided it needed same-day retail payments, it did not speed Bacs up. It built the Faster Payment System alongside it in 2008, and left Bacs running exactly as it was. Seventeen years later Bacs is still carrying 27 million payments a day and, in 2025, 5.03 billion direct debits — an all-time annual high for the scheme.

That is what a design choice looks like from the outside. Two systems, two different points on the same trade-off curve, both running at national scale, because the trade-off is real and different users sit at different points on it.

Read that table across and the logic is visible. CHAPS eliminates settlement risk and pays for it in liquidity. Bacs eliminates most of the liquidity requirement and pays for it in time and in the machinery — caps, prefunding, cycle discipline — needed to keep the risk contained. Faster Payments sits in between, buying speed for the customer while keeping the netting benefit for the banks, and paying for it with a value cap and three settlement cycles a day instead of one.

There is no free position on that curve. There never was. Netting is the mechanism by which the industry chooses where on it to sit, and every timing behaviour that irritates the public — the three days, the cut-off, the settlement that happens on Monday for a payment made on Saturday — is a consequence of that choice being made deliberately, by people who understood what they were trading away.

The next chapter leaves the interbank layer entirely and goes down to something much smaller and much more likely to break your code: what a currency amount actually is, and why the type you use to hold it matters more than almost anything else you will write.

8.98 Common wrong ideas#

Wrong: Netting means somebody gets paid less. Right: Netting operates only on the obligations between banks; the customer layer stays gross and every payment is credited and debited in full.

Wrong: Netting is just arithmetic. Right: It is a legal act, and without designation under the Settlement Finality Regulations 1999 an administrator could cherry-pick the gross obligations and ignore the net figure.

Wrong: There is a pot in the middle holding the netted money. Right: In most net settlement systems the operator sends positions to the Bank of England, which debits and credits in a single simultaneous posting with nothing held anywhere in between.

Wrong: Netting reduces risk. Right: It concentrates it, replacing several small diversified credit exposures with one large exposure to a mechanism and to a single moment.

Wrong: A bank with little exposure to the failed participant is safe in an unwind. Right: In the worked example Eastbank’s obligation rises fivefold, from £10 million to £50 million, because the effect propagates through everybody else’s positions.

Wrong: Bilateral and multilateral netting come to much the same thing. Right: Multilateral netting lets a claim on one participant fund an obligation to another, taking the four-bank example from £290 million down to £200 million.

Wrong: Unwinding the cycle is the modern remedy when a participant cannot settle. Right: Since September 2015 the British answer for Bacs and Faster Payments has been prefunding: use the money the participant already put aside.

Wrong: Bacs means the money sits in limbo for three days. Right: Nothing leaves the payer’s account on Day 1 or Day 2; the debit, the credit and the interbank settlement all fall on Day 3.

Wrong: Three days is purely a relic of magnetic tape moved around by van. Right: It is both relic and design choice, and the proof is that when Britain wanted same-day retail payments it built Faster Payments alongside Bacs in 2008 rather than speeding Bacs up.

Wrong: A central counterparty takes risk out of the market. Right: It removes bilateral credit assessment and makes multilateral netting legally trivial, but it becomes the largest single concentration of risk, which is why cover-two requirements and CCP resolution regimes exist.

8.99 Chapter summary in 20 lines#

1. On an average working day in 2024 Bacs carried about 26.8 million payments worth roughly £23.0 billion, and about £5,265 million actually moved between settlement accounts.
2. Nobody was short-changed, because netting operates on the obligations between banks and never on the payments between customers.
3. Bilateral netting offsets obligations one pair at a time; multilateral netting offsets each participant against all the others simultaneously.
4. In the four-flatmate example, twelve debts and £102 of cash became four movements and £18 on the kitchen table, with every debt discharged in full.
5. The sum of all net debit positions always equals the sum of all net credit positions, because every debt is counted once as a plus and once as a minus.
6. In the four-bank matrix, £3,090 million of gross flows became £290 million bilaterally and £200 million multilaterally, or 9.4 and 6.5 per cent of gross.
7. Multilateral netting wins because a claim on one participant can fund an obligation to a different one, which bilateral netting cannot achieve.
8. Netting works best when flows are balanced, which is why Faster Payments eliminates something like 88 per cent of value while directional Bacs payroll traffic eliminates about 77 per cent.
9. The longer the window before squaring up, the more cancels out, so waiting is itself part of the compression.
10. Money in a settlement account is money doing nothing else, so netting is the difference between a payment system a country can afford to run and one it cannot.
11. The cost is that netting concentrates risk: each participant swaps several small exposures for one large exposure to a calculation.
12. Netting is a legal act as well as an arithmetical one, and only statutory protection stops an administrator cherry-picking the gross obligations.
13. Position netting leaves the gross obligations legally alive and gives rise to unwind risk, while netting by novation and close-out netting extinguish them.
14. Multilateral netting cannot be built from bilateral novation alone and needs either a central counterparty or statutory designation, or both.
15. An unwind redistributes obligations unpredictably, and it does so at the worst possible moment, when every bank in the country is already scrambling for liquidity.
16. The Lamfalussy Report of November 1990 set six minimum standards, of which cover one — settling on time despite the largest single debtor failing — changed the industry.
17. That standard is implemented through net debit and net sender caps, through prefunding introduced for Bacs and Faster Payments in September 2015, and through survivors-pay loss sharing where prefunding is absent.
18. A central counterparty is netting taken to its limit: novation makes it buyer to every seller and seller to every buyer, defended by initial margin, variation margin, its own skin in the game and a mutualised default fund.
19. LCH resolved Lehman’s 66,390 cleared interest rate trades of $9 trillion notional in about three weeks, using 35 per cent of the posted initial margin and no mutualised default fund resources at all.
20. Bacs takes three days because the netting window, the exception-handling window and a published calendar are worth more to its users than speed, and Britain proved the trade-off was real by building Faster Payments alongside it rather than instead of it.

Sources used: Bank of England, “Payment and settlement” and “A brief introduction to the Real-Time Gross Settlement system and CHAPS” (including Table A, daily average RTGS settlement values and volumes 2024); Bank of England, “Bank of England Settlement Accounts” and “RTGS and CHAPS Service Description”, December 2018; Pay.UK, “Annual summary of payment statistics 2025”; Pay.UK, “Bacs Payment System” and “Understanding changes to the Faster Payment System Net Sender Caps”; Bacs, “Bacs Direct Credit — Getting started” (processing cycle); Committee on Payments and Market Infrastructures, Glossary, and Report of the Committee on Interbank Netting Schemes (the Lamfalussy Report), November 1990; CPMI-IOSCO, Principles for Financial Market Infrastructures, 2012; CPMI Red Book, “Payment, clearing and settlement systems in the United Kingdom”; The Financial Markets and Insolvency (Settlement Finality) Regulations 1999, SI 1999/2979, and Directive 98/26/EC; DTCC, UST1 FAQs and press release of 23 April 2025; CLS Group, “What are the necessary ingredients for any PvP arrangement to work?”; BIS Quarterly Review, “Uncovering FX settlement risk: new measures from the 2025 BIS Triennial Survey”; LSEG, LCH SwapClear risk management; CCP Global, “The Lehman Case”.