Two Kinds of Money

6.0 What this chapter gives you#

1. You will be able to explain why a payment between two customers of the same bank needs no cooperation from anybody, while an identical-looking payment to another bank eventually requires the Bank of England to change a number in a ledger of its own.
2. You will be able to name the two kinds of sterling — central bank money and commercial bank money — and say exactly who may hold each of them.
3. You will be able to work a day’s multilateral net position out of a list of customer payments, and explain why £2,245 of payments moved £450 in the upper ledger.
4. You will be able to explain why drawing £20 from a cash machine converts one kind of money into the other, and what the singleness of money means.
5. You will be able to say why reserves are not depositors’ money in storage, and why the total fell by £63 billion in a year without anybody being paid a penny.
6. You will be able to decide whether a payment is genuinely on-us by asking which legal entity holds each account rather than which brand is on the app.
7. You will be able to distinguish the four account types at the Bank of England and say which are remunerated at Bank Rate and which may not hold funds overnight.
8. You will be able to explain why a Faster Payment landing at two on a Sunday morning leaves one bank exposed to another, and what prefunding does about it.
9. You will be able to describe tiering, and say why an indirect participant’s settlement asset is commercial bank money rather than central bank money.
10. You will be able to answer “why is this payment behaving perversely” by asking which tier the obligation is sitting in and what has to happen before it moves.

The previous chapter established what a bank is: an institution whose fundamental product is a promise, and whose balance sheet records your deposit as its debt. This chapter takes that one step further and asks an awkward question. If your bank’s promise is money to you, whose promise is money to your bank?

The question is not philosophical. It has a specific, mechanical, unromantic answer, and that answer is the hinge on which the rest of this book turns. Almost every rule about settlement that arrives later — why a CHAPS payment cannot be recalled, why Bacs takes three working days, why a Faster Payment reaches the recipient in seconds but the banks square up hours later, why a card refund is slow, why an international transfer travels through institutions nobody asked for — is a consequence of the fact that there are two distinct kinds of money in circulation at the same time, in the same currency, denominated in the same pounds, and that they are not interchangeable in the way the word “pounds” implies.

Consider two payments that look identical from the outside. Priya banks with Lloyds and pays £120 to Sam, who also banks with Lloyds. Priya’s neighbour banks with Lloyds and pays £120 to a friend at Barclays. Both payments are for £120, both are sent from a phone, both arrive in seconds, both produce identical-looking lines on identical-looking statements. But the first payment never leaves Lloyds and requires no cooperation from any other institution in Britain. The second cannot be completed by Lloyds alone at any price, and eventually requires the Bank of England to change a number in a ledger of its own.

That asymmetry is invisible to customers, is largely invisible to bank staff, and governs everything. This chapter makes it visible.

The plain version#

Imagine a town with two banks. One is run by Ada, the other by Mercer. Between them they hold the accounts of everybody in town.

Ada’s customers include Priya, Sam and Tom. Mercer’s customers include Ravi and Nadia. Each bank keeps a book with one page per customer, showing how much the bank owes that person. That is what a bank balance is: a written record of a debt owed to you. Nobody has a personal pile of coins in a personal box. There is a book, and there is a line in it with your name on.

The easy payment#

On Monday, Priya pays Sam £120. Both of them are Ada’s customers.

Ada does not need to consult anybody. She opens her book, reduces Priya’s line by £120 and increases Sam’s line by £120. That is the whole event. Nothing has left Ada’s building. No cash has moved, no van has driven anywhere, no message has been sent to any other institution. Ada owed £120 to Priya before, and she owes it to Sam now. Her total debt to her customers is exactly the same as it was five minutes ago. What changed is the label on it.

This is worth stopping on because it is genuinely surprising the first time you see it clearly. From Ada’s point of view, a payment between two of her own customers is not a transfer of anything. It is a relabelling exercise. She has rewritten who she owes, and that is all.

The hard payment#

On Tuesday, Tom pays £800 to Ravi. Tom is Ada’s customer. Ravi is Mercer’s.

Ada can do her half easily. She reduces Tom’s line by £800, and her total debt to her customers falls by £800. She is £800 better off.

Mercer can do his half easily too. He increases Ravi’s line by £800, and his total debt to his customers rises by £800. He is £800 worse off.

And that is the problem. Mercer has just taken on a new obligation to Ravi in exchange for nothing at all. Ada has just been released from an obligation, also in exchange for nothing at all. If the story stops here, Ada has been enriched by £800 at Mercer’s expense, and Mercer would be an idiot to let it happen again.

So Ada has to give Mercer something worth £800. The obvious candidate is a promise: Ada writes Mercer an IOU. But Mercer does not want Ada’s IOU. Ada might be perfectly sound today, but if she is unsound in six months Mercer is left holding a piece of paper and a real debt to Ravi. If the town’s banks paid each other in IOUs, every bank would end up quietly lending to every other bank all day long, without ever deciding to, and a single failure would take the lot down.

What Mercer wants is money that does not depend on Ada being solvent.

The vault in the market square#

So the town builds one. In the market square there is an institution — call it the Old Vault — which is not a bank for people. It is a bank for banks. Ada has an account there. Mercer has an account there. So does every other bank in town, and nobody else.

Now Tuesday’s payment can complete. Ada tells the Old Vault to move £800 from her account to Mercer’s. The Vault reduces Ada’s line by £800 and increases Mercer’s by £800. Mercer has been paid in something he trusts, and he is content to owe Ravi the £800, because he has received £800 of something better than an IOU from Ada.

That is the two-tier system, in one picture. There are two ledgers, not one. Down at street level, banks keep books recording what they owe their customers. Up above, the Vault keeps a book recording what it owes the banks. Ordinary payments between customers of different banks cause changes in both books at once, and the payment is not really finished until the upper book has been changed.

The two kinds, named#

There are therefore two kinds of money in this town, and they are not the same substance even though they are counted in the same pounds.

The first kind is what you and I have. A balance at Ada’s Bank is a promise from Ada. It is only as good as Ada. If Ada fails, you are a creditor of a failed business.

The second kind is what banks have. A balance at the Old Vault is a promise from the Vault. Banks accept it from one another without hesitation for three reasons: the Vault is not trying to make money out of them, the Vault cannot run out of the town’s currency because the Vault is the thing that issues it, and the Vault is not going to go bust.

In the real United Kingdom the Old Vault is the Bank of England, and the accounts banks hold there are called reserves accounts. The money in them is called central bank money. The money in your account is called commercial bank money. Once you have those two phrases, an enormous amount of otherwise baffling behaviour becomes obvious.

One thing ordinary people can hold#

There is exactly one form of central bank money that a member of the public can hold, and it is in your wallet. A £20 note is a direct promise from the Bank of England. It is not a promise from Barclays or Nationwide or anybody else. As of the Bank’s most recent published figures there are over 4.98 billion Bank of England notes in circulation, together worth about £91.5 billion.

So when you draw £20 from a cash machine, something genuinely interesting happens: you convert £20 of commercial bank money into £20 of central bank money. Your claim on your bank shrinks by £20 and you acquire a claim on the central bank instead. It does not feel like a change of substance because the two spend identically, and the fact that they spend identically — that nobody in any shop in Britain asks whether your £20 is a Lloyds £20 or a paper £20 — is the single greatest achievement of the whole arrangement. It is called the singleness of money, and it is not an accident. It is engineered, and the thing that engineers it is the shared ledger in the middle.

A day, with numbers#

Take one full day in the town, so the arithmetic is concrete.

At the start of Wednesday, Ada holds £45,000 at the Old Vault and Mercer holds £60,000. Five payments happen.

Five customers have had their balances changed. In total, £2,245 of payments have been made. Every payer’s line went down and every payee’s line went up by the full amount, exactly as they would expect.

Now look at the Old Vault’s book. Two of those five payments — Priya to Sam, and Sam to Priya — were between customers of the same bank. The Old Vault never heard about them. They do not exist as far as the upper ledger is concerned. That leaves three payments that cross between banks: £800 from Ada’s side to Mercer’s, and £250 plus £1,000 coming back the other way.

Ada owes Mercer £800. Mercer owes Ada £1,250. There is no reason to move both amounts, so the Vault moves the difference: £450 from Mercer to Ada. Ada finishes the day with £45,450 and Mercer with £59,550. The Vault’s total is unchanged at £105,000, because nothing was created or destroyed up there — one bank’s pile grew and another’s shrank by the same amount.

Look at what that means. Five payments, £2,245 of customer activity, and the number that actually changed hands in the upper ledger was £450, once. The rest either cancelled out or never left the bank it started in.

That single observation explains more about how payment systems are built than anything else in this book. It is why banks care so much about which of their customers pay each other. It is why the machinery that works out the £450 exists at all. And it is why, when things go wrong, the question that matters is never “did the customer’s balance change” but “did the upper ledger change, and when”.

Where the plain version stops being true#

The vault-in-the-square picture is the right mental model, and practitioners carry a version of it in their heads. But four things about it are wrong in ways that will bite.

The reserves are not your money in storage#

The picture invites you to imagine that Ada’s £45,000 at the Vault is the town’s real money, kept safe on behalf of Ada’s depositors, and that Ada’s deposits are somehow backed by it. That is not what reserves are, and the error matters because it produces confident nonsense about banking.

Reserves are a liability of the central bank, created by the central bank, in whatever quantity the central bank chooses. They are not a warehouse receipt for anything. There is no ratio that must hold between a bank’s reserves and its deposits — the United Kingdom has had no mandatory reserve ratio for decades. The two figures are not even the same order of magnitude. At end-February 2026, total balances held in reserves accounts at the Bank of England stood at £643,547 million, while the UK’s broad money measure M4, which is overwhelmingly commercial bank deposits, was running at around £3.3 trillion. Reserves cover something in the region of a fifth of deposits, and that fraction is not a design parameter. It is a residue of monetary policy.

You can see this directly. Reserves balances fell from £706,745 million at end-February 2025 to £643,547 million a year later. No bank paid that £63 billion to anybody. It disappeared because the Bank of England was shrinking its own balance sheet through the sale and maturity of assets held in the Asset Purchase Facility. The upper ledger’s total is not a fact about the banks. It is a fact about the central bank.

The other half of the correction: reserves cannot leave the upper tier. A bank cannot lend reserves to you. Reserves circulate only among institutions that hold accounts at the central bank, plus a leak into banknotes when the public wants cash. When a bank makes a loan it creates a deposit — commercial bank money — out of nothing, as the Bank of England’s own Money creation in the modern economy set out in its 2014 Quarterly Bulletin. It does not hand over reserves. Reserves are what it needs afterwards, if and when the borrower spends the money at a customer of a different bank.

“Same bank” is a legal question, not a branding question#

The plain version says a payment between two customers of the same bank never leaves that bank. That is true, and it is the most useful sentence in the chapter. But “the same bank” means the same legal entity, and British banking brands routinely span several.

Since ring-fencing, the large UK groups are split. Barclays operates both Barclays Bank UK PLC and Barclays Bank PLC as separately authorised entities. HSBC operates HSBC UK Bank plc alongside HSBC Bank plc. Lloyds Banking Group contains Lloyds Bank plc, Bank of Scotland plc and Lloyds Bank Corporate Markets plc. A payment from a retail customer of one entity to a corporate customer of another entity in the same group is not an on-us payment. It has an interbank obligation in it, and it settles like any other.

The second half of this correction is about risk rather than plumbing. Even a genuine on-us payment does something real: it moves the exposure. Before Priya pays Sam, Priya is the bank’s creditor for £120. Afterwards, Sam is. If the bank fails, it matters enormously which of them is holding the claim, because deposit protection is per depositor, per authorised firm. In the United Kingdom that limit rose from £85,000 — the level set in 2017 — to £120,000 on 1 December 2025, with the temporary high balance limit for qualifying life events rising from £1 million to £1.4 million on the same date. An on-us payment that pushes a customer from £90,000 to £150,000 has quietly moved £30,000 out of protection. Nothing settled, nothing cleared, and the risk profile of the bank’s deposit book changed.

Nothing moves at the Old Vault when your payment arrives#

The picture implies that a cross-bank payment causes an immediate transfer in the upper ledger. For one UK payment system that is true. For the systems that carry almost all the volume, it is false, and the gap is where a whole category of risk lives.

CHAPS settles gross and in real time: each payment moves central bank money individually, when it is made. Everything else settles net and later. Faster Payments net settlement takes place three times each business day. Bacs settles once a day. The Image Clearing System for cheques settles once a day. Visa Europe and Mastercard Europe settle on twenty-four-hour cycles, with weekend and bank holiday cycles settling the following business day. PEXA settles six times a business day. LINK settles daily.

So when a Faster Payment lands in a stranger’s account at two in the morning on a Sunday, the recipient’s bank has credited a real customer with real, spendable money while holding nothing but an obligation from the sending bank, and it will hold that obligation until the next settlement cycle. That interval is a credit exposure of one bank to another, created by a customer, without either bank’s credit committee being consulted. The Bank of England’s answer is prefunding: it offers a service that lets a payment system operator cap each participant’s maximum possible net debit position at the amount that participant has locked in a segregated prefunding account in RTGS. If the participant fails, the cash is already there and settlement completes anyway.

The general rule to carry forward is this. A customer payment and its settlement are two different events with two different timestamps. Assuming they are the same event is the single most common error in payments engineering, and it is the reason Chapter 7 exists.

There is not one account per bank, and most institutions have none#

The picture has a tidy one-to-one mapping: one bank, one account at the Vault. Reality has four account types with different rights, and most regulated firms have none of them.

The Bank of England distinguishes reserves accounts, reserves-and-settlement accounts, settlement accounts, and omnibus accounts. They differ on who may hold them, whether they are remunerated, and whether balances may sit overnight. A settlement account held by a non-bank payment service provider is unremunerated and may not hold funds beyond what settlement requires. A reserves account is remunerated at Bank Rate and may hold balances overnight. That distinction is not administrative trivia; it decides whether an institution earns anything on its central bank balances, which is a first-order fact about its business model.

And the population is small. The Prudential Regulation Authority regulates around 1,300 firms. At end-February 2026, 221 participants had access to Sterling Monetary Framework facilities. More than 70 organisations use RTGS to settle directly in one or more payment systems. Just over 35 organisations access CHAPS directly. Several thousand financial institutions reach CHAPS indirectly, by sending their payments through one of those direct participants.

For all of those indirect institutions, the two-tier picture repeats one level down, and the top tier is not the central bank. A small firm that clears through Barclays settles in a Barclays deposit. Its settlement asset is commercial bank money. It has the same exposure to Barclays that Priya has, and it has it in size. This is called tiering, it is a standing concern of financial stability regulators, and it is the mechanism behind correspondent banking and the nostro and vostro accounts that Volume IV takes apart.

The technical version#

Definitions#

Central bank money is a liability of the central bank. In sterling it takes two forms: Bank of England banknotes, which are bearer liabilities available to the general public, and balances held in accounts at the Bank of England, which are book-entry liabilities available only to institutions that meet the Bank’s access criteria. Central bank money is the settlement asset for sterling: it is what an institution receives when a sterling obligation to it is finally discharged.

Commercial bank money is a deposit liability of a commercial bank, building society or credit union — an unsecured, unsubordinated claim on that institution, ranking as a depositor claim in insolvency and covered, up to the applicable limit, by the Financial Services Compensation Scheme. It is the money the public holds, the money businesses hold, and by value it is the overwhelming majority of the money in existence.

The two-tier system is the arrangement in which the central bank provides accounts to a restricted set of institutions, those institutions provide accounts to the public, and the public’s money is therefore a claim on a private firm while the private firms’ money is a claim on the state. The Bank for International Settlements described this in its September 2025 report on wholesale central bank money as a “time-tested two-tier structure”, and identified the central bank settlement asset as what preserves the singleness of money: the property that payments denominated in the sovereign unit of account settle at par regardless of which form of money is used. The foundational treatment remains the Committee on Payment and Settlement Systems report The role of central bank money in payment systems, published on 12 August 2003.

Accounts at the Bank of England#

Access is governed by the Bank’s published access policy for RTGS settlement accounts and services. Four account types exist.

A reserves account is available to PRA-authorised UK-incorporated banks, building societies and investment firms designated by the PRA for prudential supervision, to UK subsidiaries and branches of their non-UK incorporated equivalents, and to central counterparties and international central securities depositories. Balances are remunerated at Bank Rate — 3.75% at end-February 2026, having fallen from 4.5% over the preceding twelve months — and may be held overnight. Holding a reserves account requires participation in the Sterling Monetary Framework.

A reserves and settlement account is the same account used for both purposes: holding funds and settling obligations arising from direct participation in a payment system that settles in RTGS.

A settlement account is for firms not eligible for reserves: FCA-authorised non-bank payment service providers, systemically important financial market infrastructures, and entities within a wider banking group that already holds a reserves account. Settlement accounts are unremunerated and may not hold funds beyond what is needed to support settlement. The Bank opened this route to non-bank PSPs in 2017, making it the first G7 central bank to do so.

An omnibus account, offered since 2021, allows the operator of a payment system recognised by HM Treasury under the Banking Act 2009, and designated under the Financial Markets and Insolvency (Settlement Finality) Regulations 1999, to co-mingle reserves belonging to multiple participants in a single account, so that balances in the operator’s own ledger are backed by central bank money. Omnibus accounts are remunerated at Bank Rate. The first user was the Sterling Fnality Payment System, which processed its first live payments on 14 December 2023.

RTGS: the ledger that holds the upper tier#

The Bank of England’s Real-Time Gross Settlement infrastructure was introduced in 1996. It is, in the Bank’s own words, “an accounting system” — the ledger that underpins settlement of sterling payments, rather than a payment system in its own right. Responsibility for operating CHAPS, the UK’s high-value payment system, transferred to the Bank in November 2017, so that the Bank now acts as both settlement agent and system operator. On 28 April 2025 the Bank completed a multi-year renewal programme, moving to a new core ledger and settlement engine known as RT2; the participant-facing interface Enquiry Link was replaced by BERTI at the same time.

RTGS provides settlement for CHAPS, for the payment system embedded within CREST, for seven retail payment systems — Bacs, the Image Clearing System, Faster Payments, LINK, Mastercard Europe, PEXA and Visa Europe — and for the Bank’s Note Circulation Scheme. It supports four settlement models: real-time gross settlement, delivery versus payment, prefunded deferred net settlement and standard deferred net settlement.

Average daily settlement in RTGS, Q2 2026:

Two things in that table deserve attention. The first is the gulf between CHAPS at £411.6 billion a day and Faster Payments at £2.3 billion a day, when Faster Payments carries vastly more transactions. That is netting, and Chapter 8 is about it. The second is that the entire retail payments experience of sixty-odd million people — every Faster Payment, every direct debit, every card transaction settling through Visa Europe or Mastercard Europe, every cash machine withdrawal — reduces to about £12.6 billion of central bank money changing hands each day, roughly three per cent of the RTGS total.

The on-us case, stated exactly#

An intrabank payment — variously called an on-us payment, an internalised payment, or a book transfer — is a payment in which the payer’s account and the payee’s account are held by the same legal entity.

The accounting is a single pair of postings within that institution’s own ledger. The payer’s account is debited and the payee’s account is credited. Both are liability accounts. The bank’s balance sheet total is unchanged; only the composition of its deposit liabilities changes. There is no interbank obligation, and therefore nothing to clear, nothing to net and nothing to settle. No reserves move. No message reaches the Bank of England. The payment is complete at the moment the bank’s ledger says it is complete, and its finality is a matter of that bank’s own terms and systems rather than of any scheme rulebook or settlement finality designation.

Whether the bank routes such a payment through the relevant scheme’s central infrastructure anyway — for addressing, for Confirmation of Payee, for tracking, or simply because it is easier to have one code path — is an implementation choice rather than a settlement necessity, and the Payment Systems Regulator says as much in its reporting guidance on authorised push payment scams. That guidance defines an on-us transaction as a push payment between two account holders within the same provider group, and records that, depending on a provider’s commercial arrangements, such payments may pass either externally via Faster Payments or internally as a book transfer. Firms are required to report both, and to itemise the internally booked ones separately so that the regulator can check its totals against Pay.UK’s own Faster Payments figures. The settlement consequence, however, is not in doubt: an on-us payment generates no interbank obligation, and no quantity of central bank money moves as a result of it.

The best evidence that this is a material phenomenon rather than a curiosity is that the Bank of England’s own published statistics have to carve it out. The Bank states that CHAPS represents 0.4% of UK total payment volumes but 91% of total sterling payment values, and appends the qualification “excluding internalised flows within payment service providers”. The official numbers exclude on-us traffic because it never touches the settlement layer and therefore cannot be observed there. A payment that does not settle is, from the settlement system’s point of view, a payment that did not happen.

For a practitioner this has consequences beyond accounting. On-us payments do not attract interbank settlement obligations or, in card schemes, interchange between different institutions. They are not visible in scheme-level reconciliation files. They may be subject to different internal cut-offs, different limits and different fraud controls from their interbank equivalents. And in a large bank with a high market share, they can be a very large fraction of volume — which is exactly why a payments platform must be designed so that the on-us and interbank paths cannot diverge in behaviour that customers can see.

The interbank case, stated exactly#

When payer and payee bank with different institutions, the payment creates an obligation between those institutions that must be discharged in the settlement asset. There are two ways to do it.

Real-time gross settlement. Each payment is settled individually, in central bank money, at the moment it is processed. In sterling this model is used only by CHAPS. Settlement is final and irrevocable between direct participants as it occurs, which is why CHAPS is the instrument of choice for house purchases and for the sterling leg of transactions that must not unwind: CLS, the foreign exchange settlement system, completes its final sterling settlement through direct participation in CHAPS, and LCH and Euroclear Bank are also direct participants.

The CHAPS numbers for calendar year 2025: volumes grew 1.1% to a record 53.3 million payments, an average of 210,482 each working day; total value settled rose 7.3% to £93.9 trillion, an average of £371.3 billion daily. The average payment value was £1.8 million and the median was £4,586 — a spread that tells you a great deal, since 94% of CHAPS payments were for £1 million or less while financial institution payments, carried in the pacs.009 message, were 24% of volume but 73% of value. The Bank observes that CHAPS settles the annual UK GDP every nine working days. The all-time peak volume day was 2 April 2024 with 344,099 payments; the all-time peak value day was 3 October 2022 with £642.7 billion.

The CHAPS settlement day currently runs 06:00 to 18:00, Monday to Friday. Direct participants must be able to receive by 08:00 and to send by 10:00; the cut-off for customer payments is 17:40. Participants choose whether to submit a payment as urgent, settling immediately subject to available liquidity, or non-urgent, accepting some delay in exchange for better liquidity efficiency through the Liquidity Saving Mechanism; non-urgent payments settle around eight minutes after submission on average, and are auto-promoted to urgent after 17:30. From September 2027 the Bank will open CHAPS settlement from 01:30 rather than 06:00, and it has consulted on weekend settlement not before 2029 and a 22-hours-by-6-days pattern not before 2031, with 22x7 and 23.5x7 identified as possible end states.

Deferred net settlement. The system operator collects the individual payment instructions, calculates each participant’s multilateral net position at the end of a cycle, and instructs the Bank of England to settle a single figure per participant. This is dramatically cheaper in liquidity than gross settlement, and it is what makes high-volume, low-value payments economic. The cost is that customers are credited and debited before interbank settlement occurs, so participants carry an intraday exposure to one another. Prefunding removes that exposure by capping each participant’s maximum net debit at cash it has already segregated in RTGS.

The legal layer#

Two statutory mechanisms turn these arrangements from private contracts into something that survives an insolvency.

Recognition under the Banking Act 2009 brings a payment system within the Bank of England’s supervisory remit, on a recognition order made by HM Treasury. Designation under the Financial Markets and Insolvency (Settlement Finality) Regulations 1999 protects transfer orders and netting arrangements within a designated system from being unwound by ordinary insolvency law — so that a settlement which has occurred stays occurred even if a participant enters administration later the same day. Both are prerequisites for an omnibus account, and both are the reason the phrase “settled in central bank money” carries legal weight rather than merely describing where the entry was made.

The same structure elsewhere#

The two-tier arrangement is not a British peculiarity. Every major currency is organised this way, with local variations in access.

In the United States, the Federal Reserve Banks provide master accounts to legally eligible depository institutions, which are the gateway to Fedwire Funds, Fedwire Securities, the National Settlement Service, FedACH, FedCash and FedNow. Deposits of depository institutions at the Reserve Banks stood at $2,947,630 million on the H.4.1 release of 13 August 2026. In May 2026 the Federal Reserve proposed a new, more limited “payment account” as a route for institutions that do not qualify for a full master account.

In the euro area, the Eurosystem operates TARGET Services. In 2025, T2 settled a daily average of 431,067 euro payments worth €1,932.8 billion, a total of €492,859.0 billion for the year, with an average transaction value of €4.5 million; the ECB notes that T2 settles the equivalent of euro-area GDP in around eight days of operations. The instant payment service TIPS settled a daily average of 2,735,053 euro payments worth €2.2 billion — the same inversion of volume and value that Faster Payments shows against CHAPS.

Central bank digital currency, briefly#

A retail central bank digital currency would be central bank money held directly by households and businesses in electronic form. That is the entire proposition, and it explains why the topic generates the heat it does: it would be the first genuinely new form of tier-one money available to the general public since the banknote, and it would change the answer to the question this chapter is built on — who may hold a claim on the central bank.

The Bank of England and HM Treasury are in a design phase for a digital pound, examining technology and policy requirements. No decision to build or issue has been taken. The Bank has said the earliest it would issue a digital pound is the second half of this decade, that a decision on next steps would come in 2026, and that Parliament would have a say before any launch. It has also said that a digital pound would not replace cash, and that the Bank will continue to issue banknotes for as long as people want to use them. A Digital Pound Lab for practical experimentation was announced in October 2025.

The arguments are worth stating without taking sides. Proponents point to declining cash use, to the desirability of a public payment option that does not depend on any commercial firm, and to the anchoring role of a monetary asset available to everyone at par. Opponents and cautious central bankers point to disintermediation — if households can hold central bank money directly, deposits may leave banks, particularly in a stress — and to privacy, which is why most design work assumes holding limits and an intermediated model in which private firms, not the central bank, hold the customer relationship. These are open questions and this book does not pretend to settle them.

On the wholesale side, the framing is different, because wholesale central bank money already exists and is already digital: that is precisely what a reserves balance is. The live work there is about the form it takes rather than its existence — whether reserves can be represented on programmable ledgers so that a securities transfer and its cash leg settle atomically. Omnibus accounts are one answer already in production; the BIS report of 18 September 2025, produced with seven central banks, is the reference for the rest.

Why this chapter governs the rest of the book#

Almost every settlement rule that follows is an answer to two questions: in whose money, and when.

CHAPS is irrevocable because it settles in central bank money at the moment of processing, and there is no mechanism by which a completed transfer of the settlement asset can be reversed by one party’s regret. Bacs takes three working days because it is a deferred net batch system whose cycle was designed around the exchange of files, and the customer-visible timing is inherited from the settlement design rather than from any technical necessity. Faster Payments has a value cap because the recipient’s bank extends credit until the next of the three daily settlement cycles, and that exposure has to be bounded. Card refunds are slow because the refund is a new clearing item that must travel back through the scheme and settle in its own cycle, not a reversal of the original. International transfers acquire a chain of intermediaries because there is no single central bank whose money both ends can hold, so the chain is a sequence of commercial bank claims stitched together — the two-tier problem repeated at each hop. Safeguarding rules exist for e-money and payment institutions precisely because their customers’ balances are not deposits and their firms are not banks, so the money must be held somewhere that survives the firm.

None of those rules is arbitrary. Each is a consequence of a decision about which kind of money discharges an obligation and at what moment. Whenever a payment behaves in a way that seems perverse, the productive question is not “why is this system badly designed” but “which tier is this obligation sitting in right now, and what has to happen before it moves”.

6.98 Common wrong ideas#

Wrong: Reserves are depositors’ money kept safe at the Bank of England. Right: Reserves are a liability of the central bank, created in whatever quantity it chooses, and no ratio has to hold between them and anybody’s deposits.

Wrong: Banks lend out reserves. Right: Reserves cannot leave the upper tier; a bank lends by creating a deposit and needs reserves only afterwards, if the borrower pays a customer of a different bank.

Wrong: Two accounts showing the same brand are with the same bank. Right: Barclays, HSBC and Lloyds each run several separately authorised entities, and a payment between two of them carries a real interbank obligation.

Wrong: An on-us payment has no consequences because nothing settles. Right: It moves the exposure, and it can quietly push a customer past the £120,000 protection limit without anything clearing or settling at all.

Wrong: A cross-bank payment moves central bank money at the moment it arrives. Right: Only CHAPS settles that way; Faster Payments settles three times each business day, and Bacs, cheques and the card schemes once.

Wrong: Every bank has an account at the Bank of England. Right: The PRA regulates around 1,300 firms, just over 35 organisations access CHAPS directly, and everybody else reaches it through somebody’s balance sheet.

Wrong: All Bank of England accounts are the same. Right: Reserves accounts are remunerated at Bank Rate and may hold balances overnight; settlement accounts for non-bank payment service providers are unremunerated and may not.

Wrong: Faster Payments matters more than CHAPS because it carries far more traffic. Right: In Q2 2026 CHAPS settled £411.6 billion a day against Faster Payments’ £2.3 billion, because netting compresses the retail systems before they reach the Bank.

Wrong: A digital pound has been decided on and will replace cash. Right: The Bank and HM Treasury are in a design phase, no decision to build or issue has been taken, and the Bank has said it will keep issuing banknotes for as long as people want them.

Wrong: Banknotes are issued by your bank. Right: A £20 note is a direct bearer liability of the Bank of England, and it is the only form of central bank money a member of the public may hold.

6.99 Chapter summary in 20 lines#

1. Your bank’s promise is money to you, but your bank needs money that does not depend on any other bank staying solvent.
2. That requirement produces a two-tier system: banks keep books recording what they owe customers, and the central bank keeps a book recording what it owes banks.
3. A payment between two customers of the same bank is a relabelling exercise inside one ledger and requires nobody else’s cooperation.
4. A payment between customers of different banks releases one bank from an obligation and burdens the other, so something of value must pass between them.
5. That something is central bank money, held in reserves and settlement accounts at the Bank of England.
6. Commercial bank money is an unsecured claim on a private firm; central bank money carries effectively no credit risk in sterling.
7. Banknotes are the only form of central bank money the public may hold, and there are over 4.98 billion of them worth about £91.5 billion.
8. That both kinds spend identically is called the singleness of money, and it is engineered by the shared ledger in the middle rather than being an accident.
9. In the worked day, £2,245 of customer payments across two banks moved £450 once in the upper ledger, and two of the five payments never left the bank they started in.
10. Reserves are not deposits in storage: no ratio binds them, and the total fell from £706,745 million to £643,547 million in the year to February 2026 without anybody being paid.
11. Reserves cannot leave the upper tier, so a bank lends by creating a deposit and worries about reserves only when the money walks out of the door.
12. “The same bank” is a legal question rather than a branding question, and ring-fencing has split the large British groups into several authorised entities.
13. Even a genuine on-us payment moves risk, because deposit protection is per depositor per authorised firm, raised to £120,000 on 1 December 2025.
14. Only CHAPS settles gross and in real time; everything else settles net and later, and that interval is a credit exposure created by a customer.
15. The Bank of England’s answer is prefunding, which caps a participant’s maximum net debit at cash already segregated in RTGS.
16. There are four account types at the Bank — reserves, reserves and settlement, settlement, and omnibus — differing on eligibility, remuneration and overnight balances.
17. Access is narrow, and the several thousand institutions that reach CHAPS indirectly settle in a direct participant’s commercial bank money instead.
18. That is tiering: the two-tier picture repeats one level down with a commercial bank at the top of it.
19. RTGS settled about £902 billion a day in Q2 2026, of which the entire retail payments experience of the country was roughly £12.6 billion.
20. Every settlement rule in the rest of the book answers two questions — in whose money, and when — so the productive question about odd behaviour is which tier the obligation is sitting in.

Sources used: Bank of England — A brief introduction to the Real-Time Gross Settlement system and CHAPS; Payment and settlement; Payment and settlement statistics; Access policy for RTGS settlement accounts and services; Bank of England Settlement Accounts; CHAPS; Banknote statistics; Report on the Bank’s official market operations March 2025–February 2026; RTGS and CHAPS Annual Report 2024–25; Reviewing access to RTGS accounts for settlement (2024); Extending RTGS and CHAPS settlement hours – next steps (2026); Money creation in the modern economy, Quarterly Bulletin 2014 Q1; The digital pound; PRA news release and PS24/25 on depositor protection (18 November 2025); Which firms does the PRA regulate. Bank for International Settlements — Wholesale central bank money in the context of tokenisation (18 September 2025); CPSS, The role of central bank money in payment systems (12 August 2003). European Central Bank — TARGET Services Annual Report 2025; T2 facts and figures. Federal Reserve — H.4.1 statistical release, 13 August 2026; Federal Reserve Financial Services, Master Account; Federal Register, proposed revisions to the Policy on Payment System Risk and account access guidelines (26 May 2026). Pay.UK — Faster Payment System. Fnality — Sterling Fnality Payment System announcement, 14 December 2023.