A Balance Is Not a Thing

2.0 What this chapter gives you#

1. You will be able to explain why the number in your banking app is a promise rather than a location, and why the envelope-in-a-drawer picture of a bank is the wrong one.
2. You will be able to state, in the Bank of England’s own words, why bank lending creates deposits rather than deposits funding loans.
3. You will be able to say what genuinely constrains a bank’s lending — capital, liquidity, funding cost and the price of money — and why no reserve multiplier appears on that list in the United Kingdom.
4. You will be able to distinguish a ledger balance from an available balance, and diagnose the customer who says a payment “took the money twice”.
5. You will be able to cite Foley v Hill and explain why a depositor is an unsecured creditor of the bank rather than the beneficiary of a trust.
6. You will be able to state the FSCS limit precisely — £120,000 per eligible depositor per authorised firm for failures from 1 December 2025 — and explain why it attaches to the banking licence and not to the brand on the card.
7. You will be able to explain the difference between FSCS deposit protection and safeguarding under the Payment Services Regulations and Electronic Money Regulations, including the narrow case in which the FSCS looks through a safeguarding account.
8. You will be able to describe a bank run in terms of maturity transformation, sequential service and multiple equilibria, and say why deposit insurance works mainly by preventing runs rather than by paying for them.
9. You will be able to explain why the queue outside Northern Rock was an aftershock and why about $42 billion left Silicon Valley Bank in a single day without anyone queuing at all.
10. You will be able to say what legal tender actually means, why it rarely comes up, and why Scottish and Northern Irish banknotes are not legal tender anywhere.

Open your banking app. There is a number at the top of the screen. Most people, if pressed, would say that number represents money that is somewhere — in a vault, in a safe, in a very large drawer with their name on it. Some part of the mind pictures a stack of notes, slightly dusty, waiting.

There is no stack. There is no drawer. There is no vault with your name on it, and there never was.

What there is instead is a record. Your bank has written down that it owes you a certain amount, and the number on your screen is a rendering of that record. It is not a description of where anything is. It is a description of a promise.

This is not a cynical observation and it is not a warning. It is simply the mechanical truth of how deposit banking works, and almost everything else in this book depends on the reader having absorbed it properly. Payments do not move substances around. They rewrite records. Until that is clear, the rest of it — clearing, settlement, netting, why a refund takes five days — will feel arbitrary. Once it is clear, most of it becomes obvious.

The plain version#

Imagine you are going away for a fortnight and you ask your neighbour to look after £500 in cash.

There are two entirely different arrangements she might be agreeing to, and it matters enormously which one it is.

In the first, she takes your five hundred pounds, puts it in an envelope, writes your name on the envelope, and puts the envelope in a drawer. She does not touch it. When you come home she gives you back the same envelope with the same notes inside. If you asked to see it at three in the morning, it would be there. This is safekeeping. The money never stopped being yours; she was just holding it.

In the second, she takes your five hundred pounds and puts it in her purse. She spends some of it on shopping. She lends a bit to her son. She uses some to pay a plumber. When you come home, she gives you five hundred pounds — but not your five hundred pounds, because those particular notes are long gone, scattered across a supermarket, a plumber’s till and her son’s petrol tank. What she gives you is five hundred pounds of somebody else’s money that she happened to have. What you had for two weeks was not an envelope. It was a promise.

Nearly everyone believes a bank is the first arrangement. Every bank in the world is the second one.

Following £500 into a branch#

Walk into a high street branch on a Tuesday morning and hand over £500 in twenty-pound notes. Watch what actually happens.

The cashier counts the notes. She puts them in her till. At the end of the day the till is emptied, the notes are bundled, and eventually they go off to a cash centre in a van, where they are mixed in with the notes from every other branch. Nobody writes your name on them. Nobody could. There is no mechanism by which a particular note stays associated with a particular person, and there is no reason anyone would want one.

Two things get written down. The bank’s own stock of banknotes goes up by £500 — those notes are now the bank’s notes, in exactly the way the shopping money was your neighbour’s. And, separately, the number beside your name in the bank’s records goes up by £500.

That second number is not a location. It is an obligation. It says: this institution owes this person five hundred pounds, payable on demand.

Now the bank does what banks do. It has thousands of customers doing the same thing, and it knows from long experience that on any ordinary day most of them will leave most of their money where it is. So it keeps enough notes and enough instantly available funds to handle the ordinary comings and goings, and it puts the rest to work — lending it to somebody buying a house, buying government bonds, financing a business. That is the entire commercial logic of a bank. It is not a cloakroom. It is an intermediary that stands between people who want their money back at any moment and borrowers who want to keep it for twenty-five years.

The arithmetic that shows there is no vault#

Here is the number that makes the point unarguable.

Every Bank of England note that exists anywhere in the world — in tills, in pockets, under mattresses, in cash centres, in shoeboxes in Spain — adds up to about £91.5 billion, spread across roughly 4.98 billion individual notes. That is the total. Those are all of them.

And yet the money held in British bank accounts runs to trillions of pounds. The Bank of England’s own explanation puts it plainly: about 96% of the money in the UK is bank deposits and only about 4% is physical cash.

So if every account holder in the country walked into a branch tomorrow and asked for notes, the notes would not exist. Not because anyone stole them. Not because of any wrongdoing at all. Because notes were never the plan. The overwhelming majority of money in Britain has never had a physical form and never will. It is entries in the records of banks.

So who makes sure you get your money back?#

If a bank is a promise, the sensible next question is what happens when a bank cannot keep its promise.

Britain’s answer is the Financial Services Compensation Scheme, the FSCS. It is a statutory scheme, funded by levies on the industry, and it does something very simple: if an authorised UK bank, building society or credit union fails, the FSCS pays your money back up to a limit, without you having to sue anybody or wait for the wreckage to be sorted out.

Since 1 December 2025 that limit has been £120,000 per eligible depositor per authorised firm. Before that date it was £85,000, a figure that had stood since 2017. The Prudential Regulation Authority — the arm of the Bank of England that supervises banks — consulted on raising it to £110,000 in March 2025 and then settled on £120,000 in its final rules, having recalculated against inflation up to September 2025.

The FSCS aims to pay most depositors within seven days of the failure, automatically, without a claim form. For certain large one-off sums — the proceeds of selling your home, an inheritance, a redundancy payment, an insurance payout — there is extra cover called a temporary high balance, currently up to £1.4 million for six months from the date the money reaches you.

If you have more than £120,000, the answer is not to worry about it but to spread it. Two banks, two limits. That is the whole strategy, and it works.

What a bank run actually is#

Now put the two halves together, and the mechanics of a bank run fall out on their own.

A bank has promised everyone their money on demand. It cannot possibly have everyone’s money available at once, because most of it is out on loan. This is not a scandal; it is the design, and it works perfectly well as long as people withdraw at the ordinary, boring rate.

A run is what happens when they stop. If enough customers try to take their money out at the same time, the bank has to sell things quickly to raise cash. Selling quickly means selling cheaply. Selling cheaply produces losses. Losses make the bank look weaker. Looking weaker frightens more customers, who withdraw, which forces more selling.

The vicious part is that running is rational. If you think there is a serious chance the bank will run out, the sensible thing to do is get in the queue early, because the money goes to whoever asks first. Everybody knows that everybody knows this. A bank can therefore be destroyed by a belief, and the belief does not have to be true when it starts.

This is precisely why deposit protection exists. If you are certain the state-backed scheme will make you whole within a week, you have no reason to queue. Deposit insurance does not mainly work by paying people after a failure. It works by making the failure less likely in the first place, because it removes the incentive to run.

The exception that proves the rule#

There is one form of money that really is a thing you hold: the note in your wallet.

A £20 note is not a record of what anyone owes you that has to be looked up. It is the claim itself, made physical. Hand it over and the payment is complete — instantly, finally, with no institution consulted, no message sent, no possibility of reversal, and no record that you were ever involved. There is nothing to reconcile because nothing was written down.

Everything else in this book is the elaborate machinery required to reproduce, over a network, what a banknote does for free when it changes hands. Seen that way, cash is not the primitive version of digital money. Digital money is a very complicated attempt to catch up with cash on the one dimension where cash is unbeatable: handing it over settles it.

Where the plain version stops being true#

That account is good enough to explain at dinner and it is wrong in four specific places. The corrections matter, because each one is a place where confident, articulate people say things that are simply not so.

First: the bank does not lend out your deposit. That story is backwards.

The plain version said the bank keeps some of your £500 and lends the rest. Almost every school textbook says this, and the Bank of England has publicly said it is not how modern banking works. In Money creation in the modern economy, published in the Bank’s Quarterly Bulletin in 2014, McLeay, Radia and Thomas put it flatly: “Rather than banks receiving deposits when households save and then lending them out, bank lending creates deposits.” And: “Whenever a bank makes a loan, it simultaneously creates a matching deposit in the borrower’s bank account, thereby creating new money.”

Read that carefully, because it inverts the intuition. When a bank grants a £250,000 mortgage, it does not go looking for £250,000 of other people’s savings to hand over. It writes two entries: a new asset (the borrower owes it £250,000) and a new liability (the borrower’s account now shows £250,000). The deposit did not exist a moment earlier. Commercial banks create the great majority of the money in the economy, in the act of lending.

Which means the accompanying story — that banks are limited by a “reserve ratio”, multiplying up a fixed base of central bank money — is also not how the constraint works. The United Kingdom has no mandatory minimum reserve ratio. What used to look like one, the Cash Ratio Deposit scheme under which larger institutions placed non-interest-bearing deposits at the Bank of England, was a funding mechanism for the Bank’s own operations, not a lending constraint, and it was replaced by the Bank of England Levy on 1 March 2024. Banks are constrained by capital requirements, by liquidity requirements, by what they can lend profitably, and by monetary policy acting on the price of money. Not by a multiplier.

Your £500 is still a real liability and the bank still has to be able to pay it. But the sentence “the bank lends out your money” should be retired.

Second: “a row in a database” is itself a simplification, and in a revealing direction.

The chapter’s own framing needs correcting. In a well-built banking system, your balance is usually not a stored number that gets edited. It is a derived figure, computed from an append-only sequence of postings — credits and debits, each immutable once written. You do not update a balance; you add an entry and the balance follows. This is not a stylistic preference. It is what makes a ledger auditable, and it is why the accounting principle in the next chapter is the single most load-bearing idea in the book.

There is a second, more practical consequence. There is rarely only one balance. Your ledger balance reflects postings that have been made. Your available balance reflects the ledger balance adjusted for holds — card authorisations that have been approved but not yet cleared, cheques not yet through, funds earmarked for a standing order due tomorrow. These two numbers are routinely different, and the difference is not an error. When a customer says a payment “took the money twice”, the overwhelmingly common explanation is that they are comparing an available balance against a ledger balance and seeing an authorisation hold that has not yet dropped off. Anyone building a payments product who models “balance” as a single mutable integer will discover this the expensive way.

Third: cash is not outside the system. It is a claim too — just a bearer one.

The plain version implied that a banknote is the one form of money that is not somebody’s promise. Look at the note. It says, in English, I promise to pay the bearer on demand the sum of twenty pounds, above the signature of the Bank’s Chief Cashier. That is not decorative. A Bank of England note is a liability of the Bank of England, and it appears as one on the Bank’s balance sheet, alongside the reserve accounts that commercial banks hold there.

What distinguishes a note is not that it is not a claim. It is that it is a claim which transfers by physical possession, against an issuer nobody expects to fail. Nobody has to be told, nobody has to agree, nothing has to be looked up. That is a different property from “not being a promise”, and confusing the two leads people to strange conclusions about what digital cash would have to be.

Scottish and Northern Irish notes make this sharper still. Those are the liabilities of commercial banks, not of the Bank of England, and the issuing banks are required to hold ring-fenced backing assets covering their entire note issue at all times.

Fourth: the run you can photograph is usually not the run that kills the bank.

The image everyone carries is the queue outside the branch. It is a nineteenth-century image and it has become, for the most part, theatre.

Northern Rock is the case in point. The queues that formed in September 2007 are the most reproduced photographs in modern British banking. But by the summer of that year, retail deposits made up only about 23% of Northern Rock’s liabilities; the other roughly 77% was wholesale funding and securitised notes. The Bank for International Settlements published Hyun Song Shin’s analysis of exactly this, and his conclusion is worth quoting: “The Northern Rock depositor run, although dramatic on television, was an event in the aftermath of the liquidity crisis at Northern Rock, rather than the event that triggered its liquidity crisis.” The real run had already happened in the wholesale markets in August, when short-term lenders declined to roll their funding over.

Silicon Valley Bank in 2023 makes the modern version explicit. There were no queues. There was no branch network to queue at. Depositors withdrew about $42 billion in a single day, on Thursday 9 March 2023, by wire transfer, largely because they had heard the same thing at the same time on the same channels. The bank was closed the following morning. A twenty-first-century run is an API call executed by a lot of people at once, and it is faster than anything the queue-outside-the-branch mental model prepares you for.

So: the run is real, the mechanism in the plain version is right, and the picture is wrong. Runs happen at the speed of the fastest funding channel, and retail depositors are rarely the fastest channel any more.

The technical version#

The deposit as a liability#

A commercial bank’s balance sheet is the cleanest way to see what your balance is.

On the asset side sit the things the bank owns or is owed: loans and mortgages advanced to customers, holdings of government and corporate securities, reserve balances held at the central bank, and physical cash. On the liability side sit the things the bank owes: customer deposits, wholesale borrowing, issued debt securities. The residual — assets minus liabilities — is equity.

Your current account balance is a liability of the bank. Specifically it is an unsecured claim against the bank, generally repayable on demand, ranking in an insolvency according to the statutory creditor hierarchy. It is not property held on trust for you, and it is not segregated.

English law has been unambiguous about this since Foley v Hill (1848) 2 HLC 28, decided in the House of Lords, which established that the relationship between banker and customer is one of debtor and creditor rather than trustee and beneficiary. Lord Cottenham LC:

Money, when paid into a bank, ceases altogether to be the money of the principal; it is then the money of the banker, who is bound to return an equivalent by paying a similar sum to that deposited with him when he is asked for it.

That is the legal statement of everything in this chapter, written 178 years ago and never seriously disturbed since. The practical consequence is that when a bank fails, depositors are creditors in an insolvency, which is precisely why a compensation scheme exists.

The same logic applies one tier up. The Bank of England’s own consolidated balance sheet at 31 March 2025 showed, on the sterling liabilities side, reserves balances of £694,512 million and notes in circulation of £91,230 million, against total liabilities of £861,868 million. A banknote is a liability of the issuer, recorded as such.

What the record physically is#

In a core banking system, an account is an entity with attributes — an identifier, a product type, a currency, a status, an owning customer, an interest and fee configuration — and a stream of postings against it.

A posting carries at minimum a value date, a booking date, an amount with an explicit currency, a direction (debit or credit), a reference to the transaction that produced it, and a link to the contra entry required by double entry. Postings are written once. Corrections are made by writing further postings, not by editing history: a mistaken credit is reversed with a compensating debit, and both remain visible. This is the same discipline that makes any well-built audit trail work, and it is the reason a bank statement is a list of movements rather than a series of snapshots.

The balance is a function of the postings. Real systems typically maintain running totals for performance — recomputing from the beginning of time on every enquiry would be absurd — but those totals are caches, and reconciling them back to the posting stream is a routine control. If the derived total and the sum of postings ever disagree, the postings are right by definition.

Several distinct balances are maintained or derived, and the vocabulary is not perfectly standardised across the industry:

Different institutions use these terms with slightly different boundaries, and international groups frequently use them inconsistently between systems. Where a specification matters — a reconciliation file, an API contract — the definition should be stated rather than assumed.

Two further properties of the record are worth naming now because they recur throughout the book. Amounts are held in the currency’s minor units as exact integers, never as binary floating point, for reasons taken up in detail in Volume I’s chapter on the money type. And every instruction that moves money needs an idempotency discipline, because a network that can deliver a message twice will eventually deliver a payment instruction twice, and the ledger must be able to recognise the second one as a duplicate rather than post it.

Two tiers of money#

The phrase “your money” conceals a distinction that governs almost every settlement rule later in this book.

Commercial bank money is a deposit liability of a commercial bank. It is what households and businesses hold and what the number in your app represents. Its value depends on the solvency of that particular bank, which is why deposit protection exists and why the identity of the bank matters.

Central bank money is a liability of the central bank. In the United Kingdom it takes two forms: reserve balances held by eligible institutions in accounts at the Bank of England, and banknotes. Reserves are the settlement asset for the UK’s high-value systems: they sit in the Bank’s Real-Time Gross Settlement infrastructure, and a CHAPS payment is settled by moving reserves between accounts there. Settlement in central bank money is final in a way that settlement in commercial bank money is not, because the Bank of England does not fail.

That two-tier structure — customers holding claims on banks, banks holding claims on the central bank — is the skeleton of the whole payments system. When you pay someone at another bank, your bank’s liability to you falls, their bank’s liability to them rises, and the imbalance between the two banks is squared up in central bank money. Chapter 6 takes this apart properly; it is introduced here because “a balance is a liability of a specific institution” is the fact that makes it necessary.

Why the bank can lend, and what actually constrains it#

Since bank lending creates deposits, the interesting question is not where the money comes from but what stops banks creating an unlimited amount of it. Four things, in practice.

Capital. Under the Basel III framework as implemented in the UK by the PRA, banks must hold regulatory capital against risk-weighted assets, plus buffers. Every new loan consumes capital. Capital is finite and expensive.

Liquidity. The Liquidity Coverage Ratio requires a bank to hold sufficient high-quality liquid assets to survive a 30-calendar-day stress scenario, with a minimum of 100%. The standard was phased in from 60% on 1 January 2015, rising ten percentage points a year to 100% on 1 January 2019. The LCR is the direct regulatory answer to the run mechanism described above: it is a rule about surviving a month of people wanting their money back.

Profitability and competition. A bank creates a deposit when it lends, but it cannot control where that deposit goes. The borrower spends it, and it very often ends up at another bank, which the lending bank must settle in central bank money. Lending therefore has a funding cost, and imprudent lending has a credit cost.

Monetary policy. Bank Rate sets the price at which reserves are remunerated and thereby the reference price of money throughout the system. The central bank influences the quantity of broad money by influencing its price, not by rationing a reserve base.

Structurally, the largest UK banking groups have also been required since 1 January 2019 to place their retail deposit-taking business inside a legally separate ring-fenced bank, insulated from investment banking activity.

Deposit protection: what the FSCS actually covers#

Precision matters here more than anywhere else in the chapter, because this is the topic on which confident misinformation is most common.

The limit. £120,000 per eligible depositor, per authorised firm, for firm failures occurring on or after 1 December 2025. This was set by the PRA in policy statement PS24/25 (November 2025), following consultation paper CP4/25 (March 2025), which had proposed £110,000. The previous limit of £85,000 had been in place since 2017.

Per licence, not per brand and not per account. The limit attaches to the authorised firm — the banking licence — not to the trading name on the card. The FSCS states the position directly: protection is “up to £120,000 per person per banking licence”. Two of its own worked examples: HSBC trades as HSBC and as first direct on a single licence, so the £120,000 covers the total across both; Nationwide Building Society’s licence also carries the Derbyshire, Cheshire and Dunfermline brands. A depositor holding £120,000 with each of two brands that share a licence is protected for £120,000 in total, not £240,000. The authoritative check is the Financial Services Register maintained by the FCA, not the marketing.

Joint accounts. Each account holder has their own entitlement, so a joint account is generally treated as split equally between the holders for the purposes of the limit.

Temporary high balances. Up to £1.4 million for six months, on top of the standard limit, where the money arises from a defined qualifying event. The FSCS lists these as including the sale of a main residence, property purchase or equity release, inheritance, insurance payouts, retirement benefits, redundancy, divorce or dissolution of a civil partnership, and compensation for unfair dismissal or wrongful conviction. Protection runs from the point the money becomes legally transferable to the depositor or is first credited. Compensation for personal injury, disability or incapacity is protected without limit.

Timing. The FSCS works to a seven-day payout for the bulk of depositors, achieved by requiring failed firms to produce a Single Customer View file — a standardised extract of depositor identity, contact details, account details and aggregate balances — within 24 hours of the failure event. Note a genuine inconsistency in the published wording: FSCS industry materials refer to compensation “within seven working days of a deposit taker failing”, while consumer-facing materials say customers “will typically get their money back within seven days”. The operating requirement placed on firms is what matters in practice: the SCV must be ready before anything goes wrong, which is why SCV testing is a standing supervisory expectation and not a crisis activity.

Disclosure lag. Firms were given until 31 May 2026 to update their customer-facing information sheets and exclusion lists to the new limit. Documents published before that date, including some still in circulation, may quote £85,000 while the rule in force is £120,000. If you are checking a bank’s own PDF rather than the FSCS site, check its date.

What is not covered. The FSCS deposit protection scheme covers deposits with PRA-authorised deposit takers. It does not cover funds held with electronic money institutions or payment institutions. The FSCS is unambiguous: “We can’t protect the money you have with e-money institutions and payment providers.”

This is the single most consequential misunderstanding in UK consumer fintech, because e-money accounts look exactly like bank accounts. They have sort codes and account numbers, they receive salaries, they issue cards. What protects the customer’s money there is a different regime: safeguarding under the Payment Services Regulations 2017 and the Electronic Money Regulations 2011, which requires the firm either to segregate relevant funds in a designated safeguarding account with a credit institution or to cover them with an insurance policy or comparable guarantee. Safeguarding is a real protection and it is not nothing. But it is a claim on segregated assets resolved through an insolvency process, not a seven-day payout from a statutory scheme, and the FSCS warns that money held this way “could be tied up for a while during the insolvency process”.

There is one important bridge between the two regimes. Following PRA policy statement PS2/23, with rule amendments in force from 12 March 2023, the FSCS can “look through” a safeguarding account and compensate the underlying customers of an e-money or payment institution if the bank holding the safeguarded funds fails. This does not extend protection to the case where the e-money institution itself fails and its bank does not. The distinction is narrow and it is exactly the sort of thing that gets flattened in marketing copy.

Covered deposits and bail-in. Deposits protected by the FSCS are legally excluded from the bail-in tool under the UK’s special resolution regime. When the Bank of England writes down or converts a failing bank’s liabilities, protected deposits are not in scope. This is the resolution-side counterpart of deposit insurance and it is worth knowing, because it means the answer to “could they take my savings to save the bank?” is no, within the protected limit.

The mechanics of a run, stated properly#

The formal treatment is Diamond and Dybvig, “Bank Runs, Deposit Insurance, and Liquidity”, Journal of Political Economy volume 91, number 3 (1983), pages 401–419 — work for which Douglas Diamond and Philip Dybvig shared the 2022 Sveriges Riksbank Prize in Economic Sciences with Ben Bernanke.

The model isolates three features that together generate the run.

Maturity transformation. The bank issues liabilities that are short and liquid — demand deposits — against assets that are long and illiquid — mortgages, loans, held-to-maturity securities. That transformation is the economic service the bank provides, and it is inherently fragile.

The sequential service constraint. Depositors are served in the order they arrive, and the bank pays in full until it cannot pay at all. Withdrawing early is therefore strictly better than withdrawing late if you believe others will withdraw.

Multiple equilibria. Given the first two, the model has more than one self-consistent outcome. If nobody expects a run, nobody runs and the bank is fine. If everybody expects a run, running is individually rational and the bank fails — even if its assets, held to maturity, would have been worth more than its liabilities. Deposit insurance works by eliminating the bad equilibrium: if you will be paid regardless, you have no reason to be first.

The critical distinction the model forces is between illiquidity and insolvency. A bank is insolvent when its assets are worth less than its liabilities. A bank is illiquid when it cannot convert good assets into cash fast enough to meet demands. A run turns the second into the first, because forced sales into a falling market crystallise losses that a patient holder would never have taken.

The three cases already cited illustrate the range.

The SVB UK case is the one practitioners should study, because it shows the modern default. The preferred outcome is not a payout at all. It is a resolution — a transfer of the failing bank’s business to a purchaser over a weekend, so that on Monday morning the deposits still exist, still work, and simply have a different owner. Compensation schemes are the backstop for when that cannot be arranged, not the first line.

Note also the SVB deposit composition: about 95% uninsured. Deposit insurance calms retail depositors. A bank funded overwhelmingly by corporate treasurers holding tens of millions each has almost no insured base to calm, which is why concentration and insured share are supervisory concerns and not just balance-sheet trivia.

Cash, precisely#

Physical currency is the exception, and being precise about it clarifies the rule.

Bank of England notes in circulation, from the Bank’s own banknote statistics measured at the end of February each year:

A separate line in the same statistics, £4,704 million of “other notes”, covers higher-denomination notes held as backing for the note issues of banks in Scotland and Northern Ireland — the internal high-value notes long known in the trade as giants and titans.

Legal tender is where public understanding is worst, and the Bank of England’s own explainer is blunt that the term “has a narrow technical meaning that will rarely come up in everyday life”. It concerns the discharge of a debt in court proceedings. In England and Wales, Royal Mint coins and Bank of England notes are legal tender. In Scotland and Northern Ireland, only Royal Mint coins are — banknotes issued by Scottish and Northern Irish banks are not legal tender anywhere, including Scotland and Northern Ireland, and circulate on acceptance rather than statute. Even the coins carry limits: 1p and 2p coins are legal tender only up to 20p, and 5p and 10p coins up to £5. No retailer is obliged to accept any particular means of payment, and debit cards, credit cards, cheques and contactless payments are not legal tender anywhere in the UK.

The Scottish and Northern Irish issuers operate under the Scottish and Northern Ireland Banknote Regulations 2009, supervised by the Bank of England. All such notes must be fully backed by ring-fenced assets at all times, with a minimum of 60% of those backing assets held in Bank of England notes or UK coin. A Scottish banknote is therefore a commercial bank liability with a statutory asset backing behind it — a genuinely different instrument from a Bank of England note, despite looking like the same kind of object in the same wallet.

What makes cash distinctive as a payment instrument is a short list of properties that the rest of this book spends thousands of words trying to reconstruct electronically. It is a bearer instrument, so possession is title. It settles finally at the moment of handover, with no clearing cycle and no settlement lag. It carries no counterparty risk beyond the issuer. It works with no connectivity, no power and no third party. It is irreversible: there is no chargeback on a twenty-pound note. And it leaves no record, which is simultaneously its greatest privacy virtue and the reason it attracts the attention of anti-money-laundering regimes.

Counterfeiting, incidentally, is not the systemic problem folklore suggests. In 2025 the Bank of England received about 200,000 counterfeit notes with a notional face value of £3.97 million, which it reports as fewer than one in 24,390 notes in circulation.

What this means if you are building something#

Four practical consequences follow from this chapter, and they show up in real systems constantly.

Model balances as derived from an immutable posting stream, not as a mutable field. Any design in which a balance is updated in place will eventually produce a number that cannot be explained, and an unexplainable balance in a financial system is a regulatory problem as well as an engineering one.

Be explicit about which balance you mean, everywhere — in APIs, in reports, in support scripts, in the screen label the customer reads. “Balance” alone is not a specification.

Know exactly what protects your customers’ money and never let the language drift. If you are an e-money institution, your customers’ funds are safeguarded, not FSCS-protected, and saying otherwise in a marketing email is a regulatory failure and not a wording preference. If you are a bank, know your licence, know which brands share it, and know that the limit is £120,000 per depositor per licence.

And carry the central idea into everything that follows. A payment is never the transportation of a substance. It is the coordinated amendment of at least two records, held by at least two institutions, each of which must end up agreeing about what happened. The next chapter is about the discipline that makes that agreement possible, and it is seven hundred years old.

2.98 Common wrong ideas#

1. Wrong: the bank keeps some of your deposit and lends the rest of it out. Right: bank lending creates deposits, because a new loan writes a new asset and a matching new liability, and the deposit did not exist a moment earlier.
2. Wrong: banks are limited by a reserve ratio that multiplies up a fixed base of central bank money. Right: the United Kingdom has no mandatory minimum reserve ratio, and the real constraints are capital, liquidity, profitability and monetary policy acting on the price of money.
3. Wrong: a shop has to take cash because cash is legal tender. Right: legal tender has a narrow technical meaning about discharging a debt in court proceedings, and no retailer is obliged to accept any particular means of payment.
4. Wrong: a balance is a number stored in a field and edited when money moves. Right: in a well-built system it is derived from an append-only sequence of immutable postings, which is what makes a ledger auditable and a statement a list of movements.
5. Wrong: the app showed two different figures, so the payment was taken twice. Right: the overwhelmingly common explanation is an authorisation hold that has not yet dropped off, seen by comparing an available balance against a ledger balance.
6. Wrong: a banknote is the one form of money that is nobody’s promise. Right: a Bank of England note is a liability of the Bank of England and is recorded as one; what distinguishes it is that the claim transfers by physical possession against an issuer nobody expects to fail.
7. Wrong: FSCS cover is per account, or per brand, so two accounts mean two limits. Right: it is £120,000 per eligible depositor per authorised firm, so brands sharing a licence — HSBC and first direct, for instance — share a single limit.
8. Wrong: money in an e-money or payment institution account is FSCS-protected, because it has a sort code, receives salaries and issues cards. Right: those funds are safeguarded under the Payment Services Regulations 2017 and the Electronic Money Regulations 2011, which is a claim on segregated assets resolved through insolvency rather than a seven-day statutory payout.
9. Wrong: a bank run looks like a queue outside a branch. Right: runs happen at the speed of the fastest funding channel, which at Northern Rock was the wholesale market in August 2007 and at Silicon Valley Bank was about $42 billion of wire transfers on 9 March 2023.
10. Wrong: when a bank fails, depositors get compensated. Right: the preferred outcome is resolution — Silicon Valley Bank UK was sold to HSBC on 13 March 2023 — and a compensation payout is the backstop for when a transfer cannot be arranged.

2.99 Chapter summary in 20 lines#

1. The number at the top of your banking app is not a description of where anything is; it is a description of a promise.
2. Handing £500 to a neighbour who seals it in a labelled envelope is safekeeping, and handing it to one who spends it and repays you later is deposit banking.
3. Every bank in the world is the second arrangement, and English law has said so since Foley v Hill in 1848.
4. Notes paid into a branch become the bank’s notes, and the only thing written beside your name is an obligation payable on demand.
5. Bank deposits are about 96 to 97 per cent of the money in the United Kingdom, against roughly £91.5 billion of Bank of England notes in existence anywhere in the world.
6. If every account holder asked for notes tomorrow the notes would not exist, not because of any wrongdoing but because notes were never the plan.
7. The textbook claim that a bank lends out your deposit is backwards, because the Bank of England’s own bulletin states that bank lending creates deposits.
8. Nothing multiplies up a fixed reserve base, and the binding constraints are capital, liquidity, funding cost and Bank Rate.
9. A balance is best modelled as a figure derived from an immutable posting stream rather than as a mutable field.
10. There is rarely only one balance: ledger, cleared, available and value-dated balances differ routinely, and the differences are not errors.
11. A banknote is a claim too, and what makes it special is that it transfers by possession rather than that it is not a promise.
12. Scottish and Northern Irish notes are commercial bank liabilities with statutory ring-fenced backing, and are not legal tender anywhere.
13. Legal tender is a narrow point about discharging debts in court and obliges nobody to accept anything in a shop.
14. A run follows from three features: liabilities that are short against assets that are long, service in order of arrival, and the rationality of going early.
15. Diamond and Dybvig showed that this produces more than one self-consistent outcome, and deposit insurance works by eliminating the bad one.
16. Northern Rock’s real run happened in the wholesale market in August 2007, and the photographed queues came after the announcement of support.
17. Silicon Valley Bank lost about $42 billion by wire in a day, with roughly 95 per cent of its deposits uninsured, which is what a run looks like when the branch is an API.
18. The FSCS covers £120,000 per eligible depositor per authorised firm from 1 December 2025, per licence rather than per brand, with temporary high balance cover of up to £1.4 million for six months.
19. E-money and payment institutions are outside the deposit scheme entirely, and their customers rely on safeguarding, which is a real protection but a slower and different one.
20. A payment is never the transportation of a substance; it is the coordinated amendment of at least two records held by at least two institutions that must end up agreeing.

Chapter sources: Bank of England — Quarterly Bulletin 2014 Q1, McLeay, Radia and Thomas, “Money creation in the modern economy”; “What is money” and “What is legal tender” explainers; banknote statistics (page updated 23 April 2026); consolidated balance sheet at 31 March 2025; Bank of England Levy pages; statement on Silicon Valley Bank, 13 March 2023; Scottish and Northern Ireland banknote regime documentation; Freedom of Information response on bail-in and customer deposits. Prudential Regulation Authority — CP4/25 and PS24/25, Depositor protection; PS2/23. Financial Services Compensation Scheme — deposit limit increase, banking licences, temporary high balances, Single Customer View, and e-money protection pages. Financial Conduct Authority — safeguarding requirements for payment and e-money institutions. Bank for International Settlements — Hyun Song Shin, “Reflections on Northern Rock”. Basel Committee on Banking Supervision — Liquidity Coverage Ratio executive summary. Foley v Hill (1848) 2 HLC 28. Diamond and Dybvig, Journal of Political Economy 91(3), 1983. US House Financial Services Committee background materials on the failure of Silicon Valley Bank. UK Parliament research briefing SN04478 on Northern Rock.