What a Bank Is

5.0 What this chapter gives you#

1. You will be able to explain why your bank balance is a debt the bank owes you rather than a pile of notes held somewhere with your name on it.
2. You will be able to draw a bank’s two lists and say which side a customer deposit sits on and which side a mortgage sits on, without hesitating.
3. You will be able to show, in two entries, how a bank makes a loan by typing a number into an account, and why that leaves its equity unchanged.
4. You will be able to say why the goldsmith-and-vault story and the money multiplier are wrong, and point to what the Bank of England published in 2014 instead.
5. You will be able to separate the constraints on the banking system as a whole from the constraints on one bank, and explain why a treasurer who says “we lend out deposits” is describing their job accurately.
6. You will be able to explain why a bank can report a capital ratio of fourteen per cent while the owners’ money is about five per cent of its balance sheet, and why both figures are true.
7. You will be able to say why capital did not save Silicon Valley Bank, and keep the defence against losses distinct from the defence against withdrawals.
8. You will be able to describe what happens to depositors, shareholders and AT1 holders when a British bank is resolved over a weekend, using SVB UK as the worked case.
9. You will be able to state what the FSCS protects, why “per banking licence” is not the same as “per brand”, and where an uncovered corporate depositor ranks in an insolvency.
10. You will be able to explain why a payment institution is not a bank, has neither the capital nor the liquidity regime, and must therefore safeguard its customers’ funds rather than lend them.

Ask a hundred people in a British high street what a bank is and roughly ninety-five will describe a building where money is kept. The remaining five will say something about apps. Both answers describe the furniture.

A bank is not a place where money is kept. Very nearly the opposite is true: a bank is an institution that has spent the money you gave it, is legally obliged to give it back the moment you ask, and stays alive by being extremely careful about the gap between those two facts. Everything difficult about banking lives in that gap. Every rule, every regulator, every capital ratio, every crisis of the last four hundred years.

This is not a criticism. If banks did nothing but hold money they would be warehouses, they would charge you rent, and no house in Britain would ever be bought with a mortgage. The whole point of the institution is that it takes money people want back at a moment’s notice and turns it into money other people can borrow for twenty-five years. That transformation is useful, dangerous, and impossible to perform without the possibility of failure. Banking regulation is an attempt to make failure rare, survivable and orderly, not to abolish it, because abolishing it would mean abolishing the bank.

Two earlier chapters carry the weight here. Chapter two established that your balance is not a substance but a record. Chapter three established that no record of money may be changed by writing down one fact; every change is two entries that balance. This chapter puts them together: if your balance is a record in somebody else’s book, whose book is it, what else is in it, and what happens when it stops adding up?

Along the way we will demolish the most widely taught and most thoroughly wrong story about banking: the goldsmith, the vault, and the fraction kept in reserve. It appears in school textbooks, in economics degrees, in newspaper explainers and in a great many bank employees’ understanding of their own employer. The Bank of England published a paper in 2014 saying, in effect, that it is not how anything works. We will get to that paper. First, the plain version.

The plain version#

Your money is somebody else’s promise#

Start with the thing in your hand. Open your banking app. It says, let us say, £1,842.60.

That number is not a picture of money sitting somewhere with your name on it. There is no drawer. There is no bundle of notes in a vault in Edinburgh labelled Ms A. Okonkwo. The number is a promise: the bank saying we owe you one thousand eight hundred and forty-two pounds and sixty pence, and we will hand it over, or send it wherever you tell us, whenever you ask.

Your bank balance is a debt. It is a debt the bank owes to you.

This feels alarming the first time you hear it, then stops feeling alarming when you notice it is exactly what you already believed. You have never once thought a specific pile of notes was yours. You have never asked which notes. What you care about is that when you tap your card for £3.20, the coffee shop gets £3.20 and your app shows £3.20 less. A promise that is always kept is indistinguishable, in daily life, from a thing you own.

The coat-check that lends the coats#

Here is the analogy, and it is a good one for about four paragraphs.

Imagine a cloakroom at a theatre. You hand over your coat, you get a ticket, and at the end of the evening you swap the ticket back for the coat. It would be outrageous if, halfway through the second act, the attendant lent your coat to somebody walking past outside. Your coat is a particular coat. You want that one back.

Money is not like a coat. Any ten-pound note is as good as any other. So a cloakroom that took in money instead of coats could lend it out during the evening and nobody would mind, provided it could hand a ten-pound note back to each ticket-holder at the end. Nobody checks the serial numbers. That is roughly the story people tell about the goldsmiths of seventeenth-century London: they took in gold for safekeeping, issued receipts, noticed that most of the gold sat there untouched, and started lending it.

Hold that picture for one more minute, because we are going to break it. It gets you to the first important idea: a bank has two lists, not one.

The two lists#

Every bank keeps two lists, and understanding a bank means understanding both at once.

The first is what the bank owes. Your £1,842.60 is on it. So is every other customer’s balance, plus money the bank has borrowed from other banks, plus bonds it has issued to investors.

The second is what the bank owns. Mortgages it has lent out and expects to be repaid. Business loans. Credit card balances people owe it. Government bonds it has bought. Actual cash. A pile of computers and some rather expensive branch leases.

In a healthy bank, the second list is bigger than the first. It has to be. The difference between them is the owners’ money — the shareholders’ stake, built from what they originally put in and from profits kept rather than paid out. It is the cushion. If the bank owns £100 of things and owes £95, the cushion is £5.

That cushion is not a fund sitting in an account somewhere. It is not money set aside. It is simply the gap: what would be left over for the owners if every asset were turned into cash and every debt paid. Its entire job is to absorb bad news. When a borrower does not pay, the first list stays exactly where it is — the bank still owes you every penny of your £1,842.60 — and the second list shrinks. The loss comes out of the owners’ cushion, not out of the depositors’ promises. The cushion is there precisely so that depositors never notice. A bank fails when the losses are big enough to eat the whole cushion and start eating the promises.

Building a small bank with real numbers#

Meet Priya, who is starting a very small, very simplified bank in a town called Little Sutton. The numbers are tiny and invented so they can be followed; everything structural about them is true of a real bank.

Priya puts in £10,000 of her own money. On day one the bank owns £10,000 of cash, owes nothing, and the owners’ cushion is £10,000. Then two hundred residents open accounts and pay in £90,000 between them.

Notice that the totals still match, and why. Priya received £90,000 of cash and simultaneously took on £90,000 of promises. Two entries, as chapter three insisted. The cushion did not change, because taking a deposit does not make a bank richer or poorer. It makes it bigger.

Now Priya lends. Mr Hollis wants £60,000 to expand his bakery. Here the coat-check analogy has to be retired. Priya does not hand him sixty thousand pounds in notes from the drawer. She opens him an account and types £60,000 into it. That is the loan.

Watch the lists. The bank now owns something new: Mr Hollis’s promise to repay £60,000, an asset, because it is money owed to the bank. And it owes something new: £60,000 sitting in Mr Hollis’s account, which he can spend, because it is money owed by the bank.

Read that table twice. The bank is now £60,000 bigger on both sides. Nothing came in the door. Nobody deposited anything. Priya created £60,000 of spendable money by typing it, and created £60,000 of debt owed to her at the same instant, and the two exactly cancel, so her cushion is still £10,000 and her books still balance.

This is not a trick and it is not fraud. It is what banking is. When the Bank of England explains this to the public it says, in plain words, that most money in the economy is created not by printing presses but by banks when they make loans, and that when a loan is repaid the money is deleted and ceases to exist. Around eight in every ten pounds in the British economy is money of this kind. Notes and coins are around three per cent of the total.

So the goldsmith story is backwards. Banks do not take in money and lend it out. They lend, and the lending is what puts the money there.

So what stops Priya typing numbers forever?#

Four things, and each is a whole industry.

First, the money walks out. Mr Hollis did not borrow £60,000 to look at it. He spends it — on ovens from a supplier who banks with Barclays, on a builder who banks with Santander. Real money must then move from Priya’s bank to those banks. She cannot type that; she has to have it, or borrow it. So while the banking system as a whole creates deposits by lending, any individual bank constantly leaks. Managing that leak is called funding, and it is why banks compete so hard for your savings account even though they do not need your savings in order to lend.

Second, some borrowers do not pay. If Mr Hollis’s bakery fails and Priya recovers only £35,000, she has lost £25,000. Her cushion was £10,000. It is gone, and then some, and her bank is bust — not because it ran out of cash, but because what it owns is worth less than what it owes. This is why regulators insist the cushion be large relative to how risky the lending is.

Third, everyone might want their money at once. Priya has £100,000 of cash and £150,000 of promises. If more than £100,000 of promises are called in on the same afternoon she cannot pay, even though she is perfectly solvent and Mr Hollis is making excellent sourdough. Being unable to pay today is fatal regardless of being rich tomorrow. This is why regulators separately insist that banks hold a stock of things that can be turned into cash immediately.

Fourth, somebody is watching. In the United Kingdom that is the Prudential Regulation Authority, part of the Bank of England. It sets how big the cushion must be, how much instantly-sellable stuff must be held, and what happens if either falls short. It can and does stop banks paying dividends, and in the last resort it can close them.

What a run looks like#

The third danger is the one that actually kills banks, and nobody believes it will happen until it is happening.

On Friday 14 September 2007, queues formed outside branches of Northern Rock across Britain. The bank had asked the Bank of England for emergency liquidity support two days earlier and the news had leaked. Those queues are commonly described as the first run on a British bank since Overend, Gurney and Company collapsed in 1866. Northern Rock was not obviously insolvent that morning. It was illiquid: it had funded long-dated mortgages with short-dated wholesale borrowing that had suddenly stopped being available. The Chancellor guaranteed the deposits the same day to stop the queues. The bank was taken into state ownership on 22 February 2008.

Sixteen years later the same thing happened at hundreds of times the speed, because the queue was now an app. Silicon Valley Bank’s customers attempted to withdraw more than forty billion dollars on 9 March 2023 and its management expected over a hundred billion more the following day. It was closed on 10 March. Its British subsidiary lost about a third of its deposits in two days. Nobody needed to stand in a queue.

The two words#

Accountants and bankers do not say “what it owns” and “what it owes”. They say assets and liabilities. The cushion is called equity, or capital.

And now the sentence beginners find upside down, which chapter three prepared you for: your deposit is the bank’s liability, and the bank’s loans are its assets. Your asset — the money you think of as yours — is the bank’s debt. The mortgage you think of as your debt is the bank’s asset. Every payment in the rest of this book is a movement between somebody’s assets and somebody’s liabilities, and getting the direction right is the difference between a payments system that reconciles and one that does not.

Where the plain version stops being true#

The plain version above is honest as far as it goes. Four things in it will actively mislead you if you carry them forward unamended.

Correction one: there is no fraction, and there is no reserve#

The version taught in most schools says banks must keep a fixed fraction of deposits — the classic figure is ten per cent — and may lend the rest, and that this fraction, applied repeatedly as money is redeposited, “multiplies” an injection of central bank money into a larger stock of deposits. The ratio is the reserve requirement; the mechanism is the money multiplier.

Two separate things are wrong with this.

The first is empirical. The United Kingdom has had no mandatory reserve ratio since 1981. The number is not ten per cent; there is no number. The Bank of England does require banks above a size threshold to place non-interest-bearing cash ratio deposits with it, but that is a levy funding the Bank’s policy functions, not a liquidity rule. The United States abolished its requirements outright: the Federal Reserve Board reduced all reserve requirement ratios to zero per cent effective 26 March 2020. The euro area has kept a requirement, but it is one per cent of a defined deposit base, unchanged since 18 January 2012, and since the maintenance period beginning 20 September 2023 the European Central Bank has remunerated those reserves at zero per cent. A one per cent ratio is not a binding constraint on lending; it is a small tax.

The second is causal, and matters more. The multiplier story has the arrow pointing the wrong way. It says central bank money comes first and deposits are built on top. In practice the central bank supplies reserves on demand to keep the policy rate where it has decided it should be. If banks collectively need more reserves, they get more reserves, at the policy rate. The quantity of reserves is an outcome, not a constraint.

Correction two: “banks lend out your savings” is wrong, but so is “banks can lend without limit”#

Having demolished the vault, it is easy to overcorrect into an equally false position: that because lending creates deposits, a bank is unconstrained and money is conjured at will. The distinction that resolves it is between the system and the firm.

For the banking system as a whole, lending creates deposits and no prior pot of savings is required. That is the Bank of England’s point, and it is correct.

For an individual bank, the picture is much tighter, because the deposits it creates walk straight out to other banks. Priya’s £60,000 becomes Barclays’ problem within a week, and when it leaves she must settle in central bank reserves, which she cannot create. If she persistently lends more than she attracts, she must buy funding in the wholesale market, and that funding has a price and, crucially, an availability that can vanish overnight. Northern Rock died of exactly this. So a bank treasurer who tells you “we lend out deposits” is describing their daily job accurately even though the sentence is wrong as monetary economics. Practitioners and economists talk past each other about this constantly.

A further constraint the plain version understates: capital is scarce and expensive. Every new loan consumes capital, because the regulator requires the cushion to scale with the risk taken. A bank wanting to lend another billion pounds must find the capital to support it, and capital comes from retained profits or shareholders who want a return. This, not any reserve ratio, is the operative brake on lending in a modern bank.

Correction three: capital is not cash, and capital ratios do not measure survival#

The plain version described the cushion as “what would be left over”. That is right, but three things about it routinely mislead.

Capital is not a pot of money. You cannot spend it, hold it or point at it. It is an accounting difference between two other numbers. When a regulator says a bank must hold more capital, it is not saying “set aside more cash”; it is saying “fund yourself with more shareholder money and less borrowed money”, which changes nothing on the asset side.

Capital ratios are measured against risk-weighted assets, not actual assets. A mortgage at fifty per cent loan-to-value and an unsecured loan to a start-up both sit on the balance sheet at face value, but the second absorbs far more capital. This is why a bank can report a capital ratio of fourteen per cent while its equity is about five per cent of its actual balance sheet. Both numbers are true and answer different questions. Confusing them is the commonest error in commentary about banks.

And most importantly: capital does not stop a run. Capital protects against losses; liquidity protects against withdrawals. They are separate defences against separate deaths, and the second kills faster. Silicon Valley Bank’s capital ratios were above the required minimums when it failed. It failed because its depositors — approximately ninety-four per cent of whom were uninsured, and who knew each other — asked for their money at once, and selling assets fast enough to meet them would have crystallised losses that then did exceed the capital. A bank can be killed by either.

Correction four: the balance sheet is not the whole bank#

The two-list picture is complete only in the sense that a photograph is complete. It omits three things that matter.

It omits time. Both lists have maturities. A bank’s assets are typically long — a twenty-five-year mortgage — and its liabilities short, or instant. That mismatch is maturity transformation, and it is not a flaw in the business model, it is the business model. It is also why banks are fragile in a way a manufacturer is not.

It omits what is not on it. Undrawn credit facilities, loan commitments, guarantees and letters of credit produce no balance-sheet entry until somebody draws on them — which they do precisely when things are going badly. Regulators handle this with credit conversion factors, a technical way of saying “assume some of it becomes real”.

And it omits valuation. “What the bank owns” is an estimate, governed by accounting standards, involving forecasts of who will default. Under IFRS 9 a bank must book expected credit losses before anything has gone wrong. Two honest banks with identical loan books can report different asset values. The balance sheet is a considered opinion presented in the grammar of fact.

The technical version#

The identity, and a real balance sheet#

Everything above is a restatement of the accounting identity introduced in chapter three:

Assets = Liabilities + Equity

For a bank, the composition of those three terms is unusual in a way no other industry shares: its principal liability is money, and its principal asset is credit. Here is a real one — Lloyds Banking Group plc, statutory consolidated position at 31 December 2025, as reported in the group’s 2025 results announcement of 28 January 2026.

Total equity of £47.9 billion against total assets of £944.1 billion is approximately 5.1 per cent — the honest, unweighted answer to “how much of this bank is the owners’ money”.

On the liability side, customer deposits were £496.5 billion — retail current accounts £102.8 billion, retail savings £212.5 billion, Wealth £9.9 billion, Commercial Banking £171.1 billion. Wholesale funding (money markets, debt securities in issue, covered bonds, securitisations, subordinated debt) was £99.4 billion, of which £37.0 billion had a residual maturity under one year. Also here sit derivative liabilities, liabilities arising from insurance and investment contracts — Lloyds owns Scottish Widows — and repurchase agreements.

On the asset side, underlying loans and advances to customers were £481.1 billion: UK mortgages £323.1 billion, credit cards £17.3 billion, UK retail unsecured loans £10.5 billion, UK Motor Finance £16.4 billion, Business and Commercial Banking £28.3 billion, Corporate and Institutional Banking £62.0 billion. Financial assets at amortised cost totalled £554 billion; at fair value through profit or loss £240 billion; at fair value through other comprehensive income £36 billion; derivative financial assets £20 billion.

The loan-to-deposit ratio was 97 per cent — the clearest single statement of what this institution is. It has deliberately turned very nearly every pound the public entrusted to it into somebody else’s long-dated debt. The expected credit loss allowance of £3,353 million, 0.7 per cent of loans and advances, is the accounting recognition of correction four: the asset side is net of a forecast.

Money creation: the 2014 Quarterly Bulletin#

The definitive modern statement is Money creation in the modern economy, by Michael McLeay, Amar Radia and Ryland Thomas of the Bank of England’s Monetary Analysis Directorate, in the Bank of England Quarterly Bulletin 2014 Q1, published 14 March 2014. Anyone working in payments should read it once. Its central claims, in its own words:

“In the modern economy, most money takes the form of bank deposits. But how those bank deposits are created is often misunderstood: the principal way is through commercial banks making loans.”

“Whenever a bank makes a loan, it simultaneously creates a matching deposit in the borrower’s bank account, thereby creating new money.”

“Rather than banks lending out deposits that are placed with them, the act of lending creates deposits — the reverse of the sequence typically described in textbooks.”

And, explicitly refuting both traditional models: banks “do not act simply as intermediaries, lending out deposits that savers place with them, and nor do they ‘multiply up’ central bank money to create new loans and deposits.”

Mechanically, at the moment of origination, the lending bank posts:

Two entries. Balance sheet grows on both sides. Equity unchanged. Nothing has been taken from anyone. On repayment the entries reverse and the deposit ceases to exist — money destruction, which is why aggregate deposits fall when credit contracts even though nobody has withdrawn anything.

The Bank’s public explainer gives the composition of UK money as approximately 3 per cent notes and coin, 18 per cent central bank reserves and 79 per cent commercial bank deposits. Reserves and deposits are different instruments circulating in different closed circuits, which is the subject of the next chapter.

The article is equally clear about what does constrain creation. Banks must lend profitably in a competitive market. They must manage the credit and liquidity risk new lending creates. Prudential regulation constrains their activities in the interest of system resilience. And “monetary policy acts as the ultimate limit on money creation” — in normal times through the price of reserves, in unconventional times through asset purchases.

The Basel framework in outline#

The rules implementing those constraints are, in outline, international. The Basel Committee on Banking Supervision, hosted at the Bank for International Settlements, comprises 45 members from 28 jurisdictions, consisting of central banks and authorities with formal responsibility for banking supervision. Its standards have no legal force whatsoever. They are agreements between supervisors, which member jurisdictions then implement — or do not, or implement late, or differently — through their own law. In the United Kingdom the binding instrument is the PRA Rulebook, made under the Financial Services and Markets Act 2000 as amended. Anyone who says a British bank is “required by Basel” to do something is speaking loosely; it is required by the PRA, which has chosen to implement Basel.

Capital: the stack#

Regulatory capital is a hierarchy ordered by how readily it absorbs losses.

Common Equity Tier 1 (CET1) is ordinary shares, share premium and retained earnings, less regulatory deductions — goodwill and other intangibles, deferred tax assets relying on future profitability, prudent valuation adjustments, the excess of expected losses over accounting provisions, defined benefit pension surpluses and significant investments in financial sector entities. This is the loss-absorbing core.

Additional Tier 1 (AT1) is perpetual, deeply subordinated instruments with no maturity date and fully discretionary coupons, which convert to equity or are written down when CET1 falls through a contractual trigger — the instruments the press calls “cocos”. Tier 2 is dated subordinated debt, absorbing losses only in resolution or liquidation.

Lloyds’ regulatory capital position at 31 December 2025 makes the structure concrete:

Risk weighting, and why £944 billion becomes £235 billion#

Note the gap between total assets of £944.1 billion and risk-weighted assets of £235.5 billion. Two effects produce it.

First, risk weighting proper. Each exposure is multiplied by a weight reflecting its assessed riskiness, either from the standardised approach’s look-up tables or from the bank’s own internal ratings-based models where the PRA has granted permission. A low loan-to-value residential mortgage attracts a small weight; an unsecured corporate exposure a large one. Cash and central bank reserves attract zero.

Second, scope of consolidation. The regulatory consolidation is not the accounting consolidation. Lloyds’ insurance business is not consolidated for capital purposes; instead the group deducts £4,708 million of significant investments from CET1. The same effect explains why the leverage exposure measure is smaller than accounting total assets.

Because two banks with identical books could report very different RWAs, the Basel III finalisation package of December 2017 imposed an output floor: RWAs must be at least 72.5 per cent of what the standardised approaches would produce, phasing in from 50 per cent over five years. It also barred the advanced IRB approach for large corporates, banks and equities, imposed input floors (probability of default at five basis points for corporates, loss-given-default at 25 per cent for unsecured corporate exposures), and removed internal models from operational risk and CVA entirely. In the United Kingdom this package is called Basel 3.1; the PRA announced on 17 January 2025 a one-year delay to 1 January 2027, with full implementation still set for 1 January 2030.

The required ratios#

Basel III’s Pillar 1 minimums are 4.5 per cent CET1, 6 per cent Tier 1 and 8 per cent total capital, all as percentages of risk-weighted assets. On top sit buffers, which are not minimums in the same sense: breaching a buffer does not make a bank unlawful, it triggers automatic restrictions on dividends, buybacks and discretionary bonuses through the maximum distributable amount mechanism.

The countercyclical buffer is the interesting instrument: the Financial Policy Committee raises it when it judges risk to be building and cuts it in stress, so that capital accumulated in good years can be released rather than forcing banks to shrink lending exactly when the economy needs it.

The leverage ratio#

Because risk weighting can be gamed and models can be wrong, Basel III added a deliberately crude backstop: Tier 1 capital divided by a total exposure measure that includes on-balance-sheet assets, derivative exposures, securities financing transactions and off-balance-sheet items converted at prescribed factors, with no risk weighting at all.

The UK framework, set by Financial Policy Committee Direction and implemented in the Leverage Ratio (Capital Requirements and Buffers) Part of the PRA Rulebook, imposes a minimum of 3.25 per cent of Tier 1 capital, plus a countercyclical leverage ratio buffer and, for systemic firms, an additional leverage ratio buffer. Following PRA Policy Statement 22/25 of 12 November 2025, the retail deposits threshold for application of the requirement rose from £50 billion to £75 billion with effect from 1 January 2026, calculated on a three-year moving average rather than a point-in-time value; the separate threshold of £10 billion of non-UK assets was left unchanged. Firms below the thresholds face a supervisory expectation, in Supervisory Statement 45/15, that their leverage ratio should not ordinarily fall below 3.25 per cent. Lloyds reported a UK leverage ratio of 5.4 per cent at 31 December 2025.

Liquidity#

Capital answers “can this bank absorb losses”. Liquidity answers “can this bank pay today”. Basel III introduced two standards.

The Liquidity Coverage Ratio requires the stock of high-quality liquid assets to be at least 100 per cent of total net cash outflows over a 30-calendar-day stress scenario. HQLA are assets convertible into cash quickly, by sale or by being pledged as collateral, without significant loss of value. Level 1 assets — cash, central bank reserves, certain sovereign debt — count without limit or haircut. Level 2A assets take a 15 per cent haircut. Level 2 assets in aggregate may not exceed 40 per cent of the stock, and Level 2B may not exceed 15 per cent. The standard took effect on 1 January 2015 at 60 per cent, rising ten percentage points a year to 100 per cent on 1 January 2019.

The Net Stable Funding Ratio works over a one-year horizon: available stable funding divided by required stable funding, minimum 100 per cent. ASF weights liabilities and capital by how likely they are to stay put — long-dated funding and stable retail deposits score highly, overnight wholesale funding zero. RSF weights assets by illiquidity and by how long they must be funded for. It became a minimum standard for internationally active banks on 1 January 2018.

Lloyds reported an LCR of 145 per cent at 31 December 2025 on LCR-eligible assets of £131.4 billion, and an NSFR of 124 per cent. From those figures the modelled 30-day net outflow can be inferred at roughly £91 billion — about 18 per cent of its customer deposit book. Hold that against Silicon Valley Bank UK, which lost about 32 per cent of its deposits in two days. That is the honest limit of the LCR: it is calibrated to a severe scenario, not the worst imaginable one, and a modern digital run can outrun it.

Why a bank failing is not like a company failing#

A car parts manufacturer that becomes insolvent enters administration. An officeholder is appointed, trading may continue, assets are realised, creditors queue in a statutory order, and the process takes months or years. Nobody outside the supply chain much notices on the day.

A bank cannot be treated this way, for one reason: its liabilities are the money supply and the payments system. Freezing a bank’s balance sheet does not merely disappoint creditors, it deletes the current accounts from which salaries are paid, direct debits collected and card transactions settled. It destroys money and severs access to the payment rails simultaneously, for people who never chose to be creditors of anything.

Britain’s answer is the Special Resolution Regime in the Banking Act 2009, administered by the Bank of England as resolution authority, which sits alongside and largely displaces ordinary insolvency law for banks, building societies and certain investment firms. It provides five stabilisation options:

1. Bail-in — write-down of the claims of the bank’s unsecured creditors, including holders of capital instruments, with conversion to equity, recapitalising the firm in place.
2. Transfer to a private sector purchaser, executed without the consent of shareholders or counterparties.
3. Transfer to a bridge bank controlled by the Bank of England pending an onward sale.
4. Transfer to an asset management vehicle to run down a problem portfolio.
5. Temporary public ownership, exercised by HM Treasury as a last resort.

Where resolution is not justified — typically a small firm whose failure poses no systemic threat — the bank insolvency procedure applies instead: modified insolvency, with a liquidator holding two objectives in strict priority. First, work with the Financial Services Compensation Scheme so that protected depositors are paid out within seven days or their accounts transferred elsewhere. Only then, wind the firm up in the general creditors’ interest.

Depositor preference#

Insolvency law itself has been rewritten to put depositors near the front of the queue. Under Schedule 6 of the Insolvency Act 1986 as amended by the Banks and Building Societies (Depositor Preference and Priorities) Order 2014, paragraph 15B makes the portion of an eligible deposit not exceeding FSCS compensation an ordinary preferential debt — in practice the FSCS pays the depositor and stands in their place, which is why it is sometimes described as super-preferred. Paragraph 15BA makes the portion of an eligible deposit exceeding the compensation limit, owed to eligible persons — broadly individuals and micro, small and medium-sized enterprises — a secondary preferential debt. Paragraph 15BB extends secondary preference to deposits made through non-UK branches which would have been eligible had they been made in the UK.

Secondary preferential debts rank after ordinary preferential debts, but both rank ahead of floating charge holders and ordinary unsecured creditors. A large corporate depositor, by contrast, is an ordinary unsecured creditor and ranks with the bondholders.

FSCS cover#

The Financial Services Compensation Scheme protects eligible deposits up to £120,000 per eligible person per banking licence, a limit that took effect on 1 December 2025, having previously been £85,000. Joint accounts are covered to £240,000. Qualifying temporary high balances — house sale proceeds, redundancy payments, insurance settlements — are protected up to £1.4 million for six months from the date first deposited.

The phrase “per banking licence” is doing real work and is routinely misread. Several familiar high-street brands may share a single authorisation, in which case the limit applies once across all of them, not once per brand.

MREL and loss-absorbing capacity#

For resolution to work without public money, there must be enough liabilities available to be written down. That is the minimum requirement for own funds and eligible liabilities: own funds plus eligible liabilities, positioned appropriately in the creditor hierarchy, sufficient to absorb losses and recapitalise the firm. Lloyds reported total MREL resources of £75,732 million at 31 December 2025, an MREL ratio of 32.2 per cent of risk-weighted assets — well over twice its total capital ratio, because MREL includes £31,232 million of other eligible liabilities issued by the holding company.

Resolution carries a statutory safeguard: no creditor worse off. Following a bail-in or partial transfer, no pre-resolution shareholder or creditor may be left worse off than in an ordinary insolvency. An independent valuation determines this after the event, and compensation is payable if the test is failed.

The regime in action: SVB UK#

The clearest worked example in British experience is the resolution of Silicon Valley Bank UK Limited, documented by the Bank of England in its report under section 79A of the Banking Act 2009.

SVB UK operated as a branch from 2012 and converted to a UK subsidiary in July 2022. Its parent was closed by the California Department of Financial Protection and Innovation on 10 March 2023. At 11pm that night the Bank of England announced its intention to place SVB UK into the bank insolvency procedure. Over the weekend it changed course, determining that a private sector purchaser transfer better served the special resolution objectives. At 7am on Monday 13 March 2023 the Bank exercised its transfer powers and moved the whole of SVB UK’s share capital to HSBC UK Bank plc.

The AT1 and Tier 2 instruments were mandatorily written down as required by section 6 of the Act, wiping out £355 million of investors’ claims. The equity was transferred for one pound. Every depositor — covered and uncovered alike — retained uninterrupted access to their accounts. No UK public funds were used. Both independent valuations concluded that FSCS-covered depositors would have recovered 100 per cent of their claim in a counterfactual insolvency, so the no-creditor-worse-off test was comfortably satisfied for them.

Read that table as a summary of the chapter. A bank with £8.8 billion of assets and £1.4 billion of net assets — an owners’ cushion of over 16 per cent of its balance sheet, far above any regulatory minimum — was resolved over a weekend because roughly a third of its deposits left in two days. Solvency did not save it. What saved its depositors was a statutory regime built on the assumption that this would happen to somebody, eventually.

Ring-fencing#

One further piece of British structural regulation belongs here. The Financial Services (Banking Reform) Act 2013, implementing the recommendations of the Independent Commission on Banking, requires the largest UK banking groups to separate core retail banking — deposit-taking from individuals and small businesses, and the payment services attached to it — from investment banking activity, into a legally distinct, separately capitalised, separately governed ring-fenced body. The regime came into force on 1 January 2019, by which point approximately £1.2 trillion of core deposits had been placed inside ring-fenced entities. The threshold, set by the Financial Services and Markets Act 2000 (Ring-fenced Bodies and Core Activities) Order 2014, is now £35 billion of core deposits, above which a group conducting material investment banking activity must ring-fence, subject to a trading assets condition allowing exit where trading assets are less than 10 per cent of Tier 1 capital. Reforms described by the government as “smarter ring-fencing” came into force on 4 February 2025.

The practical consequence for anyone building payment systems is that “Barclays” and “HSBC” are not single legal entities. A group may contain a ring-fenced bank, a non-ring-fenced bank, an overseas subsidiary and a service company, each with its own authorisation, balance sheet and, in some cases, scheme participation. Which entity a payment is with has a real answer and real consequences.

What to carry forward#

Your balance is the bank’s liability and the bank’s loans are its assets. Every payment message in Volumes III and IV is an instruction to move a liability from one institution’s book to another’s, and the direction of every posting depends on remembering whose book you are in.

Lending creates deposits; deposits are not lent out. This is why the money supply expands and contracts with credit, and why “where did the money come from” is usually the wrong question to ask about a payment.

Capital protects against loss and liquidity against withdrawal, and they are entirely different defences. A payments business — an authorised payment institution, an e-money firm, a fintech — is not a bank and has neither regime, which is precisely why its customers’ funds must be safeguarded in a segregated account rather than lent.

And commercial bank money is not the only kind. When Priya’s £60,000 walks out of Little Sutton to Barclays, something else has to move — something no commercial bank can create. That something is central bank money, and it is the subject of the next chapter.

5.98 Common wrong ideas#

Wrong: A bank keeps your money in a vault and lends out a fraction of it. Right: A bank has already spent what you gave it; your balance is a promise, and lending creates new deposits rather than passing on old ones.

Wrong: British banks must hold ten per cent of deposits in reserve. Right: The United Kingdom has had no mandatory reserve ratio since 1981, and the Federal Reserve reduced all its ratios to zero per cent on 26 March 2020.

Wrong: Because lending creates deposits, a bank can lend without limit. Right: The deposits an individual bank creates walk out to other banks and must be settled in reserves it cannot create, and capital, funding and the regulator bind long before anything else does.

Wrong: Capital is a pot of money the bank sets aside for a rainy day. Right: Capital is the accounting difference between two other numbers, and requiring more of it means funding yourself with more shareholder money and less borrowed money.

Wrong: A well-capitalised bank cannot fail. Right: Capital absorbs losses and liquidity meets withdrawals; Silicon Valley Bank’s capital ratios were above the required minimums on the day it was closed.

Wrong: A capital ratio of fourteen per cent means fourteen per cent of the balance sheet is the owners’ money. Right: Ratios are struck against risk-weighted assets, which is why Lloyds could report 14.0 per cent while £47.9 billion of equity against £944.1 billion of assets is about 5.1 per cent.

Wrong: Basel rules bind British banks directly. Right: Basel standards have no legal force whatsoever; what binds a UK firm is the PRA Rulebook, made under the Financial Services and Markets Act 2000.

Wrong: FSCS cover of £120,000 applies to each of your bank’s brands. Right: It applies per eligible person per banking licence, and several familiar high-street brands may share a single authorisation.

Wrong: A failing bank goes into administration like any other company. Right: Its liabilities are the money supply and the payments system, so the Special Resolution Regime in the Banking Act 2009 largely displaces ordinary insolvency law.

Wrong: The balance sheet tells you what a bank owns. Right: It is a considered opinion presented in the grammar of fact, net of IFRS 9 expected credit losses, silent on undrawn commitments and silent on the maturity mismatch that makes the business fragile.

5.99 Chapter summary in 20 lines#

1. A bank is not a place where money is kept but an institution that has spent what you gave it and is obliged to return it the moment you ask.
2. Your balance is a record of a debt the bank owes you, not a claim on any particular notes.
3. Every bank keeps two lists: what it owns, called assets, and what it owes, called liabilities.
4. The difference between them is equity, the owners’ cushion, and its whole job is to absorb losses so that depositors never notice.
5. Taking a deposit makes a bank bigger rather than richer, because the cash coming in and the promise going out are two entries of the same size.
6. When a bank lends it types a number into the borrower’s account, creating an asset and a liability at the same instant and leaving its equity untouched.
7. Lending therefore creates deposits and repayment destroys them, which is the reverse of the sequence taught in most textbooks.
8. The Bank of England said exactly that in Money creation in the modern economy in the 2014 Q1 Quarterly Bulletin.
9. Around seventy-nine per cent of UK money is commercial bank deposits, eighteen per cent central bank reserves and about three per cent notes and coin.
10. There is no reserve ratio in the United Kingdom and no money multiplier, because reserves are supplied on demand to keep the policy rate where the central bank wants it.
11. An individual bank is nonetheless tightly constrained, since the deposits it creates leave for other banks and must be settled in reserves it cannot create.
12. Capital is scarce and expensive and must scale with the risk taken, and that, not any reserve ratio, is the operative brake on lending.
13. Capital is not cash but an accounting difference, and capital ratios are measured against risk-weighted assets rather than actual ones.
14. That is why Lloyds reported a CET1 ratio of 14.0 per cent against £235.5 billion of risk-weighted assets while its equity was about 5.1 per cent of £944.1 billion of actual assets.
15. Capital protects against loss and liquidity protects against withdrawal, and they are separate defences against separate deaths.
16. Northern Rock in 2007 and Silicon Valley Bank in 2023 both died of the second, the latter’s British subsidiary losing about a third of its deposits in two days.
17. A bank cannot be wound up like a car parts manufacturer, because its liabilities are the money supply and the payment rails at the same time.
18. Britain’s answer is the Special Resolution Regime of the Banking Act 2009, offering bail-in, private sector transfer, a bridge bank, an asset management vehicle and temporary public ownership.
19. SVB UK showed the regime working: £355 million of AT1 and Tier 2 written down, the share capital transferred to HSBC for one pound, and every depositor’s access unbroken with no public funds used.
20. Carry forward that your deposit is the bank’s liability and its loans are its assets, and that when money leaves one bank for another something no commercial bank can create has to move.

Chapter sources: McLeay, Radia and Thomas, “Money creation in the modern economy”, Bank of England Quarterly Bulletin 2014 Q1; Bank of England explainer “How is money created?”; Bank of England, “The Bank of England’s approach to resolution” (2023), the section 79A report on the transfer of SVB UK to HSBC UK, the ring-fencing pages and CP10/26, and the countercyclical capital buffer pages; PRA CP2/25 and PS22/25 on the leverage ratio retail deposits threshold, PS1/26 on Basel 3.1 final rules, and the news release of 17 January 2025 on the Basel 3.1 delay; Basel Committee on Banking Supervision membership pages and the December 2017 “Basel III: Finalising post-crisis reforms” high-level summary; Financial Stability Institute executive summaries of the LCR and NSFR; Lloyds Banking Group plc 2025 results announcement, 28 January 2026; Federal Reserve Board reserve requirements page and its 28 April 2023 review of the supervision and regulation of Silicon Valley Bank; Deutsche Bundesbank and ECB on minimum reserves, including the Governing Council decision of 27 July 2023; Insolvency Act 1986 Schedule 6 as amended and the Banks and Building Societies (Depositor Preference and Priorities) Order 2014, legislation.gov.uk; Financial Services Compensation Scheme cover pages; Banking Act 2009; Financial Services (Banking Reform) Act 2013.