The Banking
System

A bank is not, as many people implicitly believe, a warehouse for money, a place where savers deposit funds that are then lent out to borrowers, with the bank acting as an intermediary. This intuitive model is called the "financial intermediation" model, and it is wrong in a precise and consequential way. Understanding exactly how it is wrong is the entry point to understanding how the modern monetary system works.

A bank is a balance-sheet machine. On one side of its balance sheet are its assets: loans it has made, securities it holds, reserves it keeps at the central bank. On the other side are its liabilities: deposits owed to customers, bonds it has issued, short-term borrowings from other banks. The bank's fundamental business is to borrow short and lend long, to accept short-term liabilities (deposits that can be withdrawn on demand, or at short notice) and deploy them as long-term assets (mortgages that may run for thirty years, business loans that may run for ten).

This structural mismatch (short liabilities, long assets) is the source of both the banking system's economic value and its inherent fragility. It is valuable because it transforms the maturity preferences of savers (who want liquidity) and borrowers (who want certainty of long-term funding) into a mutually beneficial arrangement. It is fragile because the bank can never, at any given moment, liquidate all its long-term assets quickly enough to repay all its short-term liabilities simultaneously. A bank that must repay all depositors at once will always be insolvent, not because it is badly run, but because that is what a bank is.

The Balance Sheet as the Unit of Analysis

The Bank of England's 2014 paper "Money Creation in the Modern Economy" (among the most important and most ignored documents in public understanding of finance) stated with unusual directness something that economics textbooks routinely obfuscate: "Commercial banks create money, in the form of bank deposits, by making new loans. When a bank makes a loan, for example to someone taking out a mortgage to buy a house, it does not typically do so by giving them thousands of pounds worth of banknotes. Instead, it credits their bank account with a bank deposit of the size of the mortgage."

This is the reality that the textbook model conceals. The bank does not take in deposits and then lend them out. It makes loans and, in doing so, creates deposits. The deposit is not the source of the loan, the loan is the source of the deposit. Understanding this reversal is the most important conceptual shift in monetary literacy.

Theory is best grounded in concrete mechanics. Let us trace, step by step, the accounting entries that occur when a bank makes a loan, entries that collectively demonstrate how money is created, and why the intuitive model of banking gets this precisely backwards.

Step 1: A Customer Deposits £1,000 Cash

A customer walks into Hypothetical Bank and deposits £1,000 in cash. The bank's balance sheet changes as follows:

Step 2: The Bank Makes a £900 Loan (Holding 10% as Reserve)

The bank decides to lend £900 to a borrower, retaining £100 (10%) as a reserve. In the textbook model, this means the bank hands £900 of the deposited cash to the borrower. In reality, what happens is this:

Let that settle. The original depositor still has £1,000 in their account. The new borrower has £900 in their account. There is now £1,900 of deposits in the economy where before there was £1,000 of cash. The bank holds £1,000 of cash and £900 of loans on its asset side, and owes £1,900 of deposits on its liability side. £900 has been created from nothing except a balance-sheet entry.

Step 3: The Borrower Spends and the Deposit Migrates

The borrower spends their £900, say, to pay a tradesperson, who banks at a different institution (Rival Bank). The £900 deposit moves from Hypothetical Bank to Rival Bank, and settlement is required between the two banks. This settlement occurs in central bank reserves, the interbank settlement asset that sits at the apex of the monetary hierarchy. Hypothetical Bank must transfer £900 of central bank reserves to Rival Bank.

Notice what this reveals about the reserve constraint: the bank needed central bank reserves not to make the loan, but to settle the resulting payment when the borrower spent the proceeds. The reserves are required for settlement between banks, not for the act of credit creation itself. This is the key insight that the money multiplier model misses.

Step 4: The Process Continues: Money Multiplies

Rival Bank now holds £900 of fresh reserves. Applying the same 10% reserve ratio, it can extend an £810 loan, creating £810 of new deposits. The recipient of that loan spends it, creating reserves at a third bank. And so the process continues.

The money multiplier is the textbook mechanism by which a given quantity of base money supports a much larger quantity of broad money. It is both genuinely important and widely misunderstood, particularly in respect of the direction of causation.

The Multiplier in Action: Step by Step

Why the Textbook Multiplier Overstates Central Bank Control

The money multiplier model implies that the central bank, by controlling the quantity of base money (the "monetary base"), can precisely control the broad money supply. Expand the monetary base and the broad money supply expands by a predictable multiple. This is the theoretical foundation of Friedman's monetarism. It does not describe how modern banking systems actually work.

Three observations break the model. First, reserve requirements in many advanced economies are zero or functionally irrelevant. The Federal Reserve reduced reserve requirements to zero in March 2020. The Bank of England has had no formal reserve requirement for decades. The Reserve Bank of Australia has no reserve requirement. If the textbook multiplier model were correct, these regulatory changes would have allowed the money supply to expand to infinity. It did not.

Second, the direction of causation runs from loans to reserves, not reserves to loans. When a bank makes a loan, it creates a deposit simultaneously. The resulting payment, when it settles between banks, creates a demand for central bank reserves to facilitate settlement. The central bank, in practice, supplies these reserves to prevent the settlement system from seizing up. The monetary base is endogenous (determined by lending activity) not exogenous, not a policy lever that controls lending.

Third, the real constraint on bank lending is not reserve availability but capital adequacy. Under the Basel regulatory framework (Basel I introduced in 1988, Basel II in 2004, Basel III from 2010), banks must hold equity capital as a percentage of their risk-weighted assets. Making a loan increases risk-weighted assets, which requires additional capital. Banks that are short of capital cannot lend regardless of how many reserves are available. The 2008 financial crisis destroyed bank capital on a massive scale, which is precisely why the enormous post-crisis increase in base money (through quantitative easing) did not produce the high inflation that simple multiplier reasoning would predict.

The concept of endogenous money (the idea that the money supply is determined by credit demand rather than by central bank fiat) represents the most consequential gap between standard economics education and the actual functioning of the monetary system. Post-Keynesian economists, Hyman Minsky, and more recently the Bank of England itself have articulated this view with increasing clarity. Mainstream macroeconomics has been slow to integrate it.

The Post-Keynesian Account

Post-Keynesian monetary economists: Basil Moore's 1988 work Horizontalists and Verticalists is the canonical treatment, argue that money is "horizontally" supplied: banks create as much money as the creditworthy demand for loans requires at the prevailing interest rate. The central bank sets the interest rate (the "price" of money) but cannot directly control the quantity. The money supply is determined by the interaction of borrower demand, bank willingness to lend, and the credit conditions prevailing in the economy.

This has precise implications. Credit expansions are driven by animal spirits: Keynes' term for the irreducibly uncertain expectations about future returns that motivate investment. When expectations are positive, credit demand rises, banks lend more, the money supply expands, and economic activity accelerates. When expectations turn negative (when uncertainty spikes, when asset prices fall, when defaults begin) credit demand collapses, banks tighten lending standards, the money supply contracts, and the economic deceleration becomes self-reinforcing.

Money creation and destruction are therefore inherently procyclical: the banking system amplifies both expansions and contractions. This is not a bug, it is a feature of a credit-money system. It creates the conditions for Minsky dynamics, which we will examine in depth in Artifact V.

Capital Requirements as the Binding Constraint

If reserves do not primarily constrain lending, what does? The answer in the modern regulatory framework is capital. Under Basel III (the international capital standard developed by the Basel Committee on Banking Supervision following the 2008 crisis) banks must hold Tier 1 capital (primarily common equity) equal to at least 6% of risk-weighted assets, with an additional 2.5% capital conservation buffer. Systemically important banks face additional surcharges.

Capital is not the same as reserves. Capital is the bank's equity, the difference between the value of its assets and the value of its liabilities. A bank that makes a loan that subsequently defaults loses the loan from its asset side but still owes the deposit to its liability side; the difference comes out of equity. If loan defaults exceed the bank's equity buffer, the bank becomes insolvent: its liabilities exceed its assets. The bank fails.

This is why capital requirements matter more than reserve requirements for constraining credit expansion. A well-capitalised bank can always obtain reserves, either from the interbank market or, in extremis, from the central bank. But a poorly-capitalised bank is constrained regardless of reserve availability, because each additional loan increases its risk-weighted asset base and further erodes its capital ratio. The 2008 crisis was, at its core, a capital crisis: the discovery that the banking system's actual capital was far less than its reported capital, because the risk weights assigned to mortgage-backed securities had been catastrophically wrong.

A bank run is not an irrational panic. It is a rational response to the discovery of a structural reality that the banking system normally keeps obscured: that a bank can never simultaneously repay all its depositors. A bank run is what happens when enough depositors discover this fact and act on it simultaneously.

The Logic of the Run

Consider the situation of a depositor who hears a rumour that their bank is in trouble. They have two choices: withdraw their deposits now, or wait to see if the rumour is true. If the rumour is false and they withdraw: they lose nothing, since they simply hold cash instead of deposits, and they can redeposit when the situation is clear. If the rumour is true and they withdraw: they protect themselves. If the rumour is false and they do not withdraw: fine. If the rumour is true and they do not withdraw: they lose their deposits.

The dominant strategy for any individual depositor is to withdraw at the first sign of trouble. This is not irrationality, it is the correct individual response to the structure of the situation. The bank run is not caused by irrationality; it is caused by the rational application of individual decision-making to a system whose stability depends on collective irrationality, specifically, on enough depositors not acting in their individual best interest simultaneously.

The Nobel Prize-winning economists Douglas Diamond and Philip Dybvig formalised this logic in their 1983 paper "Bank Runs, Deposit Insurance, and Liquidity", the theoretical foundation of modern understanding of banking fragility. The Diamond-Dybvig model shows that a bank run is a self-fulfilling prophecy: if enough depositors believe a bank is about to fail, their simultaneous withdrawal will cause the bank to fail, even if it would have been perfectly solvent had depositors not withdrawn. The belief causes the reality it predicts. This is financial reflexivity in its purest form.

The Northern Rock Case, 2007

Northern Rock (a British mortgage lender that demutualised and became a bank in 1997) provides the most recent British case of a classic bank run, and it illustrates the modern version of the mechanism with particular clarity. Northern Rock's business model was to originate mortgage loans and fund them not primarily through retail deposits (as traditional banks do) but through the wholesale money markets: it issued short-term bonds to institutional investors and used the proceeds to fund long-term mortgages. This was maturity transformation on a grand scale, with the added vulnerability that wholesale market funding could disappear more suddenly than retail deposits, since institutional investors can withdraw funding with no regulatory protection and often in seconds.

When the US subprime crisis began to freeze wholesale money markets in August 2007, Northern Rock's funding model collapsed. On September 14, 2007 (the first day news of the Bank of England's emergency support broke publicly) queues of depositors formed outside Northern Rock branches across Britain. Some 24 hours later, approximately £1 billion had been withdrawn, and the queues continued to lengthen. The Bank of England and Treasury were forced to guarantee all Northern Rock deposits to stop the run.

It was the first British bank run since Overend Gurney in 1866. It would not be the last event of 2008.

[ NORTHERN ROCK BANK RUN: SEPTEMBER 14, 2007 ]

The recognition that fractional reserve banking is inherently susceptible to runs led directly to the development of central banking's most important function: the lender of last resort. The intellectual foundation was laid in 1873 by Walter Bagehot in Lombard Street: A Description of the Money Market, the foundational text of central banking practice.

Bagehot's Rule

Bagehot's prescription for managing banking crises was precise and elegant, and its precision is worth preserving. The central bank, during a liquidity crisis, should: lend freely, at a high rate of interest, against good collateral.

Each element is essential. Lend freely: in a panic, the central bank must be willing to provide unlimited liquidity to solvent institutions. Half-measures are worse than none, uncertainty about whether the central bank will provide support amplifies the panic. At a high rate of interest: to penalise borrowers for needing emergency support, to deter borrowing by institutions that do not genuinely need it, and to prevent the central bank facility from becoming a permanent subsidy. Against good collateral: to distinguish between solvent institutions facing a liquidity crisis (temporarily unable to convert assets to cash, but fundamentally sound) and insolvent institutions facing a solvency crisis (whose assets are genuinely worth less than their liabilities). Only the former should receive central bank support, the latter should be allowed to fail.

The Bagehot rule contains, in its third element, a crucial distinction that would be systematically violated in the 2008 crisis and its aftermath: the distinction between illiquidity and insolvency. An illiquid bank has assets worth more than its liabilities but cannot convert those assets to cash quickly enough to meet immediate demands. An insolvent bank has liabilities exceeding its assets, it has made bad loans, and no amount of liquidity support will make it solvent. Bagehot's prescription is that only illiquid, solvent banks deserve support. In 2008, it became very difficult to tell which category any given institution belonged to. Which is itself part of what made the crisis systemic.

Banking crises are not aberrations or surprises. They are recurring features of the credit-money system, with a recognisable anatomy that repeats across centuries and geographies. Carmen Reinhart and Kenneth Rogoff, in their comprehensive 2009 study This Time Is Different: Eight Centuries of Financial Folly, examined 800 years of financial crises across 66 countries. Their central finding (expressed in the book's mordantly ironic title) is that the belief that the current situation is immune to the patterns of history is itself a consistent feature of the period preceding a crisis.

The common anatomy of banking crises, distilled from this history: (1) a period of benign conditions reduces risk perception; (2) credit expands, financing rising asset prices; (3) rising asset prices validate the credit expansion and encourage further lending; (4) at some point (often triggered by an external shock, a rate rise, or simply the exhaustion of the self-reinforcing dynamic) asset prices stop rising; (5) borrowers who had expected to repay loans from capital gains cannot do so; (6) defaults rise, bank assets fall in value, and capital is impaired; (7) credit contracts, asset prices fall further, and the feedback loop runs in reverse. This is the Minsky moment, the subject of Artifact V.

"Too big to fail" is a phrase that entered mainstream discourse in the 1980s but describes a problem as old as fractional reserve banking: some financial institutions become so deeply embedded in the functioning of the monetary system that their failure would cause damage beyond their own balance sheets, propagating losses and liquidity crises across the entire economy.

The Interconnectedness Problem

In a simple banking system with many small, independent banks, bank failures are painful for their depositors and shareholders but not systemically devastating. The failed bank's assets are acquired by surviving banks, its depositors (if uninsured) take losses, and the system continues. This is close to the world of the nineteenth century, many small banks, limited interconnection, bank failures frequent but generally local.

The modern financial system is fundamentally different. Large financial institutions are connected to each other through an enormous web of bilateral counterparty relationships: interbank loans, derivatives contracts, repurchase agreements, securities lending, payment system dependencies. When Lehman Brothers filed for bankruptcy on September 15, 2008, it had approximately $600 billion in assets and counterparty relationships with virtually every major financial institution in the world. Its failure immediately created uncertainty about the value of assets and the creditworthiness of counterparties across the entire financial system. The failure of a single institution triggered a global financial crisis not because of its size alone, but because of its centrality in a network of interconnections.

The "Too Big to Fail" Subsidy

The expectation that systemically important banks will be rescued by governments creates a structural competitive advantage that is both economically significant and difficult to eliminate. If creditors believe that a large bank's liabilities will be honoured by government guarantee in a crisis, they will accept a lower interest rate from that bank than they would from a smaller institution. This is not a theoretical observation, it has been quantified. Various studies estimate the implicit "too big to fail" subsidy received by the largest US and European banks at between $30 billion and $80 billion per year in reduced funding costs prior to the post-2010 regulatory reforms.

This subsidy does several harmful things simultaneously: it gives large banks a competitive advantage over smaller rivals, encouraging further concentration. It gives large bank shareholders an incentive to pursue risk, since they capture the upside but are protected from the full downside. It transfers risk from bank shareholders and creditors to taxpayers, without compensating taxpayers for bearing that risk. It is, in economic terms, a large and recurring transfer from the public to large financial institutions, disguised as a regulatory arrangement.

The 2008 financial crisis is the most thoroughly documented and studied banking crisis in history. It deserves detailed treatment not as history but as a case study in the mechanics of how the banking system can amplify shocks into systemic failures, and how the institutional response reveals the real architecture of power and obligation in the modern financial system.

Phase I: The Asset Price Inflation (2002–2006)

The US housing market experienced a sustained asset price inflation between 2002 and 2006, driven by a combination of historically low interest rates (the Federal Reserve maintained the federal funds rate at 1% from June 2003 to June 2004, the lowest level since 1958), financial innovation in mortgage-backed securities and collateralised debt obligations, and the systematic deterioration of lending standards as securitisation separated mortgage originators from the risk of default.

The securitisation chain worked as follows: a mortgage originator (Countrywide, IndyMac, New Century) extended a mortgage loan to a borrower. Instead of holding the loan, it sold it to an investment bank (Goldman Sachs, Bear Stearns, Lehman Brothers), which pooled thousands of mortgages into a mortgage-backed security (MBS). The MBS was then tranched, divided into senior tranches (rated AAA by credit rating agencies, first in line for payment) and junior tranches (rated BB or lower, last in line). The senior tranches were further pooled into CDOs (collateralised debt obligations), which were again tranched and sold to pension funds, insurance companies, money market funds, and European banks globally.

The rating agencies, paid by the issuers of the securities they rated, applied models that systematically underestimated default correlation, the tendency for mortgages to default simultaneously when housing prices fell nationally rather than locally. The AAA ratings were wrong. The senior tranches of CDOs were not as safe as AAA-rated corporate bonds. But as long as housing prices rose, the models appeared validated, and the securitisation machine accelerated.

Phase II: The Inflection and First Cracks (2006–2007)

US national house prices peaked in April 2006. The Federal Reserve had been raising rates since June 2004, reaching 5.25% by June 2006, the highest since 2001. Adjustable-rate mortgages, many of which had been extended with low "teaser" rates for the first two years, began resetting to higher rates. Borrowers who had been extended credit on the assumption that rising house prices would allow refinancing found themselves unable to refinance when prices began to fall. Default rates on subprime mortgages began to rise.

The early 2007 failures of subprime mortgage originators (New Century Financial filed for bankruptcy in April 2007) seemed containable. The consensus view was that subprime problems were "well-contained." They were not. The problem was not the subprime losses themselves, they were large but not, in isolation, catastrophic. The problem was the uncertainty about where those losses resided. The securitisation machine had distributed subprime exposure globally and opaquely. Nobody knew what their counterparties actually held. When uncertainty rises to this level, the rational response is to trust nobody, and that is what happened.

Phase III: The Freeze and the Cascade (2007–2008)

In August 2007, the interbank lending markets began to seize. The TED spread (the difference between three-month LIBOR (the rate at which banks lend to each other unsecured) and the three-month Treasury bill rate (the risk-free rate)) began to widen dramatically. Banks stopped lending to each other because they could not assess each other's exposure to subprime-related losses. The shadow banking system (the money market funds, SIVs (structured investment vehicles), and repo markets that had funded much of the mortgage securitisation) began to contract.

The collapse of Bear Stearns in March 2008 (rescued by JP Morgan in a Fed-facilitated deal at $2/share (later raised to $10)) was the first crisis moment. The Fed's decision to rescue Bear but not Lehman (which filed for bankruptcy on September 15, 2008) sent the financial system into its deepest shock since 1933. The money market fund Reserve Primary Fund, which held Lehman commercial paper, "broke the buck" (its net asset value fell below $1) on September 16. This triggered a run on the entire money market fund industry, which began liquidating assets simultaneously, crashing prices further. The shadow banking run was happening in real time across every major financial market on earth.

Phase IV: The Response and Its Costs

The US government response, the $700 billion TARP (Troubled Asset Relief Program) passed in October 2008, the nationalisation of Fannie Mae and Freddie Mac in September 2008, the AIG bailout at a cost of $182 billion, the emergency guarantee of money market funds, the Fed's expansion of its balance sheet from $900 billion to $2.3 trillion by end-2008, stabilised the system but at an enormous cost. The ultimate total of US government commitments exceeded $13 trillion, though most was recovered. The crisis destroyed approximately $11 trillion in household wealth in the United States alone. Global GDP fell by approximately 2% in 2009, the first global recession since the Second World War.

The political consequences (the rise of the Tea Party and Occupy Wall Street, the profound loss of faith in financial institutions, the conditions for the subsequent populist political movements of the 2010s) are arguably larger in long-run terms than the economic consequences. When the public discovered that the same institutions whose risky lending had caused the crisis had been rescued by the government while ordinary homeowners lost their houses, the social contract of the financial system was severely damaged. That damage persists.

We have now established the mechanical reality of the banking system. Commercial banks create the majority of the money supply through lending. The multiplier is real but not mechanically fixed. Capital, not reserves, is the binding constraint. The system is structurally fragile and susceptible to self-fulfilling runs. Systemic institutions receive implicit government guarantees that socialise risk while privatising returns. And periodic crises, with a recognisable anatomy, are recurring features rather than accidents.

What does this mean? Several implications follow that are routinely obscured in public discourse.

Money Creation Is a Private Function With Public Consequences

In most advanced economies, approximately 95-97% of the money supply consists of commercial bank deposits (money created by private lending decisions. These decisions are made by private institutions, motivated by profit, subject to the procyclical dynamics of credit expansion and contraction. The social consequences) the inflation or deflation, the boom or bust, the distribution of new purchasing power to borrowers rather than savers, are enormous, and yet the decisions are private.

This is one of the least-examined features of the modern monetary system. We accept, without much discussion, that the majority of money creation (the expansion of the economy's purchasing power) is determined by the lending decisions of private banks. The central bank influences the price of credit (the interest rate) but does not directly control its allocation. Who gets credit (what projects are financed, what assets are inflated, who is excluded) is determined by private banks subject to market incentives and regulatory constraints.

The System Requires Perpetual Growth

Because most money is created as debt (because every pound of new money in the economy represents a loan that must eventually be repaid with interest) the system has a built-in requirement for perpetual expansion. If credit stops expanding, the money supply stops expanding. If the money supply stops expanding in an economy in which most actors are servicing existing debts, the result is deflation and economic contraction. Debt repayment destroys money in exactly the same way that lending creates it: when a loan is repaid, the deposit created by that loan is extinguished. The system requires new lending to replace retiring loans and provide net growth in purchasing power. An economy that stops borrowing tends toward stagnation.

This structural feature is the deep foundation of Ray Dalio's debt cycle analysis and Minsky's financial instability hypothesis, the subjects of Artifacts IV and V in this series. The banking system is not merely the infrastructure of the economy; it is the engine. And like all engines, it has failure modes. Understanding those failure modes is the beginning of economic literacy.

The Next Layer: Central Banks as System Managers

The banking system creates money through credit but cannot stabilise itself. Its inherent fragility (the maturity mismatch, the susceptibility to runs, the procyclical dynamics) requires an external institutional backstop: the central bank. What the central bank actually does, how it does it, and what the limits of its power are, these questions are the subject of Artifact IV.