The PUNZI Papers Essay Three · The Banking System Is a Ponzi

This essay is part of The PUNZI Papers, a series of essays explaining the reason for our project, PUNZI WARS, and the potential we think it points to.

Help us determine which tokens make the game.

Visit Us: https://punzi.xyz     Follow us on X: @PUNZIwars

Advisory: PUNZI WARS is a game and an on-chain research experiment. It is not an investment. Participate at your own risk.

Modern banking honours withdrawals from a deposit base it has lent out many times over. The structure has been load-bearing for three centuries, and it is now failing the one test it was never built for: a run that moves faster than the institutions sworn to catch it.

Thirty-Six Hours in March

On the morning of Thursday, 9 March 2023, Silicon Valley Bank was a regulated, federally insured commercial bank with approximately $209 billion in assets, the sixteenth-largest in the United States, and the principal banking relationship for a large share of the American venture capital economy. By the close of business Friday it no longer existed as an independent institution. The California Department of Financial Protection and Innovation seized it, the Federal Deposit Insurance Corporation was appointed receiver, and roughly thirty-six hours had passed between the first public sign of distress and the moment the doors closed.

The proximate cause was on file. During the venture funding boom of 2020-2021 the bank had taken in unusually large, uninsured corporate deposits, and it had placed much of the money in long-duration Treasury and agency securities, on the sound principle that the safest asset in the world was safe. When the Federal Reserve raised the federal funds rate from near zero to a target range of 4.25%-4.50% by the end of 2022, the value of those securities fell, as bonds in such a situation would do. The losses sat quietly on a held-to-maturity book that was not marked through earnings. They were real only if someone asked for the money back.

On Wednesday, 8 March, the bank announced an equity raise and the sale of $21 billion in available-for-sale securities at a realised loss of approximately $1.8 billion. By Thursday morning, several prominent venture investors had advised their portfolio companies to move their cash. The advice travelled through group chats, Slack channels and Twitter at a speed unavailable to any prior generation of depositor. By Thursday night, depositors had requested approximately $42 billion in withdrawals in a single day, roughly a quarter of the bank’s deposits, with a further $100 billion reportedly queued for Friday. The bank could not meet the calls and was closed before the market opened.

Two facts about that weekend are worth pausing on. First, the bank was not, by ordinary measures, insolvent on 8 March. It was insolvent only in the sense that its long-dated assets could not be turned into cash on the timeline its depositors demanded. Second, on Sunday, 12 March, the Treasury, the Federal Reserve and the FDIC jointly announced that all depositors at Silicon Valley Bank would be made whole regardless of the $250,000 statutory cap on deposit insurance, through a systemic risk exception not invoked at this scale since 2008.

The structural fact the weekend made unusually legible is the one that concerns us. A regulated, supervised, federally insured institution, holding the operating cash of a substantial fraction of American technology firms, could not honour withdrawal requests from a normal customer base on normal terms, because the assets backing those deposits were not in a form that could be returned on demand. This was not a defect of Silicon Valley Bank, and not a failure of supervision. It is the design of modern banking, working exactly as designed, in full view of every regulator paid to watch it.

What Banks Actually Are

Strip the regulatory apparatus away from a commercial bank and what remains is a Ponzi-like structure in the sense developed in the opening essay of this series. All three features are present.

1. Earlier participants are paid from later ones. A depositor at the teller’s window receives cash from the bank’s reserves, replenished at the margin by the deposits of others, by interbank borrowing, and, in extremis, by the central bank. The loan book cannot be liquidated quickly enough to honour any large fraction of depositor claims at once.

2. Operation depends on continued inflow and non-panic. Withdrawal requests are statistically uncorrelated across the depositor population. Most depositors leave most of their balance with the bank on most days. What the structure cannot survive is coordinated, simultaneous withdrawal.

3. The bank does not hold the assets its liabilities promise. A deposit is a callable loan from the depositor to the bank. The matching assets are loans to borrowers, securities, and a comparatively thin sliver of cash and central-bank reserves, of considerably longer average duration than the liabilities. The transformation between the two is precisely the business.

This is not a description of a fraud. It is the legal structure of every commercial bank in every developed economy, disclosed in every quarterly filing, supervised by national and supranational regulators, and protected at the margin by deposit insurance and the discount window. The intent is benign. The disclosure exists, in a technical sense, in every prospectus and account agreement the bank produces.

The participant classes are likewise distinct from the participants in a swindle. Depositors hold callable claims, often insured up to a cap ($250,000 in the United States, £120,000 in the United Kingdom, €100,000 across the European Union banking union). Borrowers hold liabilities at longer durations than the deposit base that funds them. Shareholders absorb losses first, bondholders next. Deposit insurance funds and, behind them, taxpayers and the central bank as lender of last resort backstop the structure when its own resources are exhausted. Banking is not a fraud. It is something more disquieting: a fully legal, openly disclosed, socially endorsed implementation of a Ponzi-like structure, sanctioned by every authority a depositor is taught to trust, and its failure modes are predictable along the three axes the first essay in our series set out.

What the Banking Structure Solves

A fair critique must concede the legitimate function, which is real. The technical term is maturity transformation. Most savers want their money back on short notice. However, most productive uses of capital (for example: a mortgage, a factory, a fleet of trucks or ships) require funding at durations no individual saver will tolerate. Without an intermediary that aggregates short-term claims and disburses long-term loans, most loans would never be made, leading to less economic activity.

The canonical account is Douglas Diamond and Philip Dybvig’s 1983 paper, ‘Bank Runs, Deposit Insurance, and Liquidity’. A bank can offer everyone the option of withdrawing on demand while investing the pool in longer-term assets, because most depositors will not, on any given day, exercise the option. It sells liquidity insurance against the statistical regularity of the pool. The same model produces the bank run as a second equilibrium of the depositors’ coordination problem. Each individual’s decision to withdraw, given the belief that others are withdrawing, is rational. Collectively, the result is catastrophic. The good equilibrium and the bad one are both available at all times; which one obtains depends on what depositors believe the others believe.

A Brief History of the Pattern

The structure is older than the term ‘central bank’. Its modern shape is conventionally traced to the goldsmith bankers of seventeenth-century London, who noticed that depositors rarely reclaimed all their gold at once and began to lend out more receipts than they held in vault gold. The pattern predates the regulatory apparatus now built around it by roughly three centuries.

The Bank of England was chartered in 1694. Walter Bagehot’s Lombard Street, published in 1873, formalised the central bank as lender of last resort: in a panic, lend freely, at a penalty rate, against good collateral. The doctrine is more conservative than what subsequent central banks have actually done.

The American history is choppier. The First and Second Banks of the United States (1791-1811, 1816-1836) were short-lived. The Panic of 1907, in which J. P. Morgan personally organised a private rescue, motivated the Federal Reserve Act of 1913. The Fed was meant to prevent panics. It did not. Roughly nine thousand American banks closed between 1929 and 1933, on its watch, aggravated, in the account associated with Milton Friedman and Anna Schwartz’s A Monetary History of the United States (1963), by the Fed’s failure to act as lender of last resort at the scale Bagehot prescribed. Ben Bernanke’s 1983 paper extended the diagnosis: the destruction of banks slowed the recovery for years after the panic ended.

The institutional response is the apparatus most depositors take for granted. The Federal Deposit Insurance Corporation, established in 1933, insures depositors up to a statutory cap that has risen over time. Continental Illinois National Bank, which failed in May 1984, was the first major postwar resolution in which the FDIC made all depositors whole. ‘Too big to fail’ entered the lexicon.

The Basel Accords, beginning with Basel I in 1988, layered capital requirements on top of the older reserve requirements, shifting the binding constraint from the asset side (cash on hand against deposits) to the equity side (capital against risk-weighted assets). In March 2020, the Federal Reserve reduced reserve requirement ratios to zero across all transaction-account tranches, effective 26 March 2020. The ratio that had been ten percent on most transaction deposits for most of the twentieth century became, on paper, nothing.

Then 2023. Silicon Valley Bank failed on 10 March, Signature Bank on 12 March, First Republic Bank on 1 May: together, three of the four largest bank failures in American history by assets. The FDIC invoked the systemic risk exception for the first two and resolved the third through an acquisition by JPMorgan Chase. What uninsured depositors at the first two received was, in operational terms, an extension of deposit insurance to the entire deposit base of two institutions that regulators determined to be systemically important after the fact. The formal line between insured and uninsured deposits held. The line that mattered moved.

How Banking Works

The textbook account of how banking works is worth recovering, because the textbook is only half right, and the other half that is wrong genuinely matters. A depositor delivers $100. The bank books a $100 asset and a $100 liability, holds a fraction (say ten percent) as reserves, and lends $90. The $90 is spent, deposited at another bank, which holds $9 and lends $81, and so on. The geometric series converges to $1,000 in deposits on the original $100 of reserves. This is the money multiplier, the inverse of the reserve ratio.

Nonetheless, as a description of how banks currently operate, this description is obsolete. Reserve requirements in the United States were reduced to zero in March 2020 and have not been restored. The Bank of England’s 2014 paper ‘Money creation in the modern economy’ argued, with substantial subsequent uptake, that modern banks do not lend out a reserve already received but create a deposit by extending credit, settling the reserve question afterward through interbank markets and central-bank operations.

The active constraint is no longer the reserve requirement but the capital requirement, under Basel III. The shift matters because it changes the failure mode. A reserve-constrained bank fails when a withdrawal exceeds its thin slice of cash. A capital-constrained bank fails when its asset values deteriorate enough to wipe out the equity buffer, at which point the withdrawal is merely the trigger that converts an asset-value problem into a cash problem. Silicon Valley Bank was an example of the second case.

Applying The Three-Axis Test

Applying the diagnostic the first essay of our series set out, the banking system fails on all three axes, and it fails them through the same underlying machinery: a functional opacity about what a deposit actually is, welded to an authority overlay that licenses the opacity to operate. In banking that overlay has institutional names. There are five of them, and naming them sharpens the diagnosis. Deposit insurance, statutorily capped, is the most visible. The central bank, as lender of last resort against par-value collateral, is the second. Bank regulators – the OCC, the Federal Reserve in its supervisory capacity, the FDIC, the Basel apparatus – are the third, certifying that institutions meet standards treated as load-bearing. The systemic risk exception, invoked twice in March 2023, is the fourth: a discretionary backstop that erases the line between insured and uninsured at any institution found, after it has failed, to after-the-fact deem is as ‘too important’. This implicit too-big-to-fail guarantee is the fifth. In short, each layer signals to the depositor that someone responsible has already done the analytical work. The signal is the product. Implied understanding that the work was done is the thing it is sold in place of.

On efficiency, the cost question is unusual because the intermediation is genuinely productive. A bank that allocates credit well creates real value, and the net interest margin compensates that function. The failure is not that banks make money. It is that the productive function and the subsidy capture sit inside the same legal entity, where they cannot be separated. American commercial banks held approximately $24 trillion in assets at the end of 2024, the four largest – JPMorgan Chase, Bank of America, Citigroup, Wells Fargo – accounting for roughly $11 trillion. Returns on equity at the largest American banks have held in the low-to-mid teens through most of the post-2010 period, above what a sector of comparable size and risk would earn without deposit insurance, capital requirements that double as entry barriers, and the understanding that institutions of a certain size will not be allowed to fail. Deposit insurance lowers the bank’s cost of funds. Lender-of-last-resort access lowers its tail risk. The bank captures the benefit. The public, as taxpayer and depositor, is handed the cost of the implicit option without being asked. The share of profitability owed to allocation skill, as opposed to public subsidy, is no longer separable from the public record, which is itself the point: an extraction that cannot be measured cannot be priced, and cannot be refused.

On transparency, the failure is functional rather than formal, and severe once stated. A depositor who places funds in a commercial bank is, in legal substance, an unsecured creditor of the bank. The deposit is a callable loan. In failure, insurance pays up to the cap; amounts above it are creditor claims in receivership, which may pay out at par, at a discount, or not at all. Almost no retail depositor understands this. The deposit is presented, in marketing and in ordinary language, as a stored asset, money that is ‘in’ the bank in something like the literal sense in which gold was once held in a vault. The Federal Reserve and the FDIC are aware of the misunderstanding and depend on it. The explicit purpose of deposit insurance is to make the misunderstanding harmless under ordinary conditions, by removing the depositor’s incentive to discover what they are actually holding. The disclosures are not concealed (most sit in the account agreement and the FDIC’s public materials). The structure does not merely assume that depositors will not read them, will not understand them if they do, and will rely instead on the system to contain the risk. It requires exactly that, and is built to keep it so. Depositors who held more than $250,000 at Silicon Valley Bank learned only over the weekend of 11-12 March 2023 whether they would be made whole, by a discretionary decision announced in a joint statement of three officials. Under the formal rules they had agreed to, they had no such guarantee. Only later their position depended on three people agreeing over a weekend.

On longevity, the 2023 episode placed a question on the table that it did not retire. Modern banking was designed under one set of communication conditions and now operates under another. The Northern Rock run in September 2007, the first British bank run since the 1860s, unfolded over five days. Queues outside branches became visible to television cameras on 14 September; the Bank of England announced liquidity support that day; the bank was nationalised in February 2008. What stands out now is how slow it was. The depositors physically travelled to branches, the failure was visible because the queue was, and the intervention had time to organise itself across the working day. The Silicon Valley Bank episode unfolded over thirty-six hours. The depositors did not queue; they instructed treasury operations to move funds by wire and ACH from corporate dashboards, and the information that others were moving travelled through social media in close to real time. The run equilibrium is still the run equilibrium. What has changed is the speed at which it crystallises. The two technologies that have most plainly broken the regularity of depositor behaviour – the smartphone and the social network – are contained by neither the deposit insurance system nor the lender-of-last-resort doctrine. They are contained, currently, by the willingness of the central bank to lend against par-value collateral on weekends, and by the willingness of the political branches to ratify those decisions afterward. The structure is passing the longevity test on a series of unanimous discretionary decisions taken under acute time pressure by a handful of officials. That is not the same as being built to pass it. It is the absence of failure standing in for a design that would survive one.

There is a further turn the five layers conceal, and it is the most damaging. The same backstop that steadies the structure also instructs it. An institution that knows it is too large or too interconnected to be allowed to fail has been handed an incentive, not merely a cushion: the larger and more entangled it becomes, the more certain the rescue, and the more certain the rescue, the less reason it has to price its own risk honestly. Gains stay private; losses are socialised. Under that asymmetry the rational move is to take more risk, not less – to reach for the leveraged, the illiquid, the duration-mismatched position that pays in the good years and is made whole by the public in the bad ones. The guarantee built to contain the failure mode quietly subsidises and encourages the behaviour. In short, the safeguards have become accelerants, corroding the two axes at once: it lets the largest banks capture rent the market would otherwise price away, and it loads the system with ever-increasing risk the longevity test is meant to measure.

Who Pays When It Fails

The cost-bearing structure in banking has been deliberately engineered, and the engineering has shifted over time. Insured depositors are made whole by the FDIC up to the cap. The fund stood at approximately $137 billion at the end of 2024, against roughly $10.7 trillion in insured deposits, a ratio of about 1.28%. When drawn down, it is refilled by raising the assessment on surviving banks, which pass the cost to their customers in slightly higher loan rates and slightly lower deposit rates.

Uninsured depositors used to bear the loss above the cap. The 2023 episode confirmed what 2008 had strongly suggested: this is no longer reliably the case. At an institution found important after the fact, uninsured depositors are now substantially likely to be made whole through the systemic risk exception, the cost socialised through the same assessment mechanism. There is a pattern in who gets caught. The uninsured depositors made whole at Silicon Valley Bank were overwhelmingly venture-backed firms and the funds standing behind them – the same well-connected cohort whose principals had spent Thursday telling one another to pull their cash, and who spent the weekend publicly demanding the backstop they then received. The formal line between insured and uninsured did not move. The informal line – between those close enough to the decision to be made whole and those left to the statutory cap – is the one that did the work, and it is written down nowhere a depositor can read it. The rule held. The practice rewarded proximity to power.

Shareholders and unsecured bondholders absorb losses first in the formal resolution waterfall, and this part was preserved: the equity of Silicon Valley Bank’s holding company was wiped out, its bondholders received cents on the dollar. The taxpayer is the residual backstop. The Treasury’s line of credit to the FDIC was not exhausted in any recent episode, exactly what allows the FDIC to commit on a weekend to a backstop it does not yet have the assessments in hand to fund. Beyond the taxpayer is the citizen, in their capacity as holder of a currency whose value depends on the central bank’s commitments not metastasising into a balance-sheet problem of their own. The Federal Reserve’s balance sheet expanded from approximately $900 billion in 2008 to roughly $9 trillion at the peak in 2022, and the political economy of its normalisation remains unresolved.

The final cost-bearing population is the one least often named. Every household that does not own significant financial assets pays, in higher housing costs and lower real wages, when monetary intervention to contain a banking crisis lifts asset prices faster than the price of labour. These losses appear on no resolution waterfall. They appear in the gap between asset-price indices and wage indices over the decade after a crisis, and the 2008-2020 period exhibits the gap in unusually clean form. The structure resembles an insurance pool whose actuarial assumptions have drifted away from its premium base: the insured population has expanded informally, the premiums have not been repriced, and the gap is closed, when needed, by drawing on the residual claim of the citizen. That residual claim is not a separate, deeper pocket. It is the same currency, the same sovereign credit, the same trust the opening essay watched being rolled forward. The pools are not independent. They drain into one another, and a single drained pool pulls the rest down with it.

What the Three Axes Reveal

The legitimate function is real. Maturity transformation finances activity that would otherwise be unfinanceable. Liquidity insurance solves a coordination problem individual savers cannot solve. The structure has been load-bearing for three hundred years, and the developed economies that operate without it are non-existent. The three-axis test does not invalidate any of that. It clarifies where the implementation has decayed, and the decay has a single shape: the safeguards, stripped of matching accountability, have inverted into the cause of the thing they were built to prevent.

A guarantee that absorbs losses without exacting a price changes the calculation of everyone it protects. Risk becomes free to take and expensive only to forgo, and the risk-and-reward arithmetic meant to discipline a financial institution becomes a one-way bet: the upside privately kept, the downside publicly assumed. A system that rewards its participants for taking risks it has promised to socialise will get more of them, larger and more correlated, until the inefficiency is structural and the fragility permanent. Risk is no longer being mitigated. It is being systematised. Each intervention that works is folded into the structure as a standing expectation, so the next crisis begins from a higher baseline of assumed rescue and a thinner margin of genuine reserve. The exception becomes the mechanism.

Modern banking is therefore a Ponzi-like structure whose three-axis profile has shifted, beneath the people inside it, in ways most participants were never told and never agreed to. The implementation has accumulated public commitments – deposit insurance, lender-of-last-resort facilities, systemic risk exceptions, balance-sheet expansions – that were once exceptional and are now routine. Each was justified in its moment. Their cumulative shape is the structure depositors now stand inside, and it is not the structure described in any deposit account agreement they have signed. The answer the record gives is not a clean failure. It is something more uncomfortable: a structure that works, for now, only through a thickening lattice of public commitments most of its participants do not know they have made, and would not have consented to if they had.

What banking holds is not an abstraction. It is the deposit base of the economy, the operating cash of firms and the savings of households, and the money people believe they simply have. The failure mode now takes thirty-six hours, and the only thing standing between a single uncontained run and the evaporation of that money is the willingness of a central bank to conjure liquidity over a weekend, and the willingness of everyone else to keep believing the conjuring is solvent.

That backstop is neither free nor infinite. It rests on the same sovereign credit the last essay in our series described, and each rescue enlarges a central-bank balance sheet whose own credibility is among the last things holding the currency up. A lattice of structures that each lean on the next is sturdy until one of them slips; then the leaning becomes falling. The system has caught every fall so far. It was not built to catch the one that arrives faster than a weekend, in more than one place at once, after the catching itself has begun to look unsound.

Whether that combination is removable in the institution of banking (or in any Ponzi-like structure of any kind) is the question this essay does not answer. It leaves that with the reader, who keeps their money in a bank because there is nowhere else to keep it, and who will be standing inside the structure on the day a run finally outruns the weekend the rescue needs.

Conclusion · The PUNZI Papers

This completes the present essay of The PUNZI Papers, a series of essays explaining the position behind our project, PUNZI WARS, and the potential we think it points to.

In the next essay we address how The Precious Metal Market Is a Ponzi

Visit us: https://punzi.xyz Follow us on X: @PUNZIwars

Advisory: PUNZI WARS is a game and an on-chain research experiment. It is not an investment. Participate at your own risk.

Sources

1. Walter Bagehot, Lombard Street: A Description of the Money Market (1873)

2. Douglas Diamond and Philip Dybvig, ‘Bank Runs, Deposit Insurance, and Liquidity’, Journal of Political Economy 91, no. 3 (1983): 401-419

3. Milton Friedman and Anna Schwartz, A Monetary History of the United States, 1867-1960 (1963)

4. Ben Bernanke, ‘Nonmonetary Effects of the Financial Crisis in the Propagation of the Great Depression’, American Economic Review 73, no. 3 (1983): 257-276

5. Michael McLeay, Amar Radia, and Ryland Thomas, ‘Money creation in the modern economy’, Bank of England Quarterly Bulletin 2014 Q1

6. Federal Reserve Board, ‘Reserve Requirements’, release of 15 March 2020 (effective 26 March 2020)

7. Federal Deposit Insurance Corporation, joint statement with the Department of the Treasury and the Federal Reserve, 12 March 2023

8. Federal Deposit Insurance Corporation, FDIC’s Supervision of Silicon Valley Bank (Material Loss Review, 2023)

9. Federal Reserve Board, Review of the Federal Reserve’s Supervision and Regulation of Silicon Valley Bank (April 2023)

10. HM Treasury Select Committee, The Run on the Rock, Fifth Report of Session 2007-08 (2008)

11. Federal Reserve Bank of St. Louis, FRED data series on required reserve ratios and the deposit insurance fund balance