The structure built up in the first three modules — instant-callable deposits, illiquid loans, thin capital, and a web of mutual credit — is extraordinarily useful and inherently fragile. This module is about how it breaks. We separate the two ways a bank can die, walk through the mechanics of a run and why running can be perfectly rational even at a sound bank, show how a liquidity problem turns into outright insolvency, and trace the recurring history of panics from 1907 to the Great Depression to 2008 to Silicon Valley Bank in 2023 — each of which left behind a piece of the safety net that Module 05 examines.
Before any history or mechanics, fix a distinction that almost every confused discussion of banking failure gets wrong. A bank can fail in two fundamentally different ways, and telling them apart is the whole game.
Insolvency is a problem of value. A bank is insolvent when its assets are worth less than its liabilities — when the loans and securities it owns have fallen so far that they no longer cover what it owes depositors and other creditors. From Module 02, this means losses have chewed through the thin capital cushion and kept going. An insolvent bank is broken in substance: even given unlimited time, it cannot pay everyone back in full because the money simply is not there.
Illiquidity is a problem of timing. A bank is illiquid when it cannot meet the withdrawals being demanded right now, even though its assets, given time, are worth more than its liabilities. From Module 02's maturity transformation: the money is real but tied up in loans that repay over years, and you cannot hand a depositor a slice of a thirty-year mortgage. A perfectly sound bank can be illiquid if enough depositors show up at once.
Insolvency is "the assets are worth too little." Illiquidity is "the assets can't be sold fast enough." A bank can suffer either. The danger — and the heart of this module — is that the two are not cleanly separate: a liquidity problem, handled badly, can manufacture an insolvency that did not exist before. Keeping the pair straight is what lets you understand both runs and the safety net built to stop them.
A bank run is illiquidity in action. Recall the ingredients from earlier modules: depositors are creditors who can demand repayment instantly (Module 03); the bank holds only a small fraction of deposits as liquid reserves (Module 02); and the rest is locked in loans that cannot be called back on short notice. Now add one more feature — first-come, first-served. Reserves are paid out in the order depositors arrive. Whoever asks first gets cash; whoever asks after the reserves run dry is left waiting for the bank to liquidate assets, possibly receiving less, possibly much later.
That ordering rule is what makes a run explosive. Suppose you hear a rumor that your bank is in trouble. You may not even believe it. But you know that if others believe it and rush to withdraw, the reserves will be exhausted before everyone is paid — and you do not want to be at the back of the line. So the rational move, regardless of whether the rumor is true, is to withdraw first and ask questions later. Everyone reasons the same way. The rush itself drains the reserves and can topple the bank, making the feared outcome real.
This is why a run can ignite from a spark as small as a rumor, a falling share price, or a queue that forms for unrelated reasons. The bank does not have to be unsound for a run to start. It only has to be the kind of institution — every conventional bank — that promises instant repayment while holding illiquid assets and paying out first-come, first-served.
The most important idea about bank runs, and the one that earned a Nobel Prize, is that a run can be entirely rational and self-fulfilling — it does not require anyone to be foolish or the bank to be bad. The economists Douglas Diamond and Philip Dybvig formalized this in 1983 (work recognized with the Nobel Prize in 2022), and the insight reframes everything.
Their model shows that a bank funding illiquid assets with demand deposits has two possible equilibria — two stable states the depositors can collectively land in:
Which equilibrium prevails depends on beliefs about other depositors' beliefs, not necessarily on the bank's fundamentals. This is a coordination trap: each person doing the individually sensible thing produces a collectively disastrous outcome. It explains why confidence is the real asset a bank holds, why runs can strike institutions that were fundamentally fine, and why the cure has to work on the coordination problem itself — which is precisely what deposit insurance does, by removing any reason to be first in line.
A run can be a self-fulfilling prophecy. Because withdrawing first is rational whenever you fear others will withdraw, belief alone can topple a sound bank. The fragility is not (only) about bad lending — it is baked into the combination of demand deposits and illiquid assets. Fixing it means changing the incentive to run, not just making banks lend more carefully.
Section 01 insisted that illiquidity and insolvency are different. Now comes the cruel twist: a liquidity crisis can create the very insolvency it feared, through a mechanism called the fire sale.
Picture a sound but illiquid bank facing a run. Its reserves are exhausted, and depositors are still demanding cash. To raise money, it must sell its less-liquid assets — loans and longer-dated securities — immediately. But forced, urgent selling fetches bad prices. Buyers know the seller is desperate and bid low; selling a large block at once pushes the price down further. The bank receives far less than the assets are truly worth held to maturity. Those realized losses eat into capital. If they are large enough, they push the bank's asset value below its liabilities — and a bank that was merely illiquid at breakfast is genuinely insolvent by dinner. The run manufactured the insolvency.
Worse, the fire sale spills onto others. When a distressed bank dumps assets and drives their market price down, every other institution holding similar assets must mark its own holdings to the new, lower price — weakening their capital too, and possibly tipping some of them toward distress. This is the asset-price channel of the contagion introduced in Module 03: one bank's fire sale becomes every similar bank's loss. Liquidity trouble, insolvency, and system-wide contagion are thus tightly linked — which is why authorities treat a run not as one bank's problem but as a threat to the whole system.
For most of banking history, a run looked like a physical queue: anxious depositors lining up outside a branch, the line itself becoming the signal that fed the panic. That friction — the need to physically show up during banking hours — gave runs a natural speed limit and gave authorities time to respond.
That speed limit is gone. Today deposits move with a tap. A frightened depositor does not queue; they open an app and transfer the money in seconds, at any hour, from anywhere. And the rumor that triggers them no longer spreads by word of mouth down a city block — it spreads through group chats, social media, and financial news to thousands of depositors simultaneously. The coordination trap of Section 03 now resolves into the bad equilibrium almost instantly, because everyone gets the signal and can act on it at the same moment.
The deeper point is that none of the underlying mechanics changed — the mismatch, the first-come-first-served incentive, the coordination trap are all exactly as Diamond and Dybvig described. Technology did not create the fragility; it removed the friction that used to slow it down, compressing days into hours. This is one reason the digital era has pushed both regulators and innovators to revisit the basic design, a thread the companion innovation track picks up.
Banking crises are not rare aberrations; they recur across centuries and countries with striking regularity, because the fragility is structural. What is most instructive is that each major panic left behind a permanent piece of the safety net — the response to one crisis became the architecture that shaped the next. The pattern is the through-line of this whole track.
| Era | Crisis | What broke | What it left behind |
|---|---|---|---|
| 1907 | Panic of 1907 | Runs on trust companies with no central bank to supply liquidity; a private financier organized the rescue | The realization that the U.S. needed a public lender of last resort — leading to the founding of the Federal Reserve in 1913 |
| 1930s | Great Depression bank failures | Thousands of U.S. banks failed in waves of runs; depositors lost savings en masse | Federal deposit insurance (the FDIC, 1933) — removing the incentive for small depositors to run |
| 2007–08 | Global financial crisis | A run on the UK's Northern Rock; Lehman's failure; runs on money-market funds; frozen interbank lending | Sweeping new capital and liquidity rules (Basel III) and expanded resolution powers |
| 2023 | Regional-bank stress | The fastest digital run in history at Silicon Valley Bank; failures of Signature and First Republic; Credit Suisse rescued in Europe | Emergency full-deposit backstops and new central-bank liquidity facilities; a live debate about insurance caps and run speed |
Read down the right-hand column and you are reading the construction history of the modern safety net: a central bank to supply liquidity (1907 → 1913), deposit insurance to stop small-depositor runs (1930s), capital and liquidity rules to make banks more shock-resistant (2008), and emergency backstops calibrated to digital-speed runs (2023). Each was a response to a fragility that the previous architecture had not anticipated. The next module takes that safety net apart piece by piece.
Runs and crises are universal, but they take locally distinctive forms, and a comparative eye sees patterns a single-country view misses.
Gather the module together. The conventional bank is fragile in a deep, structural way: it promises instant repayment while holding illiquid assets (maturity mismatch), it rests on a thin capital cushion (leverage), its depositors face a first-come-first-served race that makes running rational (the coordination trap), forced sales can turn illiquidity into insolvency (the fire-sale doom loop), and the web of credit spreads any one failure to the rest (contagion). Modern technology has stripped away the friction that used to slow runs down. History shows this is not hypothetical — panics recur across centuries and continents.
No society tolerates recurring banking collapse for long, because banking failure means the payment system and the money supply themselves seize up. So every developed financial system has built a safety net to contain the fragility — and, crucially, each component targets a specific failure mode from this module:
This is the bargain we first glimpsed in Module 01: the state stabilizes banking, and in exchange it regulates banks heavily. Module 05 takes the safety net apart piece by piece — how deposit insurance is funded and capped, how the lender of last resort is supposed to operate, and what the capital rules require. And it sets up the sting in the tail that Module 06 develops: the same safety net that tames fragility also raises a wall around the industry, because only a tightly regulated, chartered, central-bank-connected institution can be trusted with it — which is exactly why newcomers find banking so hard to enter.
Six questions on fragility, runs, and the mechanics that turn one bank's trouble into a crisis. The questions test the reasoning rather than memorization of dates.