A bank run is the oldest failure mode in finance and the purest example of a self-fulfilling prophecy. Banks hold long-dated, illiquid assets against short-dated, redeemable liabilities. That maturity transformation is the service they sell — and the vulnerability they carry. If every depositor asks for cash at once, no bank can pay, however sound its loan book.
The mechanism is game-theoretic rather than fundamental. Each depositor’s rational move depends on what they believe other depositors will do. In the good equilibrium, everyone leaves their money in place and the bank is fine. In the bad equilibrium, the mere expectation of a run produces one. News, rumour, and the sight of a queue outside a branch are enough to flip the system from one equilibrium to the other.
The classic sequence appears in the Panic of 1907, where trust companies outside the clearing-house system faced queues that spread institution by institution, and in the American banking panics of 1930–1933, where thousands of banks closed in successive waves. The 2023 US regional banking stress showed the modern variant: uninsured deposits moving by phone at a speed no nineteenth-century teller ever faced, emptying Silicon Valley Bank in hours rather than days.
Two standing remedies emerged from the historical record: deposit insurance, which removes the incentive to join the queue, and a lender of last resort, which makes the queue pointless by ensuring cash is always available. Both work by changing beliefs, not balance sheets — which is why credibility is the whole game.