The Right Question After the Line Forms

The subject of this series has been the passage from a bank run to a bank failure, and the final question is how to judge that passage without collapsing several different events into one. A run is a visible act of withdrawal. Failure is a permanent institutional outcome. Economic damage is a further consequence that may or may not follow. The most defensible conclusion is therefore conditional: a run can occur at a strong bank, but it becomes a severe and lasting crisis mainly when weak fundamentals, concentrated or fragile funding, and insufficient response mechanisms allow withdrawal pressure to pass through into failure and disrupted credit.

That answer is shorter than the evidence needed to support it. The point of the final chapter is to make the answer usable without turning it into a universal prediction. When a headline announces a bank run, the reader should ask what the withdrawals mean, how much resilience the bank had before they began, which resources could interrupt the sequence, and what happened after the immediate panic.

The shortest accurate answer

A bank run is not the same thing as bank failure because the two terms describe different points in a process. The run describes depositor behavior: a substantial number of people or organizations suddenly demand cash or another bank’s liabilities. Failure describes the bank’s fate: a suspension becomes permanent, a receiver or comparable resolution process takes control, and the institution no longer continues as an operating bank.

The historical record makes the distinction concrete. From 1863 through 1934, the New York Fed study identifies 3,984 recorded runs across bank types. Of those, 2,074 did not involve failure and 1,910 did. The database also contains 10,341 failures, many without a newspaper-recorded run. A run was therefore neither necessary for every failure nor sufficient to produce failure in every episode.

That does not make a run a minor event. In the national-bank sample, a run was associated with roughly a 38-percentage-point increase in failure probability, compared with an unconditional annual failure probability below 1 percent. The average effect is large precisely because a run puts a bank under unusual pressure. But the average hides the pass-through condition. In the weakest fundamentals decile, the increase associated with a run was about 59 percentage points; in the strongest decile, it was about 10 points. Within roughly 40 days, very weak banks reached about a 56 percent cumulative failure probability after a run, while very strong banks were around 9 percent.

The right summary is not “runs are harmless” or “runs cause failures.” It is: withdrawals reveal and amplify vulnerability, but bank condition and institutional response determine how far the damage travels.

A withdrawal wave is information, not yet a diagnosis

The first question after a run is what triggered it. A withdrawal wave may follow information specific to the bank, a deteriorating economy, a run at another bank, a false rumor, or a mixture of these. The visible behavior is the same—depositors reduce their balances—but the signal carried by the behavior differs.

A run at another bank can make depositors jittery. The New York Fed study finds that runs were more likely after a run on another bank in the same city, and that such runs were associated with a lower probability of failure than other runs. That pattern is consistent with information contagion drawing some healthy banks into a withdrawal wave. It does not prove that the target bank was sound; the reported failure probability remained material. It does show why the first bank’s failure cannot automatically be treated as evidence that every nearby bank has the same balance-sheet problem.

A rise in nonfinancial business failures has a different implication. It can signal that borrowers are failing and bank assets are deteriorating. In the historical analysis, a one-percentage-point rise in the state business-failure rate—roughly a one-standard-deviation increase—was associated with about a seven-percentage-point increase in the pass-through from a run to failure. The signal may change depositor beliefs, but it may also describe a real decline in the bank’s ability to repay.

False-rumor episodes provide a useful boundary case. The study identifies 53 national-bank runs that newspapers explicitly attributed to misinformation or confusion, without bank-specific adverse news, distress at other banks, systemic panic, or a local economic shock. Their failure association was about 11 percentage points, much lower than the average run. Strong banks in this small subset had approximately zero estimated failure probability. Yet a weak or fraudulent bank could still fail when a random run exposed problems that were already there.