Why a Bank Run Is Not the Same Thing as Bank Failure
A bank run and a bank failure are often reported as if they were two names for the same event. They are not. A run describes what depositors do: a substantial number of them suddenly try to convert their claims into cash or another bank’s liabilities. Failure describes what happens to the institution: a suspension becomes permanent, a receiver or comparable resolution process takes over, and the bank no longer continues as an operating intermediary. This series examines the relationship between those events, with a practical aim: to help readers judge when a withdrawal wave is a contained liquidity shock, when it exposes deep weakness, and when it becomes a broader economic crisis.
That distinction matters because the visible scene—a line at a branch, a rush of transfers, or a flood of alarming headlines—does not by itself tell us which outcome is coming. A run can be followed by a temporary suspension and reopening. It can be absorbed while the bank remains open. Or it can end in permanent closure. The same word, run, can therefore describe episodes with very different implications for depositors, lenders, and the surrounding economy.
One word, several institutional outcomes
The historical study at the center of this series uses a deliberately practical vocabulary. A run is an episode in which a substantial number of depositors suddenly demand cash or another bank’s liabilities. A suspension is a bank closing its doors, whether that closure is temporary or permanent. If the bank later resumes business, the episode includes a reopening. A failure is a suspension that becomes permanent. For national banks in the study, the operational marker is an OCC receivership.
These categories describe a sequence, not a verdict. A run is an episode of pressure. A suspension is a decision or condition in which ordinary payments stop. Reopening means that the interruption did not become permanent. Failure is the terminal institutional outcome. Keeping the terms separate allows us to ask a more useful question than “Was there a panic?” We can ask: What made this bank able or unable to absorb the pressure?
The distinction is visible in the study’s historical counts. Across all bank types from 1863 through 1934, the researchers identify 3,984 bank-run episodes. Of those, 2,074 did not involve failure and 1,910 did. The database also contains 13,772 suspensions and 10,341 failures. In other words, most recorded runs did not end in failure, while most recorded failures did not include a newspaper mention of a run. These facts do not make runs unimportant. A run still sharply raises the risk of failure. They show instead that the risk is conditional: a run changes the situation, but it does not determine the ending by itself.
Temporary suspension is especially important to this distinction. Among runs on national banks that did not end in failure, about 28 percent involved a temporary suspension. Closing the doors for a time can be a serious event—it may unsettle customers and reduce confidence—but it is not identical to permanent liquidation. A bank that reopens has passed through a different institutional state from one that enters receivership, even if both appeared in the news under the headline “bank run.”
Why the difference matters beyond terminology
The difference changes what we should expect to happen next. If a run always meant failure, then counting runs would be enough to estimate the damage. But the historical evidence shows that outcomes diverge. The New York Fed study finds that runs on weak banks are associated with much larger declines in deposits, lending, and manufacturing activity than runs on strong banks. That pattern makes the bank’s condition part of the explanation, not a detail to be added after the fact.
The timing of a run can also be misleading. A run may be the moment when a bank’s underlying weakness becomes public, the moment when a marginal institution loses the time it had been using to postpone closure, or the moment when an otherwise sound bank must demonstrate that it can meet withdrawals. In each case, the withdrawal wave is real. But its meaning differs. One episode may reveal a solvency problem; another may be a temporary information shock; a third may accelerate a failure that was already developing.
This is why “the run caused the failure” is often too simple a conclusion. It may be true that withdrawals forced an institution into a damaging sequence. It may also be true that the bank was already weak and that the run determined the timing of an outcome that was likely to arrive later. The evidence can support a run as a trigger or amplifier without showing that it was the root cause of every loss. That distinction becomes decisive when we move from the bank itself to local lending and real economic activity.
The historical rates reinforce the conditional picture. For national banks, the average annual probability of a run in the 1863–1934 sample was 0.39 percent, compared with 0.89 percent for failure and about 0.20 percent for a run without failure. In years classified as banking crises, the run rate rose to 1.42 percent, from 0.23 percent outside crisis years. The failure rate rose as well, from 0.49 percent to 3.45 percent. Runs cluster in bad periods, but clustering does not erase the difference between an event that produces pressure and an institution that permanently closes.
Three reasons depositors may run
A visible withdrawal wave can have more than one underlying explanation. The historical study uses three broad theories to organize the possibilities.
The first is a non-fundamental or self-fulfilling run. Depositors may withdraw because they expect other depositors to withdraw. Their collective action can create a liquidity problem even when the bank might otherwise have remained viable. This mechanism allows runs on relatively healthy banks after a small shock. It makes the run itself look like the decisive event.
The second is a fundamental-based panic. Here, the bank’s condition matters more directly. Poor capitalization, high leverage, limited liquid assets, or dependence on more expensive and risk-sensitive funding can make depositors more likely to withdraw. The run is still a sudden event, but it is responding to a bank that has less room to absorb it.
The third is an information-based run. Depositors receive bad news about the economy or the banking system and struggle to distinguish strong banks from weak ones. A decline in the stock market, rising business failures, or a run on another bank can change how safe depositors believe their own bank to be. This mechanism can send money out of a sound bank as well as a troubled one. It also explains why a run need not end in failure: the bank may be able to reassure depositors, obtain support, or otherwise show that the alarming signal does not describe its own condition.
These theories are not three labels that the evidence assigns with perfect certainty. They are competing expectations. A self-fulfilling account predicts that even a strong bank can be pushed toward failure. A fundamentals-based account predicts a closer relationship between weakness and the onset of a run. An information-based account predicts runs after broad signals and allows withdrawals from banks that ultimately survive. The value of the framework is that it tells us what to look for after the headline appears.
What a broad historical record adds
For decades, discussion of bank runs often depended on famous crises or detailed accounts of particular episodes. Those cases are valuable, but they cannot by themselves tell us whether the same pattern recurs across ordinary banks, smaller cities, quiet periods, and runs that never became failures. The New York Fed study expands the comparison by using historical newspapers to assemble a near-universe of recorded bank distress in the United States from 1863 to 1934.
The researchers combine newspaper accounts with bank balance sheets, official closure records, local business failures, and measures of city manufacturing activity. The result is not simply a longer list of newspaper stories. It links the occurrence of a run to the condition of the bank and to what happened afterward. That makes it possible to compare a run that ended in failure with one that ended in reopening, and to compare both with episodes that involved no recorded run.
The database is substantial, but it is not omniscient. A newspaper-recorded run is a binary event in the study; it does not directly measure how much money left or how long the pressure lasted. Smaller-bank events were less likely to be reported. Some distress could remain quiet. The researchers validate the database against four ground-truth exercises, with national-bank matching rates around 95 percent overall, but validation supports usefulness rather than perfection. A missing newspaper account is not proof that no run occurred, and a recorded account is not a complete measure of severity.
Those limits make the central result more credible, not less, because the result does not depend on pretending that the data are flawless. The evidence is strongest when the question is one the dataset was designed to answer: how the likelihood and consequences of recorded runs vary with bank fundamentals, public signals, failure, and local conditions. It is weaker when we try to infer an unobserved depositor motive or apply historical rates mechanically to a modern institution.
The question to carry forward
When a bank run is reported, four questions are more informative than the headline alone. What triggered the withdrawals: evidence about the bank, bad news about the wider system, or confusion? What was the bank’s condition before depositors moved? Did the bank remain open, suspend temporarily, reopen, or fail permanently? And what happened to deposits, lending, and economic activity afterward?
The first chapter of this series therefore does not ask readers to choose between “runs are harmless” and “runs cause crises.” Both slogans are too broad. The evidence supports a more precise judgment: runs can occur at strong and weak banks, but weak fundamentals make them more likely to occur, more likely to pass through into failure, and more likely to be followed by severe economic consequences. Stabilizing mechanisms can interrupt that path, while a run can still leave lasting damage even when the bank survives.
The next step is to examine the pressure itself. Once depositors demand payment, what happens to a bank’s cash, loans, and available options? That mechanism explains why the same visible act—withdrawal—can end in survival for one bank and permanent failure for another.