Bank Runs and Panics: A Primer
“… you're thinking of this place all wrong. As if I had the money back in a safe. The money's not here. Your money's in Joe's house .... and a hundred others.” George Bailey, Manager, Bailey Brothers’ Building and Loan; in Frank Capra, It’s a Wonderful Life.
A bank promises its clients immediate access to cash. Depositors can redeem their funds on demand at face value—first come, first served. Other short-term creditors can do the same, albeit at varying speeds, by not rolling over their loans. And, households and firms that pay a fee for a credit commitment can take down their loans at will.
For banks that hold illiquid assets—like the Bailey Brothers’ Building and Loan in Frank Capra’s classic film (see opening quote)—these promises of liquidity on demand are the key source of vulnerability. The same applies to other financial institutions (de facto or shadow banks) that perform bank-like services, using their balance sheets to transform illiquid, longer-maturity, risky assets into liquid, short-maturity, low-risk liabilities.
A bank run occurs when depositors wish to make a large volume of withdrawals all at once. A bank that cannot meet this sudden demand fails. Even solvent banks—those whose assets exceed the value of their liabilities—fail if they cannot convert their assets into cash rapidly enough (and with minimal loss) to satisfy their clients’ demands.
A banking panic is the plural of a bank run: when clients run on multiple banks. We call the spread of runs from one bank to others contagion—the same term used to describe the spread of a biological pathogen.
The modern history of bank runs and panics begins in the 17th century and extends to the present (for a chronology, see here and here). No country is immune: even advanced economies with sophisticated financial systems experience banking crises (see Laeven and Valencia). Importantly, these disruptions are typically associated with some of the deepest peacetime economic downturns on record (see chart).
Advanced economies: Output loss (percent of potential GDP) in the largest banking crises since 1970
Notes: Output loss is the cumulative difference from potential GDP over the three years from the start date of the crisis (shown in parentheses beneath each country).
Source: Laeven and Valencia, Systemic Banking Crises Database, 2018.
In this primer, we characterize the sources of bank runs and panics, as well as the tools we use to prevent or mitigate them.
Theory of Bank Runs and Panics. The modern theory of bank runs originates with Diamond and Dybvig. In their model, the fragility of a bank comes directly from the liquidity services that it provides. While holding illiquid assets, the bank offers risk-averse depositors immediacy: namely, they can make withdrawals at face value on demand at any time. While the bank is solvent—the value of its assets exceeds that of its liabilities—it is illiquid because it cannot convert its assets immediately into cash without suffering large losses. In this setting, the sequential process of redeeming deposits at face value creates a first-mover advantage: those who get to the bank first get paid in full, while those who are patient (or just slow) may receive nothing.
In the Diamond-Dybvig model, bank runs reflect a shift from a good equilibrium in which nobody runs and withdrawals occur randomly over time, to a bad one where impatient depositors all try to make withdrawals simultaneously. The good equilibrium, where everyone trusts the bank, is fragile—just as it was with George Bailey’s Building and Loan. This means that runs are self-fulfilling: when some depositors believe that others will run, they have an incentive to run first. As Doug Diamond later put it, “even sunspots could cause runs if everyone believed that they did.”
In practice, however, sunspots generally do not cause runs (see our earlier post). Instead, runs arise when something happens that casts doubt on the value of the bank’s assets. In the U.S. national banking era (from 1863 to 1913), runs typically coincided with the peak of business cycles when—looking ahead—bank creditors might anticipate the rise of loan defaults (see, for example, Gorton). And, dramatic shocks with significant economic impact sometimes coordinated a rapid deterioration of expectations. Bruner and Carr attribute the origins of the U.S. panic of 1907 to the liquidity squeeze triggered by the San Francisco earthquake of 1906.
In the Diamond-Dybvig model, the only uncertainty about the value of the bank’s assets is whether the there will be a run. In their stylized setup, the run itself forces banks to liquidate its assets at fire-sale prices, giving depositors an incentive to run. In a more realistic setting, the difficulty creditors have in observing the value of the bank’s assets makes the bank vulnerable to runs—a difficulty that intensifies in the aftermath of a large shock.
This “hidden attribute” problem, in which a lender cannot costlessly observe the creditworthiness of a borrower, is a classic feature of adverse selection (see our earlier primer). Not knowing the value of the bank’s assets after a shock, depositors question the solvency of the bank. They also may worry that the bank is barely solvent, with only a slim equity cushion available to absorb losses from a forced fire sale (or a further shock). The smaller the perceived cushion, the greater the incentive to run. Importantly, a first-mover advantage makes it prudent to redeem deposits before others do so.
Not surprisingly, adverse selection is what causes bank runs to trigger panics. When depositors in one bank observe a run on another bank, they naturally question the solvency of their own bank. They cannot be sure because their bank is like a black box: they cannot costlessly observe the value of its assets. Put differently, the news about a bank run alerts everyone to the fact that there may be other “lemons” among the universe of banks.
Importantly, depositors need not believe their bank is insolvent to run. They need only fear that it is has insufficient capital to absorb the losses resulting from a fire sale triggered by a run. That fear will be markedly greater if the banking system as a whole is viewed as undercapitalized. In this setting, a destabilizing feedback arises in which runs spark fire sales that aggravate the depletion of capital and make the banking system even more vulnerable to shocks. To maintain bank resilience in the face of such uncertainty, there is a “market-imposed” minimum of equity financing needed.
Like biological contagion, financial contagion arises from a “common exposure” to a shock. However, unlike biological contagion, which typically results from direct transmission, financial contagion need not take the form of a sequential default cascade arising from exposure to a specific failing counterparty. Instead, it may be a common shock—such as a broad plunge of asset prices—that raises doubts simultaneously about the value of assets at many banks. The shock makes them all collectively vulnerable, boosting the cost of funding and triggering fire sales. Another difference is that, since it usually requires a period of incubation, the spread of disease takes time. Financial contagion can occur extremely quickly.
Preventing or Mitigating Runs and Panics. The primary aim of the government financial safety net is to prevent or mitigate bank runs. The first component of the safety net is the lender of last resort. Following Bagehot’s 19th century dictum, a modern financial system needs a central bank that is prepared to lend without limit to solvent firms against good collateral at a penalty rate (see Tucker). The designers of the Federal Reserve based the system’s original mandate on this Bagehot-like principle: supply an elastic currency in the face of financial disruption. Provided that banks are well capitalized, the existence of a credible lender of last resort should deter bank runs.
The second component of the safety net is a government guarantee. Government guarantees make bank liabilities information-insensitive: that is, news about the performance of a bank’s assets does not alter the attractiveness of holding that bank’s liabilities. Unless depositors doubt the willingness or ability of the government to pay, there is no incentive to run. And, even if one bank were to fail, there would be no incentive to run on any other bank. Credible guarantees prevent panics.
Deposit insurance—originating in the United States with the FDIC and now in nearly 150 countries—is the most common form that this government guarantee takes. Governments also may guarantee other bank liabilities, as the FDIC did with its Temporary Liquidity Guarantee Program in 2008.
In addition to such explicit guarantees, governments also may provide implicit guarantees. For example, the perception that some banks and intermediaries are too big to fail can lead counterparties to anticipate the provision of a bailout when insolvency occurs, even if governments promise not to do so. Such promises lack credibility if a future government will have both the authority and a powerful incentive to renege (see our primer on time consistency).
Government guarantees create incentives for the guaranteed firms to take risk. That is, they foster moral hazard (see our earlier primer). As a result, there is a tradeoff between preventing and mitigating financial disruptions. While guarantees make bank liabilities information-insensitive, they encourage banks’ owners and managers to take on greater leverage and acquire riskier assets, increasing the vulnerability of the banking system to a shock.
Consequently, the tools to prevent bank runs and crises include self-insurance by intermediaries to limit the moral hazard created by government guarantees. Just as casualty insurance deductibles and co-pays encourage the insured to take care, bank capital and liquidity requirements give banks an incentive to manage the risks they undertake. These shock absorbers also limit the destabilizing feedback across the banking system when a particular bank comes under stress.
There is little reason to think bank runs and panics are going away. Even if governments required banks to finance all their risky assets with equity—what are commonly called “narrow banks”—risk-taking could simply shift to nonbanks that perform bank-like transformation of liquidity, maturity and credit (see our earlier post). At the opposite extreme, the provision of a government backstop for all bank liabilities would make banks into public utilities and undermine the key role of market discipline in allocating credit.
In addition to managing the tradeoff between preventing and mitigating financial disruptions, societies also face a challenge of containing panics while preserving the benefits of a competitive banking system. The more control governments’ exert, the less freedom banks have to act (and the less they face the consequences of their actions), the more stable the system will be. But a financial system that never experiences stress is one that is almost surely very inefficient at performing its fundamental tasks. Ultimately, societies must make a choice of where on these tradeoffs they want their banking systems to be.