What causes bank failures? There are two main explanations. First, bank runs can cause bank failures when depositors collectively withdraw their funds from otherwise solvent banks. Bank runs have been cited as an important cause of bank failures during the Great Depression, the 2008 global financial crisis, and the spring of 2023. An alternative view is that bank failures are caused by poor fundamentals, such as realized credit risk, interest rate risk, or fraud, which trigger insolvency irrespective of whether a bank run occurs.
Our research examines which explanation is more common. We constructed a database with information on the balance sheets of most banks in the United States since the Civil War—around 37,000—of which more than 5,000 failed. Our data consist of a historical sample that covers all national banks from 1865 to 1941 and a modern sample that covers all commercial banks from 1959 to 2023. Thus, our database includes failures both before and after the founding of the Federal Reserve System and the introduction of deposit insurance from the Federal Deposit Insurance Corporation (FDIC). Hence, our data allowed us to study bank failures during historical periods when bank runs could plausibly have been a common cause of bank failures.
Our findings reveal three commonalities about failing banks. First, failing banks experienced a rise in nonperforming loans and deteriorating solvency several years before failure. Second, failing banks increasingly relied on expensive and risky sources of funding in the run-up to failure. Third, failing banks experienced rapid growth in assets during the decade before failure.
Furthermore, our research finds that bank failures were remarkably predictable using accounting metrics from publicly available financial statements that indicated deteriorating fundamentals. The probability of failure could be predicted using measures of insolvency risk and funding vulnerabilities. For example, a bank in the top fifth percentile of both insolvency risk and funding vulnerability had a probability of failure over the next three years of between 13 percent and 42 percent—a 10- to 40-fold increase in the probability of failure relative to the average bank.
Deposit withdrawals immediately before bank failures were much larger before the FDIC was founded. Prior to 1934, deposits in banks declined by 12 percent on average before failure (when deposit insurance became effective) but fell by only about 2 percent before failure between 1959 and 2023. Nonetheless, even before 1934—when depositors typically realized large losses when banks failed—bank runs did not precede all bank failures. In around 25 percent of failures before 1934, deposits did not decline at all or only declined minimally.
Furthermore, failures with large deposit withdrawals were at least as easy to predict as failures without large deposit withdrawals. Therefore, banks that failed after a run could have been identified as weak banks based on their financial fundamentals before the run happened. Moreover, our analysis finds that weak bank fundamentals predicted waves of banking failures in addition to individual bank failures. Thus, spikes in bank failures during systemic banking crises cannot be explained merely by runs.
Historically, the Office of the Comptroller of the Currency (OCC) provided the causes of bank failures in the United States based on assessments by contemporary bank examiners. While many banks experienced large deposit withdrawals in the run-up to failure before the founding of the FDIC, the OCC classified most failures as caused by losses, fraud, or external economic shocks. Despite popular narratives about banking panics playing a key role in the historical US banking system, runs and liquidity issues account for less than 2 percent of failures classified by the OCC.
Our research supports the finding that bank runs have not been a common cause of bank failures, even before deposit insurance. Theoretically, bank failures caused by runs on otherwise solvent banks should exhibit three characteristics. First, failures caused by runs should feature large deposit withdrawals before failure, as large deposit outflows are required to force an otherwise solvent bank into responses that reduce the value of its assets, such as selling assets at low prices to quickly generate cash. Second, these failures should exhibit relatively low losses on assets held in failure. The value of assets not liquidated before failure should only be minimally affected by failure if held to maturity; thus, if the bank was solvent before the run, then the assets of the bank in failure should not be too troubled. Third, bank failures caused by runs should not be easy to predict, as otherwise attentive depositors would withdraw their funds before the panic. However, failures that exhibit all three characteristics are not common—they make up less than 15 percent of all failures before deposit insurance. This suggests that deposit insurance has not eliminated bank failures caused solely by runs. Rather, bank runs were an uncommon cause of bank failures even before deposit insurance. Instead, our research finds that in the majority of bank failures, depositors either did not run or withdrew their funds from banks that were most likely already insolvent. Moreover, when runs on failing banks did occur, they typically happened to banks with weak financial fundamentals.
Taken together, our findings suggest that most bank failures result from deteriorating solvency rather than bank runs. The erosion of a bank’s equity causes either a run or a supervisory decision to close a bank, with runs being more common in the historical period we studied. Additionally, depositors seem slow to react to information about bank fundamentals, thus making failures highly predictable.
