Why Banks Fail

Sergio Correia, Stephan Luck, and Emil Verner. "Failing Banks." Federal Reserve Bank of Richmond Working Paper No. 25-04, June 2025.

Why do banks fail? One popular view emphasizes the role of bank runs, when a bank's customers collectively withdraw their money on the belief that the bank will soon fail. Bank runs, according to this view, can cause otherwise healthy banks to fail — the panic itself creates a self-fulfilling prophecy. This is often cited as an important cause of bank failures during the Great Depression, the 2008 financial crisis, and the 2023 U.S. banking crisis. A recent paper, however, casts serious doubt on this narrative.

Sergio Correia of the Richmond Fed, along with Stephan Luck of the New York Fed and Emil Verner of the MIT Sloan School of Management, built a dataset with balance sheet information for most commercial banks in the United States operating between 1863 and 2024. Of the roughly 37,000 banks in the dataset, over 5,000 failed at some point in the sample. This allowed the researchers to study bank failures throughout a long history of the U.S. banking system, including before the founding of the Federal Reserve System and the Federal Deposit Insurance Corporation (FDIC). Since the Fed and the FDIC were largely designed to prevent bank runs, this large dataset allowed the researchers to study bank failures in an environment where runs were more common.

Correia, Luck, and Verner found that failing banks were characterized by weak and deteriorating fundamentals, meaning they tended to be financially unhealthy prior to failure. These banks see a rise in nonperforming loans in the years leading up to failure, forcing them to allocate more money to cover the losses from uncollected loan payments. The banks were more likely to rely on noncore funding sources like time deposits and brokered deposits, which tend to be more expensive and riskier than other sources. These factors reduce bank profitability, making it more difficult to meet financial obligations.

A result of these findings is that bank failures are highly predictable by weak bank fundamentals. For example, a bank in the 95th percentile of insolvency risk and noncore funding reliance has a 25 percent chance of failure within the next three years — 10 to 25 times higher than the probability of failure for the average bank. Bank failures were highly predictable both in the modern sample and in the pre-FDIC era. This shows that bank failures are almost always and everywhere a result of weak fundamentals, rather than a consequence of sudden panics.

When bank runs do occur, do they cause the failure of otherwise healthy banks? To find out, the authors restricted the sample to banks with large deposit outflows prior to failure, which is indicative of a bank run. They found that even where runs occur, failures are no less predictable by a bank's fundamentals. In other words, when bank runs occur, they are unlikely to be the root cause of the failure. Rather, bank runs tend to occur at banks that are already financially unhealthy.

When a bank fails, a receiver tries to recover as much value as possible from the bank's assets. The authors argued that, if the bank were solvent up to failure, the value of the recovered assets would not fall significantly short of the bank's debts. However, failed banks had notably low asset recovery rates before the introduction of deposit insurance. The authors estimated that fewer than 8 percent of failed banks in the pre-FDIC era could have plausibly been solvent and forced down by a run.

Finally, the authors examined various reports on bank failures between 1863 and 1937 from the Office of the Comptroller of the Currency (OCC). Despite the prevalence of bank runs during this period, the OCC attributed less than 2 percent of bank failures to runs. Economic conditions, losses, and fraud were found to be the most common causes of bank failures. This is further evidence that deteriorating asset quality and poor fundamentals were much more significant than bank runs alone.

Using a comprehensive dataset of U.S. bank failures, Correia, Luck, and Verner determined that bank failures are predominantly a result of poor bank fundamentals, contrary to the narrative that bank runs lead to the failures of otherwise healthy banks. When bank runs occur, it is generally at banks that are already unhealthy. Bank runs can be an immediate trigger of a bank's failure, but they are seldom the root cause. Because failures are highly predictable by a bank's fundamentals, these findings support proactive policies to prevent or mitigate bank failures, such as limiting dividend payments and the use of noncore funding for banks with weak finances. While bank runs can be extremely disruptive, polices designed to prevent bank failures must address fundamental solvency issues.