Bank failures are a key feature of banking crises that impose large real economic costs (Bernanke 1983). The US has experienced repeated waves of bank failures (Figure 1). Each crisis revives a longstanding debate: Are bank failures caused by runs? Or are they mainly caused by deeper solvency problems, which may in turn trigger runs?

The answers to these questions have first-order implications for policy (Beck et al. 2024). If runs are the main cause of bank distress, deposit insurance and central bank emergency lending may suffice to prevent costly failures. For instance, after the failure of Silicon Valley Bank and other regional US banks in March 2023, many observers argued for improvements in lender-of-last-resort policy and expanded deposit insurance. If instead insolvency is the deeper problem, emphasis must shift toward bank capital, supervision, and risk management.

In a recent paper, we survey the research on the causes of bank failures (Correia et al. 2026a). We draw on the long-run history of bank failures, considering evidence from both before and after the introduction of deposit insurance. We argue that fundamental insolvency is usually a necessary condition for bank failure. Runs are rarely a plausible standalone cause of failure for otherwise solvent banks, but they can be an important trigger of failure at already insolvent banks.

Figure 1 Bank failures in the US, 1863–2024

Source: Correia et al. (2026a).

The theory

Bank failures can stem from two related but distinct sources. Under the liquidity view, a sudden wave of withdrawals forces a bank to liquidate assets at fire-sale discounts, rendering it insolvent. Runs can thus trigger the failure of otherwise healthy banks (Diamond and Dybvig 1983) or of weak but still solvent banks (Goldstein and Pauzner 2005).

Under the solvency view, losses on loans or investments erode a bank’s equity. Once bank assets are not worth enough to fully repay depositors, the bank is fundamentally insolvent. A run may then be the final trigger that forces closure. The run can determine when and how the bank fails, but it is not the root cause of the problem.

The evidence

Finding 1: Bank failures are always and everywhere related to weak bank fundamentals.

The debate over whether bank failures are caused by insolvency or illiquidity has a long history. In A Monetary History of the United States, Friedman and Schwartz (1963) argued that many bank failures during the Great Depression resulted from “self-justifying” runs on solvent banks. However, subsequent empirical work has placed more emphasis on poor economic and bank-level fundamentals. Studies using regional and bank-level data find that banks that failed during the Great Depression were more exposed to declining local economic conditions, were less well capitalised, held more illiquid assets, and relied more on wholesale funding than surviving banks (Temin 1976, White 1984, Calomiris and Mason 2003).

The crucial role of poor bank fundamentals extends well beyond the Great Depression. In a recent paper, we extend these findings across 160 years of US banking data, covering over 5,000 bank failures (Correia et al. 2026b). Failing banks consistently exhibit declining income and capitalisation, rising asset losses, and growing reliance on expensive funding in the years before failure. A common precursor to failure is rapid asset growth, usually from aggressive lending. These patterns hold for bank failures with and without runs. They also hold across institutional regimes with and without deposit insurance or a public lender of last resort. As a result, bank failures are substantially predictable based on weak bank fundamentals. More broadly, crises in which many banks fail are often a predictable consequence of deteriorating fundamentals.

Finding 2: Recovery rates suggest most, but not all, failed banks subject to runs were fundamentally insolvent.

The evidence clearly links bank failures to weak fundamentals. But this does not settle whether most failures are caused by runs on weak but solvent banks or on fundamentally insolvent banks. Recovery rates on failed bank assets provide new insights into this question. Before the introduction of federal deposit insurance in 1934, overall creditors recovered, on average, only 75 cents on the dollar, while unsecured depositors recovered only 66 cents on the dollar (Correia et al. 2026a, 2026b). This means that failed bank assets fell substantially short of covering debt claims, indicating that most failed banks were fundamentally insolvent. Runs on weak but solvent banks therefore accounted for only a modest share of national bank failures, unless one assumes that receivership itself destroyed substantial value. While runs were an important trigger of failure at insolvent banks, low recovery rates suggest they were less often the cause of failure for otherwise solvent banks.

Finding 3: Bank examiners emphasise poor asset quality and rarely attribute failures to runs.

What do bank examiners say about the condition of failed banks and the causes of bank failure? In the pre-deposit-insurance US banking system, OCC examiners’ assessments indicate that most failed banks held assets exposed to substantial losses. On average, examiners classified only 36% of failed bank assets as “good,” while 47% were considered “doubtful” and 18% “worthless”.

Furthermore, US bank examiners historically classified the cause of death for banks. In the OCC’s bank-specific cause-of-failure reports, the most common causes were poor local economic conditions, asset losses, and fraud. Runs and liquidity issues were cited in fewer than 20 out of over 2,000 cases.

The Great Depression is a partial exception. Federal Reserve Board’s classifications of Depression-era suspensions suggest that liquidity issues played a larger role than in other periods (Richardson 2007). But even then, examiners’ assessments remained pessimistic about asset quality. As a 1936 Federal Reserve report put it, “[in] our long, failure-studded history of banking, most of the institutions which suspended business were subsequently proved to be insolvent”.

Finding 4: Strong banks usually survive runs through various mechanisms, including interbank cooperation, suspension, and examination.

Why don’t runs cause solvent banks to fail? Part of the answer is that runs are more common at weak banks. But strong banks do sometimes experience runs (Correia et al. 2026c). These runs rarely lead to failure because strong banks can employ several mechanisms to avoid costly failure.

First, in some cases, owners would provide cash to credibly signal confidence in their bank, much like how George Bailey stops the run in It’s a Wonderful Life. Second, interbank lending can provide needed liquidity, as banks are often better informed about a peer’s true condition (Blickle et al. 2024). Third, in the historical US banking system, clearinghouses acted as quasi-central banks, issuing loan certificates to provide liquidity. Finally, during severe runs, banks would temporarily suspend convertibility, both to cool panics and to allow examiners to audit their financial statements and assess solvency. Together, these mechanisms reduce the scope for runs to force healthy banks into costly failure.

Policy implications

The finding that most bank failures stem from solvency problems has important implications for financial stability policy.

Deposit insurance, introduced at the federal level with the creation of the FDIC in 1933, sharply reduced failures with runs. However, because pre-FDIC failures were rarely caused by runs on healthy banks, deposit insurance has not eradicated waves of bank failures altogether. What it did change was the way banks fail. In the absence of depositor discipline, bank failure is now more often the result of supervisory intervention (Correia et al. 2025). This framework reduces the occurrence of potentially costly runs, but it also reduces the ex-post discipline that runs impose on insolvent banks. Without this discipline, there is more onus on supervisors to prevent the emergence of insolvent zombie banks that distort credit allocation and amplify losses over time.

Lender-of-last-resort policy can also help solvent banks survive panics. A natural experiment from the Depression, comparing the Atlanta Fed’s generous lending with the St. Louis Fed’s more restrictive approach, shows that liquidity support can reduce failures (Richardson and Troost 2009). But liquidity provision cannot fix insolvency. International evidence shows that even banking distress without bank runs can produce severe contractions in credit and output (Baron et al. 2021). Likewise, targeted liquidity interventions in the 21st century have never resolved bank distress on their own; balance-sheet restructuring was always required (Kelly et al. 2025).

If most bank failures are ultimately driven by insolvency, then higher equity capital plays a key role in making the banking system more resilient (Admati and Hellwig 2014). Better capitalisation reduces both the likelihood of failure and the scope for runs to cause damage. Effective supervision also plays a central role by ensuring that banks recognize losses and by identifying when recapitalisation is needed. In crises rooted in weak solvency, recapitalisation is arguably the most effective way to restore confidence in the banking system.

Conclusion

The long-run evidence on bank failures points in one direction: bank failures usually begin with bad assets, weak earnings, and deteriorating solvency. Runs can accelerate failure and worsen the damage, but by the time depositors head for the exit, the deeper problem is usually baked into bank balance sheets. Ensuring that banks are funded with adequate capital and avoiding reckless lending booms are the most reliable ways to prevent failures from reaching that stage.

References

Admati, A and M Hellwig (2014), “The Bankers’ New Clothes: What’s Wrong with Banking and What to Do about It.”

Beck, T, V Ioannidou, E C Perotti, A Sánchez Serrano, J Suarez, and V Vives (2024), “Addressing banks’ vulnerability to runs, part 1: Facts, arguments, and policy challenges”, VoxEU.org, 9 October.

Bernanke, B S (1983), “Nonmonetary effects of the financial crisis in the propagation of the great depression”, The American Economic Review 73(3): 257–276.

Baron, M, E Verner, and W Xiong (2021), “Banking Crises Without Panics”, The Quarterly Journal of Economics 136(1): 51-113.

Blickle, K, M Brunnermeier, and S Luck (2024), “Who can tell which banks will fail?”, The Review of Financial Studies 37(9): 2685-2731.

Board of Governors of the Federal Reserve System (1936), Bank Suspensions, 1892–1935.

Correia, S, S Luck, and E Verner (2025), “Supervising Failing Banks,” NBER Working Paper No. w34343.

Correia, S, S Luck, and E Verner (2026a), “Bank Failures: The Roles of Solvency and Liquidity”.

Correia, S, S Luck, and E Verner (2026b), “Failing banks”, The Quarterly Journal of Economics 141(1): 147-204.

Correia, S, S Luck, and E Verner (2026c), “Bank Runs With and Without Bank Failure”, arXiv preprint arXiv:2601.20285.

Diamond, D W and P H Dybvig (1983), “Bank runs, deposit insurance, and liquidity”, Journal of Political Economy 91(3): 401-419.

Goldstein, I and A Pauzner (2005), “Demand–deposit contracts and the probability of bank runs”, The Journal of Finance 60(3): 1293-1327.

Kelly, S, V Arnold, G Feldberg, and A Metrick (2025), “Ad Hoc Emergency Liquidity Programs in the 21st Century”, Journal of Financial Crises 7(1): 57-160.

Richardson, G (2007), “Categories and Causes of Bank Distress during the Great Depression, 1929-1933: The Illiquidity versus Insolvency Debate Revisited”, Explorations in Economic History 44(4): 588–607.

Richardson, G and W Troost (2009), “Monetary Intervention Mitigated Banking Panics during the Great Depression: Quasi-experimental Evidence from a Federal Reserve District Border, 1929–1933”, Journal of Political Economy 117(6): 1031-1073.