Failures are overwhelmingly the result of three factors: deteriorating solvency over several years, increasing reliance on costly non-core funding and rapid growth during the decade before the failure, co-authors Sergio Correia, Stephan Luck and Emil Verner state in the study, published this month by the National Bureau of Economic Research.
Correia and Luck are economists at the Federal Reserve Board and the Federal Reserve Bank of New York, respectively. Verner is an associate professor of finance at the Massachusetts Institute of Technology's Sloan School of Management and a faculty research fellow at NBER.
Taken together, the three factors imply that it's possible to predict which banks are at the highest risk of failure, Verner said in an interview. That data should help mitigate that risk, he said.
"We're reinforcing this view that to prevent failures and crises, banks need to be really focused on solvency and capitalization," Verner said. "It's a better understanding of how to monitor risk."
The trio's research, which includes data going back to 1865, is the latest contribution to the discourse on bank failures, which has been a hot topic
Pulling information from call reports and other regulatory-agency sources, the researchers examined data for more than 37,000 banks, 5,111 of which failed.
The data does not include bank failures that took place between 1941 and 1958, the researchers noted. That's because the Office of the Comptroller of the Currency's annual reports to Congress, which were the source for bank-failure data before 1941, stopped including balance sheets that year, while call reports from the Fed are only available in a digital format starting in 1959.
Broadly, the researchers found that failing banks experience a gradual increase in insolvency and large unrealized asset losses as their profitability and capitalization decline. They also found that failing banks increasingly depend on expensive deposit funding, such as time deposits and brokered deposits, and that they tend to go through a "boom-bust" period that starts with extremely fast growth, often as a result of rapid loan growth.
The "boom-bust" pattern is especially evident between 1959 and 2023, in part because the growth of banks in the earlier period was limited by geography, and banks faced restrictions on lending against real estate, the paper said.
Rapid asset growth is usually a red flag, said Bert Ely, a bank consultant who studies failed banks and thrifts. "You grow too fast and you make mistakes along the way," he said.
At Silicon Valley Bank, which was based in Santa Clara, California,
The findings further quash the notion that deposit runs are a primary source of failures. While runs were larger during the period before deposit insurance — in the pre-1934 sample, deposits in failing banks fell by 12% on average versus around 2% on average between 1959 and 2023 — about 25% of the failures before 1934 had minimal outflows or none at all, according to the paper.
"There is a story that a deposit run can come out of the blue because there's a 'panic,' and even the word itself reflects some kind of overreaction," Verner said. "But we find that's not true. Typically, when there is a run, you can see it coming in terms of weaknesses in a bank."
Brian Graham, a partner at Klaros Group who advises banks on issues such as strategy, finance and capital, said the paper's conclusions line up with the current thinking about bank failures. In short, it's hard for banks to fail due to liquidity issues, and "this report demonstrates that," he said.
"It should be obvious, but it's great to have some statistical analysis to back that up," Graham said.
Still, the paper's analysis is missing one key element — interest rate risk — according to Graham and Ely. The way the researchers measure solvency doesn't include the impact of interest rate risk, and that can lead to some banks reporting equity that seems OK or strong but in reality is weak, Graham said.
Verner acknowledged the gap, saying in a follow-up email that he and his fellow researchers did not "account for the valuation effects that rising interest rates imply for the fall in the value of long-term assets" and adding that the group has not "looked at the specific role" of rising rates and monetary tightening.
The paper does look at interest expenses compared with interest income at failing banks, and it finds that net interest margins are stable leading up to bank failures. The group's "conclusion is that credit risk seems to be more important than interest rate risk for understanding the typical bank failure" in the past 160 years, though interest rate risk "certainly matters in some episodes and for some banks," including the bank failures last year, Verner said in his email.
Bank failures aren't always a bad thing, Graham said. Just as restaurants come and go, so too will banks if they aren't able to offer customers the products and services they want, he said.
"We operate implicitly as if we want zero bank failures, but that's not really the right spot for the economy," Graham said. "There's a level of failure greater than zero that's good, and sometimes we lose sight of that."
"No one wants a [globally systemic important bank] to fail, but if it's a $1 billion bank, the economy will do just fine," he said.
