Podcast

Important Information:
Why Do Banks Fail?
Important Information:
Sergio Correia discusses his research on the causes of bank failures and his use of a novel set of historical data to evaluate two different but intertwined potential causes — insolvency of the bank and a run on deposits. Correia recently joined the Federal Reserve Bank of Richmond as a senior economist.
Transcript
Tim Sablik: My guest today is Sergio Correia, a senior economist at the Richmond Fed. Sergio, welcome to the show.
Sergio Correia: Thanks for having me, Tim.
Sablik: We're going to be discussing a recent working paper that you wrote with Stephan Luck and Emil Verner, which tries to answer a question that might initially seem fairly straightforward, and that's "Why do banks fail?" As you and your co-authors lay out in the paper, though, the answer to this question is not at all straightforward.
Economists have developed two competing theories to explain bank failures. While those two theories aren't mutually exclusive, they emphasize different causes. So, to start, could you explain what those two theories are?
Correia: There's a classic debate between what we call liquidity stories and solvency stories. Every time there's a banking crisis, you're going to hear about this debate.
One, you have the solvency view. In this view, banks fail for almost the same reason as a restaurant fails. But why does a restaurant fail? It fails because it's empty. It's losing money day after day [and] it has no customers. Every dish it serves, it loses money. In the same way, a bank [may] lose money every quarter because of either credit risk — because it made bad loans — or because of interest rate risk — because it invested poorly. So, under this view, banks are just going to lose money until they are insolvent and then they close.
Then you have the other view, which is the bank runs view, which was famously formalized by Diamond and Dybvig. It says that banks can fail even if they're healthy, even if they are solvent.
Sablik: Could you tell us a little bit about the history behind the development of these two theories?
Correia: The bank runs theory became very prominent in the early '80s when Diamond and Dybvig developed their model. It really captured the popular imagination. People remember these stories from the Great Depression, the lines of panicked depositors waiting outside banks.
On the other hand, you have a solvency view that was always around. But it was really strengthened by the work of economic historians who went back into the archives in order to get hard data and see what exactly caused previous bank failures. For instance, Charles Calomiris and Joe Mason looked into the 1932 Chicago banking panic. They found that the banks that failed were the ones that were already weak, that were losing money, even before any panic happened. So, although the idea of a random panic is appealing, historical work has suggested that fundamentals are what's driving most of these failures.
Sablik: As you mentioned, these theories have been around for some time. Why has it been difficult to determine which of these theories does a better job of explaining bank failures?
Correia: It's a bit of a chicken and egg problem. If you think about a bank failure, banks are going to be losing money, losing money, and at some point everyone's going to realize it's losing money and it's going to fail. So, everyone runs to the bank. So, whenever you see a bank that fails, you often have these two together — you have losses and you have runs. Then, it's hard to distinguish what's the actual root cause and what's the immediate trigger.
Then you have another challenge, at least in the modern era, that we have a lot of provisions and government interventions to prevent bank runs. For instance, deposit insurance and the FDIC are designed to prevent depositors from running to the bank because they're guaranteed they're not going to lose money, at least small depositors. Therefore, we don't observe panic runs nowadays. Somebody can argue, oh, that's because of all these government programs. That's why we have to go back in time and see how things were before the safety nets were in place and see, in those cases, what was the main driver of bank failures.
Sablik: How did you and your co-authors approach trying to determine whether bank failures are primarily the result of runs or insolvency?
Correia: We took a comprehensive look at bank failures, not just a single crisis or a regulatory framework. We tried to get historical data from multiple regulatory eras, especially before the Federal Reserve and the FDIC. This allowed us to study in an environment where runs were much more plausible as a cause for banking failures.
With this data, what we look for is systematic patterns in the years before failures. If failures are caused by self-fulfilling runs that nobody could have predicted, then failures should be unpredictable. On the other hand, if you can look at a bank's publicly available information and see that there's a high chance of bank failure years before that happened, then it's more likely that failures are because of solvency issues and not because of random events.
Sablik: For your paper, you and your co-authors are looking at a really long range of data going back to 1863, more than 150 years. Where did you get the data that you used?
Correia: That was one of the most ambitious projects I've done.
One of the bread-and-butter datasets that we have at the Fed is the call reports, which have basically the balance sheets of every bank. The other bread-and-butter is the NIC [National Information Center] tables that have every event that happens to any bank: banks open, banks closed, mergers, etc.
This we have for the modern period. We don't have [this data] for the periods before the FDIC and Fed started. What helped us a lot was that the OCC [Office of the Comptroller of the Currency], which was founded in 1863, every year submitted its annual report. And in the annual report they had tables with balance sheet information for every national bank and also lists every new bank, every merger, etc. So, we saw these annual reports and said, "Can we do something with this? Can we make sense of the data and put it into Excel, into a database?"
It took quite a lot of effort — OCR, Python scraping, a lot of hand checks. Ultimately, we were able to recreate these two datasets — call reports and NIC tables — for the first 80 years of the national banking era.
Sablik: Once you gathered all this data, what did your sample consist of? How many banks and how many bank failures?
Correia: We have around 300,000 call reports in the historical era for about 15,000 national banks. I don't remember off the top of my head the number of failures, but we had about 50,000 events which include closures of any type — both failures and non-failures — as well as new banks and mergers. We have about 25,000 banks in the universe of banks regulated by the federal regulators above a certain threshold in the modern era.
Sablik: What are some of the common characteristics that you found among banks that fail in your dataset?
Correia: We found three remarkably consistent facts. First, failing banks don't just collapse overnight. They start deteriorating years before they actually fail. You can see this in their solvency, in their profitability.
Second, as banks get weaker, they shift their funding sources. Typically, a healthy bank funds itself with core deposits like savings accounts, which are their bread and butter. Those are cheap and those are stable; banks love them. But as banks become riskier, they shift away from this safer core funding into risk sensitive and more expensive funding such as time deposits, wholesale funding, etc. That, in turn, also eats into their profitability. So now they're losing money, not just because of credit losses but because their funding is more expensive. This is a signal of distress.
Lastly, there's typically a boom-and-bust cycle in their assets. There's a period of rapid growth in the decade or so leading up to failure, then you have a stagnation or a bust in their assets, and finally they collapse. So, in a way, the boom acts as the seed for the imbalances, the risks that ultimately are going to lead to the bank's failure. For instance, the last two major banking failures in 2023, these banks had growth rates over the last five years of 300, 400 percent. This is common, not just in historical periods but also in the modern era.
Sablik: Based on your findings, which theory of bank failure seems to most fit the data — the bank runs view or the solvency view?
Correia: These three facts seem consistent with the fundamental reasons for bank failures. The data doesn't match up that much with runs causing most of bank failures. I'm not talking about every case. I'm just talking about the majority of the failures. In the cases where you have deposit outflows before failure, it's often on banks that already had a lot of credit losses, not on the healthy banks. So, it's not that deposit outflows led to the failure.
This doesn't mean that panics and runs can't be disruptive. They can be extremely hurtful and they can damage the local economies and the overall financial system. They have large costs. But, ultimately, what we are finding is that the failures of the banks itself are caused by fundamentals, and that has implications on how we can prevent this failure.
Sablik: That's a perfect segue. Maybe you can talk about whether your findings suggest any lessons when it comes to bank supervision.
Correia: If fundamentals are the main reason why a bank fails, what we want to do is to ensure in supervision that banks are healthy and they are well capitalized.
We have some work on this in a new paper, which is called "Supervising Failing Banks." The idea basically is to look at how supervisors work and what exactly do they do when identifying risky banks. We see that they identify risky banks. They see that these banks actually recognize the losses. Ultimately, they close these banks and help to minimize losses to taxpayers and the economy.
And, we find something fascinating. The balance sheet information that markets rely on, that we rely on in our paper, etc, gets revised thanks to the work of examiners. So, whenever you have an examiner at the bank, especially in a failing bank, they revise downwards stuff like capital or earnings.
Sablik: That sounds really interesting. We'll definitely have to have you back on to talk about that paper as well. But Sergio, thanks so much for joining me today to talk about why banks fail.