By Byron Gilliam
Bitcoin supporters often label fractional reserve banking a Ponzi scheme, arguing that any run can bring down even the soundest bank. But a new study, which uses a large language model to scrape newspaper reports of bank runs, finds that most runs die out before threatening the bank's survival. This is a direct challenge to the anti-bank narrative that has fueled much of crypto's appeal.
"You're thinking of this all wrong, as though I had the money back in the safe," says George Bailey in the classic film It's a Wonderful Life, as he tries to calm a run on his building and loan. He eventually saves the bank by using his own $2,000 and later receiving $8,000 from friends and clients.
Austrian economist Murray Rothbard would have let Bailey's bank fail. "Fractional reserve banking is a scam, a Ponzi scheme, a fraud," he wrote. Rothbard argued that banks create money out of thin air, and that if people truly understood this, they would all demand their money back at once, causing even the best banks to collapse.
This pessimistic view found academic support in the famous Diamond-Dybvig model of bank runs. Economists Douglas Diamond and Philip Dybvig showed that banks funding long-term loans with demand deposits are vulnerable to runs even if they are fundamentally sound. A run can become self-fulfilling: depositors rush to withdraw because they expect the bank to fail, forcing the bank to sell assets at a loss and actually fail.
But the new research, published on the Bank Runs website, suggests the model doesn't match reality. The researchers compiled a database of hundreds of historical bank runs, detailing how each started and ended. The surprising finding: "There are more bank runs that did not result in bank failure than those that did." Even among banks with "very weak" fundamentals, only 59% failed after a run. The strongest banks "rarely failed" even when hit by a run. The authors conclude: "The pattern raises doubts about a strong version of the view that liquidity problems alone can cause severe financial distress."
Diamond and Dybvig were right about one thing: expectations can trigger a run, and those expectations can be based on almost anything. The Bank Runs database is full of colorful examples:
- In 1910, a run on Los Angeles' Merchants National Bank started when boxer Jim Jeffries visited the bank, drawing a crowd of fans. "Dozens of depositors thought something was wrong and began cashing in deposits," the newspaper reported. "They didn't calm down until the boxer left." Jeffries was just opening an account.
- In 1924, a run on Metals Bank & Trust Company in Butte, Montana, began when someone overheard a joke bet that the bank wouldn't open the next day. The bank stayed open four extra hours to pay out deposits, until it was too dark to continue. The joke was about Lincoln's Birthday.
- In 1929, a run on Bay Ridge Savings Bank in Brooklyn started when a rumor spread that the president had died. The bank learned of the false rumor in advance and prepared $14 million in cash. The president had actually gone to Connecticut to have a carbuncle removed from his neck (he survived the surgery).
- In 1930, a run on Chicago's Independence State Bank began when a fight outside a restaurant two doors away drew police cars, sparking rumors of a bank run. Depositors withdrew over $1.6 million from a bank with $5.6 million in deposits, forcing state officials to close it.
In many cases, simply displaying cash stopped a run. In 1907, a bank averted a run by "displaying large piles of banknotes and cash on the counter for depositors to see." In 1857, a "profitless run" on an Alabama bank ended when depositors saw a "mountain of gold and a 'Redan' of silver" on the teller's desk. In 1924, a Brooklyn bank manager stopped a run by piling banknotes of up to $1,000 in the front window, "in a great heap," for all to see.
Did bank customers understand that no matter how high the cash pile, it wouldn't be enough if everyone tried to withdraw at once? The author suspects they did. But the historical record shows that runs rarely turn into catastrophes. Banks are tougher than the Bitcoin narrative suggests.
The study does not deny that runs can sometimes be deadly. The 1930 Chicago case shows that rumors and panic can bring down a bank. But the authors' research indicates that if a bank is fundamentally sound, it will almost certainly survive. A run is more like a stress test than a self-fulfilling prophecy.
(Byron Gilliam)