More On: Douglas Diamond
Cryptocurrency transactions need to return to its decentralized roots.
Douglas Diamond, Philip Dybvig, and Ben Bernanke shared the Sveriges Riksbank Prize in Economic Sciences on October 10 for their research on banks and financial crises.
The prize committee said that Diamond and Dybvig's research showed that borrowing short and lending long makes even good banks vulnerable to runs, because even a rumor of a run can turn into "a self-fulfilling prophecy." "These dangerous dynamics can be stopped," the committee wrote, "if the government insures bank deposits and acts as a lender of last resort."
One month later, on November 10, FTX, the world's second-largest cryptocurrency exchange, stopped letting people withdraw their money after a wild three-day run in which $6 billion was withdrawn. The next day, FTX filed for bankruptcy, leaving its 100,000 remaining creditors with between $10 billion and $50 billion in unpaid debts and causing chaos in other cryptocurrency markets.
Due to the way these events happened at the same time, many people saw the FTX embroglio as a perfect example and further proof (if it was needed) of Diamond and Dybvig's theory. The author of Money Changes Everything, William Goetzmann of Yale, called FTX's collapse "A Diamond-Dybvig moment" on Twitter. Campbell Harvey of Duke University, whose area of expertise is decentralized finance, called it "a classic Diamond-Dybvig bank run." Matt Levine, writing for Bloomberg, also said that the fact that FTX lent (or "repurposed") its clients' money made it "vulnerable to runs, Diamond-Dybvig, It's a Wonderful Life, etc., everyone knows this." Dozens of other reports called FTX's problems a "liquidity crisis," which is the kind of crisis that Diamond and Dybvig's famous article says causes banks to fail.
The problem with this is that the fall of FTX doesn't show anything about the Diamond-Dybvig story. Most runs on commercial banks don't happen either. And the difference between the kind of bank run Diamond and Dybvig describe and most real bank runs, like the one on FTX, is very important. For example, government-mandated deposit insurance or public last resort lending may be the best way to stop or respond to a Diamond and Dybvig-style run, but other reforms, like market-based ones, are better at stopping the kinds of runs that happen most often in the real world, like the one on FTX.
Of course, telling you that is a lot easier than getting you to agree. This is why this post is so long.
Illiquidity and Insolvency
To understand Diamond and Dybvig’s contribution, it’s essential to recognize the difference between a bank that’s insolvent and one that’s merely illiquid. Any bank that can fully pay off its creditors once its investments mature, or after liquidating them in an orderly fashion, rather than all at once, is solvent. But whether a bank is solvent or not, if it lacks cash with which to pay its current obligations, it’s illiquid.
The Diamond-Dybvig theory of bank runs is different from other theories because it tries (but doesn't always succeed) to show that banks are vulnerable to runs just because they are illiquid, which they will be if they borrow short to lend long. Diamond and Dybvig say that the possibility of runs on otherwise solvent banks is a reason for a lender of last resort, government-backed deposit insurance, or both. This is because runs can mess up financial arrangements that would otherwise allow patient and impatient depositors to share risk in the best way.
Diamond and Dybvig make it clear that their model is meant to show how runs can be dangerous to banks that are otherwise doing well. To do this, they assume that their hypothetical bank's investments are risk-free, and that if they are left to mature, they will bring in enough money to pay off all of its debts. Its problems don't come from the chance that its investments will go bad. Instead, it's because there's a chance that depositors will freak out, forcing the bank to sell those investments too soon. Diamond and Dybvig say that as long as trust is kept,
there can be efficient risk sharing, because in that equilibrium a withdrawal will indicate that a depositor should withdraw under optimal risk sharing. If agents panic, there is a bank run and incentives are distorted. In that equilibrium, everyone rushes in to withdraw their deposits before the bank gives out all of its assets. The bank must liquidate all its assets, even if not all depositors withdraw, because liquidated assets are sold at a loss.
Importantly, it’s only “because even ‘healthy’ banks can fail” that “bank runs cause real economic problems.” That’s because runs on insolvent banks—that is, banks that have invested badly—don’t necessarily spoil an otherwise optimal risk‐sharing arrangement. Instead, by hastening the closure of banks that have invested imprudently, or otherwise mismanaged depositors’ funds, they can actually be efficient means for limiting waste.
As if to make this distinguishing feature of their theory especially clear, Diamond and Dybvig contrast it with Irving Fisher’s explanation of bank runs, in his 1911 book The Purchasing Power of Money. According to Fisher, it’s “the failure (or prospect of failure) of firms that have borrowed heavily from banks” that
induces fear on the part of many that the banks will not be able to realize on these loans. Hence the banks themselves fall under suspicion, and for this reason demand cash. Then occur ‘runs on the banks,’ which deplete the bank reserves at the very moment they are most needed.
Diamond and Dybvig compare Fisher's theory to their own and say that he thinks runs happen when a bank's "liquid but risky assets no longer cover the nominally fixed liability (demand deposits), so depositors pull out their money quickly to cut their losses." In their model, a run is caused by "a shift in expectations, which could be caused by almost anything, which is consistent with the seemingly irrational behavior of people running on banks." In fact, the only thing that can't cause a run on a Diamond and Dybvig bank is for depositors to find out that the bank is actually broke, because a Diamond and Dybvig bank is built to be a good bank. The only "shock" that its depositors should worry about is one that they could cause by losing their cool.
In the literature, Diamond‐Dybvig type runs are often referred to as “random” runs; in contrast, Fisher‐type runs are “information‐based.” In a 1988 paper, Charles Jacklin and Sudipo Bhattacharya offer a formal theory of information‐based runs, showing how their likelihood and welfare implications differ from those of runs of the Diamond and Dybvig type. Among other things, they show that, according to their model, runs aren’t a problem so long as banks’ investments are reasonably liquid and depositors aren’t highly risk averse. Runs can even prove beneficial, by forcing banks to liquidate investments that seem likely to go south. Finally, there are cases in which even risk‐averse depositors won’t be inclined to run on a bank despite hearing bad news about it. Although Jacklin and Bhattacharya’s findings don’t necessarily rule out a role for deposit insurance or last resort lending in discouraging information‐based runs, they point to other options that can work as well, or better.
One Bad Bank
Even though cryptocurrency exchanges aren't banks, some of them, like FTX, work like banks. They use the money they get from their customers, which they promise to return on demand, to make risky investments. So, their customers can run away if they have a good reason to think that some of their money has been gambled away.
Just a quick look at what led up to the run on FTX is enough to show that there was nothing crazy about it. Alameda Research was a trading company started by Sam Bankman Fried, who also started and ran FTX. On November 2, a disturbing report about Alameda Research came out. In the weeks before FTX went bankrupt, Alameda lost a lot of money and was almost certainly broke. But CoinDesk says that it filled the big hole in its $14.6 billion balance sheet at the time with FTT and other tokens that FTX made and lent to it. Alameda then put a price on the tokens based on their value on the bull's market. Even though FTX said it was just a custodian of its customers' cryptocurrency balances—that is, that the money it received from them was not its property and would not be lent—it made those customers Alameda's creditors, which meant they would be responsible if Alameda went out of business.
As soon as the truth came out, trust in both companies started to fall apart, and accusations of wrongdoing started to pile up. The run on FTX really got going when Binance, the largest cryptocurrency exchange in the world at the time, decided to sell about $500 million FTT tokens. Even though Binance later said it would buy FTX and save its creditors, it backed out of that deal on November 9. On the 10th, both FTX and its sister company, FTX.US, stopped letting customers withdraw money. On the 11th, both companies, along with Alameda and many other FTX affiliates, filed for bankruptcy in Delaware. At the beginning of November, one FTT token was worth more than $25, but as of this writing, it is worth $1.36.
So far, it looks like FTX's balance sheet only has a small amount of liquid USD assets, about $900 million. This is compared to about $8.7 billion in illiquid private equity and non-USD crypto token investments. FTX's liquidation value won't be known for a while, but calling its creditors' chances of getting their money back in full "slim" is probably an understatement. Recently, FTX account holder bankruptcy claims were selling for as little as ten cents on the dollar. In short, whatever kind of bank FTX was running, it was nothing like the one Diamond and Dybvig describe in their famous paper.
When FTX went down, it caused a lot of other things to happen, like runs on other exchanges. But so far, these also seem to be based on facts rather than fear. Most of them seem to be caused by the "high degree of interconnectedness" of cryptocurrency markets rather than just an epidemic of fear. For example, the failure of Sam Bankman Fried's firms hurt other cryptocurrency firms that had direct credit exposure to them or relied on them for financing. In turn, the firms that were directly hurt hurt other firms that were connected to them, and so on. But the waves only went so far: as Annie Lowrey of The Atlantic wrote on November 12,
the FTX debacle has thus far had no evident impact on the stock market, nor has it had any effect on the stocks of publicly traded financial firms. The Wall Street ‘fear index,’ a measure of financial volatility, went down a touch when FTX went down in flames. A number of firms wrote down or are expected to write down the value of their investments in FTX. But there is little concern about systemic risk, at least for now.
Also, cryptocurrency companies were not attacked without any reason. Fortune's Jeff John Roberts says that instead, there were both winners and losers. The winners included both centralized exchanges that "resisted the urge to chase crypto riches through leverage and offshore shell games" and decentralized exchanges that "are properly collateralized and can't be destroyed by the greed or folly of a single individual." It's hard to square this kind of bias with the idea that runs are caused by unfounded fear.
The Usual Story
Even though the run on FTX was different from most bank runs in many ways, it wasn't because it was based on information. Most runs on banks have also been runs on companies that seemed to be bankrupt or close to it before the runs started. In other words, Diamond and Dybvig's "irrational observed behavior" hasn't been seen very often, and what Matt Levine calls "a classic normal bank run" isn't normal at all.
It's possible that this is why the run on Bailey Brothers Building & Loan, as shown in the movie It's a Wonderful Life, is economists' favorite example of a Diamond-Dybvig type run, even though it's made up. It was used as an example by the Nobel Prize committee. But even that run doesn't really show how Diamond and Dybvig's theory works, because it starts when depositors find out that "Uncle" Billy Bailey lost $8,000 of the bank's cash—a lot of money for a "measly one-horse" thrift in the 1940s. Also, because thrifts didn't offer demand deposits back then, Bailey Brothers Building & Loan wasn't required to let its upset depositors cash out their balances. Mary Poppins, the second most well-known example of a Diamond-Dybvig bank run, has a run that is caused by pure panic. Noah Smith talks about only the two Hollywood runs in an article that praises Diamond and Dybvig's theory for how relevant it is.
Smith's explanation makes me wonder, "Which real-world situations match the theory?" Most studies by economic historians show that few people do. Even in the United States, which has had more than its share of bank runs, failures, and crises, this has been the case. For example, most of the banks that failed during the so-called "free banking era" (1837–1863) did not fail because of bank runs or even because of fraud. Instead, they failed because the value of their assets went down. Often, the troublesome assets were ones that bankers were forced by law to buy. There were definitely runs. But Jerry Dwyer and Iftekhar Hasan say that they were caused by other shocks and don't seem to have caused more permanent bank closings on their own.
After carefully looking at bank runs during the pre-Fed National Banking Era (1863–1914), Charles Calomiris and Gary Gorton came to the conclusion that they didn't fit the Diamond–Dybvig theory. Instead, they usually happened when "bank depositors would change their beliefs about how risky banks were in a reasonable way." Even so-called "panics" during the time of national banking were caused by "rational revisions in beliefs about bank performance" rather than "irrational behavior" on the part of depositors.
But these early events aren't what most people think of when they think of banks that would have been fine if depositors didn't panic and force the banks to close. People instead think about "the scary but rather unique experience of the Great Depression," as the late George Kaufman put it. Even so, most of the bank failures in the early 1930s were caused by things other than runs. As Kaufman points out, and as I've said in other places, they mostly
reflected the severe problems in agriculture from a continuing sharp decline in commodity and land prices after an even sharper run up. Ninety percent of the banks that failed…had capital of less than $100,000, had loans and investments of less than $1,ooo,ooo, and were located in towns of less than 5,000 in population. Even after adjusting for the sixfold [in 1983] increase in prices since that period, these were Ma and Pa banks by any measure and were unlikely to have been diversified greatly or managed professionally.
During the Great Depression, it wasn't just rural banks that went out of business. In some cases, like the Chicago banking crisis in July 1932, depositors didn't just run away from good banks. Still, Calomiris and Joseph Mason say that even in that case, the runs were "caused by fears of bank insolvency, not by depositors' outside demand for cash." If some good banks were hit by runs, it was because they were often rightly thought to have business with or be in the same boat as banks that were already in trouble. In other words, the runs were based on information, but the information wasn't complete, so some good banks were suspected even though they were safe.
Of all the bank runs that happened in the 1930s, the ones that started in February 1933 and ended when FDR declared a national bank holiday on March 6 are usually seen as evidence of a widespread panic. But Barry Wigmore argued convincingly not long after Diamond and Dybvig's article came out that those runs were caused by worries about what the next president-elect would do with the gold standard. As Roosevelt's inauguration day (then March 4) got closer, people worried that he would lower the value of the dollar if he didn't make a firm promise to keep the gold standard. So, people rushed to their banks to deposit money, not because they didn't trust the banks, but because they didn't trust paper dollars. When people got their hands on Federal Reserve notes, they went back to the Fed, especially the Federal Reserve Bank of New York, to trade them for gold. The New York Fed asked New York State Governor Lehman to declare a bank holiday, even though most of the big commercial banks in New York didn't want to. The rest of the country's banks had to close because of this holiday.
During the Great Depression, both rural and urban bank runs were led by sophisticated people with big bank accounts, not by people with small accounts who didn't know much about money. Martin Frost did a study for the Federal Reserve that looked at runs on 67 urban banks that closed at the time. He found that accounts of $25,000 or more made up more than 40 percent of withdrawals, even though they only made up 28 percent of deposits. In 1932, the president of Chemical Bank told the Senate Banking Committee, "The smart guy gets out first, and he's the big depositor." Smart people were thought to know which banks were good and which ones were not.
Many bank depositors haven't had a reason to run on their banks for a while now, thanks to deposit insurance. However, runs still happen, and in the vast majority of cases, they target banks that are thought to be insolvent, usually with good reason. In an article published the same year as Diamond and Dybvig's, Joseph Aharony and Itzak Swary look at bank stock prices to see if they show any effects from what were, at the time, the three biggest bank failures in U.S. history: the Franklin National Bank of New York (FNB), the United States National Bank of San Diego (USNB), and the Hamilton National Bank of Chattanooga (HNB). They come to the conclusion that "the data do not back up the pure contagion effect hypothesis." In particular, they found that the failures of USNB and HNB, both of which had done fraud and other bad things, had no effect on other banks. Aharony and Swary come to the conclusion that "the implication is that the failure of a dishonestly run bank, even a big one, does not have to cause panic and a loss of public confidence in the integrity of the banking system as a whole."
According to Jonathan Rose, what was true in the past was still true during the Great Financial Crisis of 2008. Large depositors led runs on Wachovia, Washington Mutual, and some smaller banks, and each one happened right after a shock that hurt the banks' stability or a revelation about it. Even though some of the runs were contagious, the spread was quite limited, both in terms of location and whether or not banks nearby had also been in the news for the wrong reasons.
History, in Theory
Even though history shows otherwise, since Diamond and Dybvig published their article, many, if not most, economists have tended to see every financial firm failure as a "Diamond Dybvig Moment." Robert Shiller, who also won the Nobel Prize, is a good example of this. In the beginning of his 2011 "Financial Markets" Open Yale Course lecture on "Banks," Shiller says that Diamond and Dybvig's 1983 work is not "a" theory of banks, but "the" theory of banks. We shouldn't be surprised, then, when he goes on to talk about banking crises and treats them all as examples of Diamond and Dybvig's theory, that is, as events that happened because bank depositors got scared and tried to take all their money out, causing "the whole banking system" to collapse.
As I've already said, the "sheer panic" explanation doesn't work well even for the 1933 banking crisis, which was different from most in that it affected the "whole banking system." Shiller can't use it to explain how other runs work either. Take his first example, which is about the Northern Rock Bank in the UK, which failed in 2007. This was the first major bank run in the UK since 1866. "A rumor started," Shiller says, "that Northern Rock had a lot of subprime securities and was going to go bankrupt, so people rushed to Northern Rock...people thought, 'Well, this is just like 1933.'" Then, according to Shiller, the run stopped when Bank of England Governor Mervyn King offered to help everyone.
To call Shiller’s account of the run on Northern Rock less than accurate is to be generous to a fault. As Hyun Song Shin explains in a 2009 article in the Journal of Economic Perspectives,
the storyline of the Northern Rock bank run does not fit the conventional narrative. On September 13, 2007, the BBC’s evening television news broadcast first broke the news that Northern Rock had sought the Bank of England’s support. The next morning, the Bank of England announced that it would provide emergency liquidity support. It was only after that announcement…that retail depositors started queuing outside the branch offices.
Nor did Northern Rock hold many subprime loans or securities. It suffered from its unusually heavy reliance on wholesale rather than retail (deposit) funding. The market for wholesale funding all but froze up after the BNP Paribas suspended redemptions on three of its funds in August 2007. The Northern Rock run was, therefore, “an event in the aftermath of [its] liquidity crisis…rather than the event that triggered its liquidity crisis. In this sense, the Northern Rock episode was not an old‐fashioned bank run of the sort we see in movies like It’s a Wonderful Life or Mary Poppins.”
Instead of being brought down by jittery retail depositors, Northern Rock’s fate was sealed “by sophisticated institutional investors.” As Hyun goes on to explain,
the irony of the images of Northern Rock’s retail customers standing in line to withdraw deposits is that retail deposit funding is perhaps the most stable form of funding available to a bank. Although retail deposits can be withdrawn on demand, bankers have been heard to joke that a depositor is more likely to get divorced than to switch banks.
Diamond and Dybvig’s research has been extremely influential, and that fact alone may justify their having received the Bank of Sweden’s prize for it. But as Shiller’s account of Northern Rock’s collapse makes clear, the theory’s influence hasn’t been entirely benign. When economists and policymakers, ignoring the work of economic historians, insist on viewing actual bank runs through Diamond‐Dybvig colored glasses, they see a warped version of history; and warped versions of history can inspire unwise policies.
Good versus Bad Runs
Even though Northern Rock's run wasn't a Diamond-Dybvig run, it was a rare case of a bank failing not because of bad investments but because it didn't have enough cash. As such, it shows how dangerous it can be when people aren't the same age. Even so, Northern Rock's liquidity problems were caused by its own carelessness. It was careless not because it used ordinary retail demand deposits to fund its investments, but because it didn't. Most bankers have known for a long time that "core" retail deposits are a much more stable source of funding than short-term wholesale funds. So, it's not clear that better solutions to Northern Rock's liquidity problem would have been more complete deposit insurance or more timely last-resort loans.
Indeed, in many cases in which a bank is about to fail because it has acted imprudently, its failure may be optimal whether runs hasten it or not: as George Kaufman explains (pp. 6–7), a central bank that keeps a reckless bank alive by lending it funds with which to pay off its anxious creditors can reduce rather than enhance social welfare, and especially so if the bank’s net worth becomes negative while it is being kept on support. When allowed to proceed, runs on insolvent banks keep them from becoming “zombie” institutions the costs of whose failure ultimately falls on taxpayers. In short,
bank runs do not deserve the bad reputation they have received. They did a dirty job in maintaining market discipline. But someone had to do it! Eliminating dirty jobs per se does not eliminate the problems that give rise to these jobs.
Even though it's true that when depositors don't know everything about their bank's investments, the failure of an insolvent bank can cause runs on some solvent banks, that risk justifies last-resort lending to the solvent banks, not the insolvent ones, and since lending to solvent banks isn't very risky, it's much more likely to come from the private sector. Lastly, there is a case for insuring unsophisticated depositors against losses even when those losses are caused by their banks' bad behavior. However, that case hardly justifies coverage of "$250,000 per depositor, per insured bank, for each account ownership category" and other policies that may keep bank depositors whole by passing the cost of bank failures on to people who are even less to blame and maybe even less well off.
Even though it was a big deal, FTX's failure seems to be a perfect example of a run on a financial institution doing its "dirty job" of keeping the market in check. It didn't change the best way to share risks; instead, it helped stop a move away from those arrangements that couldn't last. Even though the collapse of FTX shook up other crypto markets, it did not pose a serious threat to the system as a whole. Could a much bigger failure like FTX have done it? Sure, but the only way to make it less likely that a huge mess will happen is to let people who are involved in smaller ones take their lumps. In short, if the choices are the Diamond-Dybvig solution of government guarantees, with all the problems that come with them because of moral hazard, or doing nothing, it's probably best to do nothing.
Better than Bailouts
But those aren’t the only choices. Because Diamond‐Dybvig runs can happen for no reason at all, nothing short of blanket depositor guarantees, and the outright prohibition of all sorts of uninsured intermediaries that engage in maturity mismatching, can easily rule them out. In contrast, when runs are information‐based, whether they afflict sound financial firms or only unsound ones depends on the accuracy of the information available to those firms’ customers. The more accurate that information, the less likely it is that runs will harm otherwise solvent businesses.
It follows that there are ways to protect clients of cryptocurrency firms from losses without undermining market discipline. This can be done by requiring that uninsured cryptocurrency exchanges and token issuers routinely publish audited reports listing their asset holdings. The Stablecoin Transparency Act (S. 3970 and H.R. 7328) is a reform in this spirit specifically aimed at USD stablecoin issuers, as is an earlier proposal put forward by my CMFA colleagues.
A related reform is the "ring-fencing" of relatively liquid assets that can be used to back cryptocurrency balances. This ensures that holders of these balances will get their money back in full if a company goes bankrupt because of its or its affiliates' risky activities. Ring-fencing can be done by giving cryptocurrency firms' deposit-taking operations to a "bankruptcy remote" subsidiary or by putting a licensed third-party custodian in charge of protecting their customers' cryptocurrency balances.
Other good ways to change things don't need the government's help because the cryptocurrency industry and users are already doing them on their own. The first, called "Proof of Reserves" (PoR), uses cryptographic verification to show auditors of a centralized exchange where their clients' money is going or, in other words, what their account balances are backed by. PoR alone can't stop an unwise or dishonest exchange from making risky or illegal investments and failing as a result, but it can reassure customers of other exchanges that those exchanges aren't in the same boat. Soon after FTX went down, Binance said it would start using PoR on its accounts. Several other exchanges have also already used PoR in their audit reports after FXR.
Even more interesting would be to bring cryptocurrency transactions back to their decentralized roots. The original idea behind cryptocurrency was to get rid of the need for third parties, making transactions completely safe from bad actors. Using centralized exchanges like FTX may have some benefits for cryptocurrency users, but it defeats the point of cryptocurrency in the first place. But an open-source algorithm like the one that makes Bitcoin a truly decentralized currency also makes decentralized cryptocurrency exchange (DEX) possible. This means that you don't have to trust your coins to central counterparties with unclear balance sheets, which is a very risky thing to do. Campbell Harvey writes that:
I own asset A and I want to buy asset B. I send my asset A to the decentralized exchange and in return I get asset B. I see the liquidity and I see the price. I can do this 24/7. A DEX has no central counterparty. I interact directly with the algorithm. … I can see the liquidity in real time, essentially an instantaneous audit. There is no leverage. The liquidity is not lent to hedge funds. No broker/dealer is involved… .When I trade, the execution and settlement are simultaneous. Most importantly, I do not delegate custody of my tokens. The result is no waiting in line as an unsecured creditor in a bankruptcy.
There are already a lot of DEX exchanges, and the number of people using them has grown since FTX went out of business.
All of these other ways to stop people from running out of cryptocurrency have bugs that need to be fixed, if not bigger problems. DEX, for example, puts the responsibility of storing and caring for coins back on the individual, though better cryptocurrency wallets can make this job a lot easier. PoR doesn't guarantee that a company won't go bankrupt, and without ring-fencing, holders of balances with "proven" reserves will be treated as unsecured creditors if the company goes bankrupt. But as my colleagues have said elsewhere, good cryptocurrency regulations are worth waiting for, and because information-based runs on cryptocurrency firms don't pose a systemic threat to the U.S. financial system yet, regulators will have plenty of time to get things right. That is, unless they get scared, they'll have plenty of time.
** Information on these pages contains forward-looking statements that involve risks and uncertainties. Markets and instruments profiled on this page are for informational purposes only and should not in any way come across as a recommendation to buy or sell in these assets. You should do your own thorough research before making any investment decisions. All risks, losses and costs associated with investing, including total loss of principal, are your responsibility. The views and opinions expressed in this article are those of the authors and do not necessarily reflect the official policy or position of USA GAG nor its advertisers. The author will not be held responsible for information that is found at the end of links posted on this page.