Ninety-Four Percent

A bank run is not irrational. That is the thing to understand before anything else about it.

A bank takes short term deposits and makes long term loans. That is the whole business, and it means that at any moment the bank does not have your money, because your money is a mortgage on somebody's house for the next twenty years. It works because depositors do not all want their money on the same day.

If you believe that other depositors are about to ask for theirs, the correct thing to do is to ask for yours first, because there is a queue and the queue runs out. Everyone reasoning that way at once destroys a bank that might have been perfectly solvent the day before. Each person is behaving sensibly and the outcome is a disaster.

For a century, this happened over and over.

What stopped it

More than nine thousand American banks ceased operations between the stock market crash of October 1929 and the banking holiday declared in March 1933.

Roosevelt signed the Banking Act on 16 June 1933, and among other things it created the Federal Deposit Insurance Corporation. From 1934 an ordinary depositor's money was guaranteed by the federal government up to 2,500 dollars, doubled to 5,000 in July of the same year.

In 1934, nine insured banks failed.

Nine thousand, then nine. Not because the insurance paid out on a huge scale, but because it removed the reason to queue. If your money is guaranteed, the news that other people are withdrawing theirs is somebody else's problem. The insurance works by making itself unnecessary.

It is one of the most effective pieces of financial regulation ever written, and its effectiveness is invisible, because what it produces is an absence.

The limit

Deposit insurance has a ceiling. It was 2,500 dollars in 1934 and it is 250,000 dollars now.

The ceiling exists for a reason that is easy to defend. Insurance without a limit means a bank can offer any interest rate it likes on any amount, take enormous risks with the proceeds, and have the taxpayer stand behind it. The limit is there so that large depositors have a reason to care whether their bank is well run, and so that the discipline of watching the bank falls on people equipped to do it.

The theory is that small depositors get protection and large ones get an incentive. It is a reasonable design.

It also creates a category of institution that is outside the protection entirely.

Silicon Valley Bank

Silicon Valley Bank did what its name says. It banked technology companies, mostly startups, mostly funded by venture capital, mostly in one place and one industry.

A startup that has just raised twenty million dollars puts it in a bank. It is not going to sit in nineteen accounts of 250,000 dollars each. It is a corporate treasury balance and it is uninsured by definition.

By the end of 2022, about 94 per cent of the bank's deposits were above the insured limit.

That single number is the whole story. It means that essentially every depositor at that bank had a live financial interest in getting out ahead of the others, and no reason at all to sit still, because none of the reassurance built in 1933 applied to any of them.

It also means they were unusually alike. They were in the same industry, in the same city, funded by the same few dozen venture firms, on the same group chats. The thing that stops a run is a mass of depositors who do not talk to each other and are protected anyway. SVB's depositors were protected by nothing and spoke to each other constantly.

The day

The other half of the failure was ordinary interest rate risk. The bank had taken a very large deposit inflow during the boom and put a lot of it into long dated bonds. When rates rose sharply those bonds lost market value, so the assets that were supposed to cover the deposits were worth less than they had cost. That is a mistake, it was identified as a mistake afterwards by the Federal Reserve's own review, and the bank's management and board were blamed for it.

But plenty of banks were sitting on unrealised losses in 2023. What made this one fatal was who the depositors were.

On 9 March 2023, customers attempted to withdraw 42 billion dollars in a single day, roughly a quarter of all the deposits in the bank. The cash position went to about a billion dollars negative. The share price fell by 60 per cent. The following day the regulators closed it.

Compare that with 1931. A run in 1931 was a physical queue, and the queue was limited by how fast a teller could count. In 2023 the run was an instruction typed into an app by a founder who had just been told to move by an investor who had told forty other founders the same thing. The mechanism did not change. The clock speed did.

What they did about it

The FDIC and the Treasury invoked the systemic risk exception and guaranteed all deposits at the failed banks, insured or not.

Which is the part worth sitting with. The limit exists so that large depositors have a reason to monitor their bank. When a large enough failure happened, the limit was suspended, and every uninsured depositor was made whole anyway.

You can defend that decision on the day. Thousands of companies were about to miss payroll, and the second and third banks in the queue would have gone the same way for the same reason. Refusing would have been a choice to have a much larger crisis in order to make a point about incentives.

You can also see what it teaches. If the limit binds in ordinary times and is lifted whenever it would actually hurt, then the discipline it was supposed to create does not exist, and what exists instead is an insurance scheme with a published ceiling and an unpublished one.

The thing that has not been solved

The 1933 settlement assumed a depositor who is small, numerous, slow and unconnected. That depositor still exists and is still protected.

The depositor who broke Silicon Valley Bank is none of those things: a company treasury, holding a balance a hundred times the ceiling, in an industry where everyone knows everyone, able to move the entire balance in under a minute at three in the morning.

Nine thousand banks, then nine, then ninety four per cent.

Sources

• Federal Reserve History, "Banking Act of 1933 (Glass-Steagall)" (the signing of the Act on 16 June 1933 and its creation of the Federal Deposit Insurance Corporation, and the banking conditions that preceded it).
• Federal Deposit Insurance Corporation, "A brief history of deposit insurance" (the introduction of federal deposit insurance in 1934 at 2,500 dollars per depositor, its increase to 5,000 dollars in July of that year, and the effect on the rate of bank failures).
• Federal Deposit Insurance Corporation, "The First Fifty Years: a history of the FDIC 1933-1983" (more than 9,000 banks ceasing operations between the stock market crash of October 1929 and the banking holiday of March 1933, and nine insured banks failing during 1934).
• Office of Inspector General, Board of Governors of the Federal Reserve System, "Material Loss Review of Silicon Valley Bank", September 2023 (the concentration of uninsured deposits at approximately 94 per cent of total deposits as of year end 2022; the volatility of uninsured deposits in times of stress and the resulting flight risk; the mismanagement of interest rate risk on the securities portfolio; and the attribution of the failure to senior management and the board).
• Federal Deposit Insurance Corporation, "FDIC acts to protect all depositors of the former Silicon Valley Bank" (the closure of the bank, the invocation of the systemic risk exception with the approval of the Treasury Secretary and the Federal Reserve, and the decision to make all depositors whole regardless of the insurance limit).
• PBS NewsHour, "What to know about the Silicon Valley Bank collapse" (the withdrawal attempt of 42 billion dollars on 9 March 2023, amounting to roughly a quarter of the bank's deposits, and the share of deposits at both failed banks exceeding the 250,000 dollar insurance limit).