How a bank run happens at the speed of an app

Depositors tried to pull $42 billion out of Silicon Valley Bank (SVB) in less than ten hours on Thursday, 9 March, more than a quarter of its deposits. The run followed a botched capital raise meant to cover a $1.8 billion loss on available-for-sale securities the bank had sold. By the time the California regulator shut it on Friday morning, SVB was insolvent. It was the first real bank run of the smartphone era, and no bank with a $200 billion balance sheet had ever gone under so quickly.

The run exposed what happens when badly managed duration meets concentrated deposits. A normal retail bank has millions of small, insured accounts. SVB's deposits came overwhelmingly from venture capital funds and the companies they back.

Forty-eight hours

The sequence took two days. On Wednesday evening SVB said it had sold a large block of securities at a $1.8 billion loss after tax and would sell shares to plug the hole. By Thursday its share price had collapsed and the equity raise was dead. On Friday the regulator closed the bank and appointed the FDIC as receiver, making it the largest American bank failure since Washington Mutual in 2008.

The timing of the announcement did as much harm as its content. SVB disclosed the loss and the planned share sale on the same day that Silvergate, a much smaller lender to crypto firms, said it would wind itself down. Investors reading both statements together saw a pattern of banks forced to sell bonds at a loss to meet withdrawals, and the share price fell by about 60 per cent on Thursday. A falling share price made the capital raise harder, and the failed raise told depositors that no fresh equity was coming.

How the balance sheet was built

SVB's troubles began in the boom. Deposits roughly tripled between the end of 2019 and early 2022 as venture funding poured into technology start-ups, which parked the cash at the bank most of them already used. Loan demand from those same firms was modest, so SVB put most of the money into Treasuries and agency mortgage-backed securities, much of it locked into fixed rates close to the bottom of the cycle. When the Fed began raising rates last March, the market value of those bonds fell, and start-ups whose funding had dried up began drawing down their balances.

The squeeze then came from both sides. Outflows forced the bank to find cash, and the only quick source was the bond book whose value had fallen. Selling bonds turned a paper loss into a real one and dented capital, which prompted the share sale, which prompted the run.

Panic spread by group chat

More than ninety per cent of SVB's deposits exceeded the FDIC's $250,000 insurance limit. When well-known venture capitalists spotted the capital shortfall and sent out warnings on WhatsApp groups and Twitter, the panic spread in hours.

There were no queues round the block. Depositors tapped a button on their banking apps or sent wires from corporate treasury portals, and billions left in seconds. Payments moved faster than the bank could pledge collateral or tap the Fed's discount window.

The comparison with earlier runs is sobering. Washington Mutual's depositors withdrew $16.7 billion over ten days before it failed in September 2008, and Northern Rock's customers in Britain queued outside branches the year before. SVB lost far more in a single working day, and a depositor base of closely networked firms behaved like a single account holder.

The nearest precedent is Continental Illinois in 1984. That bank also relied on large, uninsured deposits from institutions rather than households, and when rumours spread about its loan losses those depositors withdrew by wire over a matter of days. The run on Continental was an early sign that wholesale funding could leave faster than branch deposits, and SVB has now shown how far that speed has increased.

Rules written for slower runs

SVB's managers made the oldest mistake in banking. They funded long-dated fixed-rate securities with flighty, unhedged corporate deposits that cost them nothing. When rates rose, the assets lost value and the deposits walked out overnight.

The post-crisis liquidity coverage ratio (LCR) did not allow for depositors moving together on the back of social media. The ratio asks whether a bank holds enough liquid assets to survive thirty days of stressed outflows. A 2018 law raised the threshold for the strictest post-crisis oversight from $50 billion of assets to $250 billion, and the tailoring rules that followed spared banks of SVB's size from the full version of the test. No calibration of a thirty-day horizon anticipated a quarter of deposits leaving in a single afternoon.

The discount window has a similar problem. A bank can borrow from the Fed only against collateral that has been pledged and valued in advance, and moving securities into position takes time and operational work. SVB ended Thursday unable to cover its outflows, according to the California regulator's closing order, even though it owned tens of billions of dollars of government bonds. Liquidity that exists on paper is useless if it cannot be turned into cash before the payments system closes.

Who is left exposed

The FDIC has promised insured depositors access to their money by Monday morning. Uninsured depositors, including start-ups that need to meet payroll next week, have been left to wait for whatever the receivership recovers from the bank's assets. Shareholders and bondholders are likely to lose most or all of their stakes.

The losses reach beyond the bank itself. Start-ups that kept their entire operating cash at SVB face the prospect of missing wages and supplier payments while the receivership works through the assets, and some venture funds had capital calls in progress through the bank. Payroll processors and lenders with exposure to SVB's clients are checking their own positions. The FDIC's usual tool is an advance dividend to uninsured depositors based on expected recoveries, and its size will set how much pain spills into the technology sector next week.

Shares in other regional lenders with large uninsured deposit bases fell hard on Thursday and Friday, and the fear is that their depositors will draw the same conclusion as SVB's. The first gauge will be how much of the uninsured money the FDIC can promise to return, and how quickly. Deposit flows at mid-sized banks over the next week will show whether the panic was contained to one idiosyncratic lender.

One-off, or a new template

The case that SVB was a one-off has some force. Few banks combined such a concentrated client base with so much of their balance sheet in long-dated bonds bought at the bottom of the rate cycle. The difficulty is that the channel of the run, fast digital transfers prompted by messages among connected depositors, is available to every bank. The obvious beneficiaries are the largest banks, whose size makes depositors assume they would be rescued, and they are likely to gain deposits from smaller rivals whatever the regulators decide.

A bank with many uninsured depositors can now be emptied between breakfast and lunch. Regulators will probably have to tighten liquidity rules for mid-sized banks, and other lenders with similar deposit bases will face scrutiny in the coming weeks.