Wake up to the dangers of digital bank runs

Last month, just after Silicon Valley Bank collapsed, a team of American economists embarked on the daunting task of crunching 5.4mn tweets.

The reason? When SVB failed it was dubbed the world’s first “Twitter-fuelled bank run” by Patrick McHenry, House Financial Services Committee chair. So, the economists — Anthony Cookson, Corbin Fox, Javier Gil-Bazo, Juan Imbet and Christoph Schiller — wanted to test if this was true, or not.

The results are striking: after perusing all tweets mentioning bank stocks since 2020, they conclude that “social media exposure led to bank run risk rather than simply reflecting it”.

More specifically, “at the hourly frequency, we show that negative returns [for banks’ share prices] emerge after periods of intense Twitter conversation,” they added. They also noted that “banks with a large pre-existing exposure to social media performed much worse during the recent SVB bank run,” since the granular pattern of tweets suggests that “investor conversations spilled over into depositor conversations”. 

In other words, and to borrow a distinction made by financial sociologist Donald MacKenzie, social media is not merely a “camera” of events, but an “engine” driving them as well when combined with 24-7 mobile banking. Hence why some $42bn (or almost a quarter of all deposits) left SVB in a few hours on March 9 — and another $100bn was poised to flee on March 10, before the bank was shut down.

So what should regulators and investors do?

Some might think (or hope) that SVB was an anomaly. After all, what made it so vulnerable was not just its mismanagement of interest rate bets, but the concentrated — uninsured — nature of its depositor base. And as the economists’ study shows, many depositors congregated online in a tribal echo chamber, fuelling feedback loops.

But while SVB was an extreme case, it also revealed a bigger pattern: neither banks nor governments are prepared to deal with a social media-infused world that operates at warp speed.

One problem exposed in March, for example, was that the Federal Reserve’s discount window, which is where troubled banks can get liquidity in a crunch, closes at 7pm Eastern time, even though social media and mobile banking operates 24-7. Another was that the staff of the Federal Deposit Insurance Corporation have hitherto assumed that they will have enough warning before a bank collapses to scrutinise its books, prior to a sale. Not so at SVB.

A third problem is that the Basel III regulatory framework only asks banks to hold enough liquidity to protect themselves from a scenario in which they lose 5 per cent of their deposits each day, for 30 days, or a quarter of their deposits; SVB lost that in a few hours.

Then there is a more generic issue: information moves so fast in cyber space that it is hard to detect what is accurate and what is not. And while this leaves some beleaguered bankers eager to muzzle social media, this seems almost impossible to do right now — particularly given that the owner of Twitter, Elon Musk, is a self-described “free speech absolutist.”

So is there anything that governments can actually do? We should hear some answers when the Fed releases its own report into SVB on May 1. But even before this, there are some obvious steps that Washington (and others) should take. First, the Fed urgently needs to ensure that its discount window can operate 24-7.

Second, regulators must improve how they track signals of impending stress, not just in markets but on social media as well. The obvious way to do this in America is to give more resources and status to the Office of Financial Research, a monitoring entity created after the 2008 banking crisis that was subsequently (foolishly) undermined by the administration of Donald Trump.

Third, regulators should change their liquidity coverage rules to prepare for faster bank runs (although that would not necessarily have protected SVB). Fourth, banks should explore the idea of introducing measures to restrict the speed of deposit outflows in a crisis.

It would be a terrible idea to do this ad hoc during a crisis, since this could fuel contagion. But it might not be so crazy to stipulate that, in the future, when new accounts are opened that exceed the $250,000 threshold for FDIC insurance, these could be gated for a few days in a shock.

However, the unpalatable fact is that even if all of these steps are implemented, it might still not be enough to stop the contagion arising from social media panics. So the final move that needs to be made is for governments to recognise that the only truly effective option to quell a cyber panic in a hurry is to backstop the system themselves, by protecting depositors.

The SVB drama suggests that governments will indeed do this — if needed. But that raises two more big questions: if the only way to quell a Twitter panic is for governments to backstop depositors, does that mean banks must become utilities? And, most crucially, will governments also backstop non-banks such as money market funds if they fall prey to a Twitter run?

Right now the answers to these questions seem completely unclear. Maybe Musk should test the market mood in a new Twitter poll.

gillian.tett@ft.com