Silicon Valley Bank’s “Twitter Fuelled Bank Run”

SVB’s collapse earlier this month was unlike any other bank run in history.

In the old days, a “bank run” meant crowds quite literally running down to a bank branch in a panic to withdraw their money before the bank went under.

You never wanted to be last in the queue because you probably wouldn’t get your money out.

The speed with which SVB’s collapse unfolded was frightening and most likely enabled by digital banking and social media. Some have even referred to it as the first Twitter fuelled bank run.

According to regulatory filings, SVB clients withdrew $42 billion in a single day prior to the bank being closed the next day. Most withdrawals would have been done online. Compare that to Washington Mutual’s 2008 collapse (still the largest bank failure in US history) where customers only managed to withdraw $16.7 billion in the 10 days before it was shut down.

Watching the collapse happen in real-time on social media and TV was surreal, and it was hard not to experience a slight feeling of nausea. People’s hard-earned money, jobs and the entire tech startup ecosystem were in jeopardy right up until regulators stepped in and guaranteed that depositors would get full access to their funds.

For many SVB clients there might also have been a feeling of sadness at trying to withdraw their funds, like they were being disloyal to a bank that had served them well for years. Commentators would describe the events as being driven by panic and hysteria.

But withdrawing your money in a bank run is anything but panic. At least, according to game theory, it’s the most rational and reasonable thing to do.

Why did the SVB Bank Run Happen?

Nobel prize winners Douglas Diamond and Philip Dybvig have explained why bank runs happen using the Diamond-Dybvig Model.

The thing that banks are really good at – and what makes them valuable in an economy – is that they transform illiquid assets (investments) by offering liabilities with a different, smoother pattern of returns over time than the illiquid assets offer.

You put your money in a bank, they go off and invest that money into illiquid assets and the bank effectively provides you with insurance that your money will be available for you when you need it. Banks can do this because they rely on the fact that few depositors will need to withdraw all of their money at the same time.

If there’s confidence in the banking system, there can be efficient risk sharing among people who need liquidity at different random times.

But when there’s panic, incentives become distorted and every one rushes in to withdraw their deposits before the bank can give out all of its assets.

The bank then has to liquidate its assets and sell them at a loss.

A Game Theory Explanation of the SVB Bank Run

We can try to unpack how this works (at a very basic level) with the following game:

• Let’s imagine SVB only has two depositors: Depositor 1 and Depositor 2. We can assume they’re completely rational humans.
• Rumours surface that SVB has balance sheet issues, its stock price is quickly collapsing and it’s urgently trying to raise funds.
• Our depositors each have $1,000,000 in their accounts. They know their deposits are FDIC-insured up to $250,000 but are concerned about the uninsured amounts.
• Our depositors can only take 1 of 2 actions:
  • Leave the funds in their accounts; or
  • Withdraw their funds exceeding $250,000.
• If both depositors leave their funds, SVB will have enough time to generate liquidity to honour their withdrawals. SVB will be saved from bankruptcy.
• If SVB declares bankruptcy, depositors will lose their funds exceeding the FDIC-insured $250,000.

We can predict the outcome of this game using a little game theory:

In game theory terms, this is a co-ordination game because the best outcome for both players is to co-ordinate strategies and both leave their funds in the bank, allowing it the time to shore up its liquidity.

If they both do so, they will keep their full deposits but SVB might need to put a limit on how much they can withdraw in the short term, say $800,000.

We can represent this in our payoff table by each player receiving 800 in the (Leave Funds, Leave Funds) cell.

But this needs trust in order to work.

Imagine you’re Depositor 1.

You know that if Depositor 2 breaks rank and withdraws their funds, they stand to gain by getting all of their money out if they’re first to act. When that happens, you also know SVB will probably go bankrupt and you’ll be left with only the FDIC-insured $250,000.

This explains the fear of being the last in the queue to get your money out.

In our payoff table, Depositor 1 receives only 250 while Depositor 2 receives 1000 in the (Leave Funds, Withdraw Funds) cell.

So, if you think Depositor 2 will withdraw their funds, what’s your best course of action?

You surely won’t do nothing at the risk of losing most of your money. Your best course of action is to also try to withdraw your funds before the other depositor. This way, there’s at least a chance you get most (if not all) of your money out if you withdraw in time.

Totally rational and reasonable.

But Depositor 2 has the exact same thought, and you both rush to withdraw your funds and the bank goes bankrupt.

Because both depositors rushed to the bank, they each only managed to withdraw a fraction of their cash, say $150,000, before SVB goes bankrupt. Along with the FDIC-insured $250,000, each depositor is left with only $400,000.

They’re each worse off than if they both left their funds, but each still better off than if they did nothing and the other player withdrew their funds.

In our payoff table, this is illustrated in the (Withdraw Funds, Withdraw Funds) cell where both depositors score 400.

Despite acting rationally, both depositors in our game still end up losing out. In fact, it’s their completely rational behaviour that causes the very outcome they were trying to escape.

In the real world, SVB had many more depositors and we can multiply this game by orders of magnitude.

Some depositors caught wind early and got large portions of their money out, but many were stuck with huge amounts frozen in locked accounts by the time the authorities stepped in on 10 March. The kind of co-ordination needed to reach an optimal outcome for all depositors simply wasn’t possible.

Because of the high concentration of SVB’s clients in the tech startup space, we actually saw an opposite kind of co-ordination. In the days before the bank run, many VCs advised their portfolio companies to withdraw their funds. The startup community is highly network driven, and fears of SVB’s collapse spread through closed startup operator groups.

No one wanted to be last in the queue.

From what we know, it seems like one of the main causes of SVB’s downfall was a lack of liquidity. Given time, it may well have been that the best outcome for depositors (and the wider ecosystem) was to coordinate and leave their funds in SVB to allow the bank to find an investor or buyer.

We can only guess.

So, where to from here?

Basic game theory helps us predict that any rational person holding cash in a bank at risk of bankruptcy will rush to withdraw their deposits, given their beliefs about what other depositors will do.

This makes government interventions through deposit insurance and guaranteeing liquidity all the more important to protecting the integrity of the banking system and helping prevent the massive economic loss that history shows us follows in the wake of bank runs.

That’s not a criticism of the free market or a call for interventionism, one way or the other.

But black swan events seem to be happening more regularly, and it’s not a stretch to assume that people will continue to act rationally and in their own interests when they do happen.

Many businesses with SVB deposits would have feared for their survival over the weekend of 10 March. We can only guess how bad things could have been had they not been able to access their cash come Monday.