Anatomy of a bank collapse
Things move slowly, then suddenly all at once.
“Every banker knows that if he has to prove that he is worthy of credit, however good may be his arguments, in fact his credit is gone.” — Walter Bagehot
It’s hard to pin down the one date that the average American might have realised they were entering the Great Depression. The signs of a weakened economy were quickly apparent in the months following the market crash of October 1929 — industrial regions were hit the hardest as automobile sales slumped, and the Smoot-Hawley act began a new age of protectionism in the US. But it was only in September of the following year that the first commercial banks started failing — at first slowly, and then suddenly all at once.
By that November came the first failure to hit national headlines, as banking and insurance giant Caldwell & Co. — the largest financial holding company in the South — collapsed spectacularly in the space of just a few weeks. First to go was their Bank of Tennessee, followed days later by affiliates in Knoxville and Kentucky. As their towns’ banks vanished overnight, communities rushed to withdraw their funds from other local banks. Word spread to nearby towns who panicked in turn, and so on, and so on. Within a few weeks, hundreds of banks had suspended operations — most never reopened.
But how is it that a bank can just run out of money to return to its depositors?
To answer this, you’ll often hear about “fractional-reserve banking”, money multipliers, and all the rest. But that goes a level too far, and skips over the key point of why your bank doesn’t want to hold all your money — ready for you to take it back. (If you know about this stuff, feel free to skip ahead!)
A bank is a business which borrows money on a short-term timeline, and lends money on a longer timeline. When you place your money with a bank, you’re lending the bank money, usually against a very low interest rate for checking accounts. It’s borrowing your money, and over no set period — it has to return it to you as soon as you want it back. With the money you provide them, banks, in turn, lend that money elsewhere. That lending takes two main forms: creating new loans for customers (mortgages, corporate lines of credit, etc.); and buying existing debt (mortgages originated by other banks, US treasury bonds, etc.). These investments pay a high return, but are difficult to convert to cash quickly.
This difference between the low cost of borrowing money (the low interest rate paid on deposits), and the high return of holding long-term assets is called the net interest spread, and is the primary way banks make money. There’s an inherent risk in this business model, though. We’ve seen that the short-term deposits (the bank’s liabilities) have to be available to withdraw at any moment, but the long-term holdings (the banks’s assets), by their very nature, tie up capital for a long time. These holdings can usually be sold off and turned into cash before their term is due, but almost never for their full value. There’s a baked in tension, then, between savers and borrowers, as depositors need quick access to funds, while the bank realises the profits of their assets much more slowly. This tension is known as a maturity mismatch.
Banks have a lot of experience in modelling roughly how many customers will call on their deposits on an average day. They then base their investment decisions around that estimate: keep enough cash on hand to return money to those customers when they want it, and invest the rest. But that tells us that even if the bank took in $1 of deposits for every $1 it lent out (it doesn’t, but just imagine), it couldn’t give all the money back at a moment’s notice, as it’s tied up elsewhere. The value of all those investments might be worth more than the amount the bank owes depositors (i.e. the bank is solvent), but that value will only be unlocked in the future: the money needed to pay everyone out at once is just not there (i.e. the bank is not liquid).
Sure enough, that model of the normal withdrawal rate can’t be perfect. And on occasion people do ask for their money back faster than planned. This puts the bank in a tricky spot. The math checks out — the bank can pay everyone back — but not if they all come at once.
Bank withdrawals work on the basis of sequential service, meaning the bank will honour withdrawal requests on a first-come-first-served basis. This creates another tension: if I have reason to believe other people will be asking for their money back, I’m now tempted to go and ask for mine back too — I don’t want to be last in the queue and risk the bank running out of money before it’s my turn. Essentially, if I think other people are going to panic, I’ll panic. I join the queue outside my local branch (or open my bank’s app) and ask to move all my money out of the bank. One by one, the queue slowly advances, withdrawal requests are honoured, and far away in a financial district skyscraper, the alarm bells are growing loud.
Apropos of nothing, Silicon Valley Bank’s had a rough week.
On Friday, Silicon Valley Bank, known as SVB, collapsed. It is the second-largest bank in the US by assets to fail after Washington Mutual. The collapse followed a run on deposits that doomed the tech focused lenders plans to raise fresh capital.
A story in three parts
Broadly, a bank run plays out in three acts. First, depositors get an idea that the deposits in their bank might be at risk. At this point, the reasons can be valid or entirely made up. Second, those customers head to their bank to move their funds away from the bank. The sight of this, added to the (true or invented) initial causes of worry, opens the third act: as the flow of worried customers turns from a stream to a cascade. Nobody wants to be the last mover (sequential service!), so they join the run — sending the bank into a liquidity spiral.
Each bank run has its own reasons and its own unique twists, but SVB’s 48h implosion followed the textbook definition almost perfectly.
By early on Wednesday 8th of March, there was reported concern by some financial officers of tech companies regarding Silicon Valley Bank’s asset portfolio. We’ll get into the details in a minute, but their assets didn’t suddenly lose their value overnight, their holdings had been under growing pressure for months. What sparked the worries was the bank’s announcement of a sale of about $21B of their securities portfolio. This forced the bank to realise a $1.8B loss on the assets, which it looked to soften with an approx. $1.75B capital raise from public and private markets.
At this point, the numbers showed SVB having to take some rough blows, but no obvious ticking time bomb.
By Thursday, Bloomberg reported that Peter Thiel’s VC firm, Founders Fund, had advised its portfolio companies to move their money out of SVB. Ever powerful, the Thiel mystique held true, and this news opened the floodgates for thousands of other founders and financiers to rush to take their money out of SVB. Matt Levine, as always, put it best:
Also, I am sorry to be rude, but there is another reason that it is maybe not great to be the Bank of Startups, which is that nobody on Earth is more of a herd animal than Silicon Valley venture capitalists. […] If all of your depositors are startups with the same handful of venture capitalists on their boards, and all those venture capitalists are competing with each other to Add Value and Be Influencers and Do The Current Thing by calling all their portfolio companies to say “hey, did you hear, everyone’s taking money out of Silicon Valley Bank, you should too,” then all of your depositors will take their money out at the same time.
By the morning of Friday 10th, the bank was bust. In the frenzy, customers had initiated withdrawals of over $42B — a quarter of all deposits — and was unable to meet requests. The Federal Deposit Insurance Corporation had taken over the bank — now sitting on an approx. $1B negative cash balance.
A brewing problem
Given how quickly things turned to disaster, the first question you might ask is “what caused the bank to fire-sale that $21B of securities and raise outside capital?”. As mentioned above, bank’s situation didn’t turn dire overnight.
Silicon Valley Bank is the go-to bank in the Bay Area for startups, and in the past few years, as the tech market boomed and investment rounds grew bigger than ever, deposits raced into SVB.
Reaching $189B in deposits in 2021, almost double the prior year’s, the bank needed somewhere to stow away all the excess liquidity. With interest rates at their lowest level in decades, high-yield, low risk investments were scarce. It picked two main targets: long-term government-backed securities, and fixed-rate mortgage bonds. These offered slightly higher yields than short term T-Bills, but meant tying up that capital for a decade-plus. The problem with highly-rated, long-term, fixed-rate assets, though, is that their value is almost entirely dictated by what happens with domestic interest rates. Sure enough, as rates rose sharply in 2022, the fixed-rate bonds’ value declined, and the bank’s holdings went underwater.
For some quick context: when a bank acquires fixed-term securities like these, it has to decide up-front whether they plan to hold them to maturity (HTM), or whether they want to keep the flexibility of being able to sell them before their term is up, called “Available for Sale” (AFS). HTM assets are not “marked to market”, meaning banks can see their bond values plummet and not worry, as it intends to hold them to maturity, collecting the planned amount. AFS assets, on the other hand, are marked to market: their value on the bank’s balance sheet is dependent on how much those assets could be sold for in the short-term.
Only around 20% of SVB’s assets were marked AFS. The remainder was all HTM — meaning that it could (in theory) stomach progressive interest rate rises, because the bonds’ value was certain, regardless of the market. But as rates rose more sharply than planned throughout 2022, mortgage-related assets were hit hard, and the bank’s HTM portfolio suffered major unrealised losses. Quoting Marc Rubinstein at Net Interest:
So big was this drawdown that on a marked-to-market basis, Silicon Valley Bank was technically insolvent at the end of September. Its $15.9 billion of HTM mark-to-market losses completely subsumed the $11.8 billion of tangible common equity that supported the bank’s balance sheet.
Not great, sure, but most of these losses didn’t need to be realised, as most of the portfolio was HTM. It’d just hold until maturity, weathering the storm. But what the bank didn’t anticipate was just how quickly withdrawals would increase as the higher rates sent a shock through the tech sector and companies drew down on their deposits. In the space of just a year, deposits fell almost 20% from $198B in March 2022 to $165B in Feb. 2023.
It’s at this point that the bank started its $21B assets sale and outside money raise, spooking influential depositors, starting the mess of the following days.
In retrospect it’s glaring how SVB always sat at an exceptionally risky intersection. It not only filled its portfolio with interest-rate-sensitive assets, but then banked only with clients whose businesses are almost entirely dependent on low interest rates. When rates rose, SVB was in the perfect position to feel both the first and second-order effects.
And SVB’s position as the go-to bank for highly-capitalised businesses brought with it an extra consideration when things turned south: most customers had account balances far in excess of what federal deposit insurance would cover.
My gains, our losses
Born in 1933, in the midst of the Great Depression-era bank failures, the Federal Deposit Insurance Corporation (FDIC) does about what you’d expect — insuring customer deposits in US commercial and savings banks. Before 1933, bank failures because of runs were common, and most clients never got their money back.
The creation of the FDIC, on its face, was a way of reassuring Americans that in the event that their bank saw a run and collapsed, they’d be made whole (up to a specified amount) by the government. But the real effect of the FDIC went deeper than that: it removed much of the reason for a bank run to start in the first place. Say a rumour goes around that the local bank might be struggling, but you know your money will be protected regardless; why even rush to the bank?
And if we’re measuring bank failures alone, it appears to have worked almost overnight. In the year the FDIC was founded, there were around 4,000 bank failures in the US. The following year, there were 9.
The coverage limit for deposit insurance has grown in real terms by almost 5x since its inception, from $2,500 in 1933 dollars, to $250,000 today, enshrined by the Dodd-Frank act. For most commercial banks, this limit (per account, per person, and per bank) covers most depositors’ needs; but most bank depositors don’t need to cash multi-million dollar VC cheques.
Sure enough, as of end 2022, among the 100 largest financial institutions in the US, only one bank had a lower percentage of customer accounts with a balance under $250k than SVB. Assuming an approximately similar ratio just a few months later, only around 3% of depositor accounts at SVB would have been fully covered by FDIC insurance. In fact, across the approx 37,000 accounts in excess of $250,000, the average balance was north of $4,2M.
A valid question at this point might be, like “why are startups all using this one bank?”. There’s no one single answer here. Mostly, startups in a constant hunt for funding will mostly just do what VCs recommend, and VCs mostly do what other VCs do — so why rock the boat? Also, by some founders’ reports, for a time, SVB was also the only bank that would issue credit cards to higher-risk clients like startup founders. SVB was a major player in the startup funding ecosystem in the Bay Area, too, offering very startup-focused products like AWS credits, venture debt, and wealth management services all rolled into one — a potentially appealing combination for fast-growing companies. It’s unfortunate that SVB only banked startups, and startups only banked with SVB, but it’s undeniably a very Bay Area phenomenon.
Onwards
Over the weekend following the bank’s collapse, we were treated to three days of all-caps fear mongering by prominent VCs, Michael Burry doing his usual thing, and arguments throughout tech Twitter over what the government reaction should be with regard to SVB’s depositors; the core of the debate revolving around whether the bank’s collapse risked sparking wider contagion.
But then came Monday morning:
[The US Treasury] approved actions enabling the FDIC to complete its resolution of Silicon Valley Bank […] in a manner that fully protects all depositors. Depositors will have access to all of their money starting Monday, March 13.
Whew!
Of course I and most everyone is happy that depositors will be made whole, even beyond the $250k normal limit. Overall, I think it’s a harsh call to say that founders should have eaten the loss just because “that’s the risk you take!” — like, sure, I lack a bit of sympathy if your money goes up in smoke because it was tied up in a crypto-focused bank literally 2 days before this all started, but SVB wasn’t some weird fringe bank. Starting its final week, SVB was the 16th largest bank in the US, around since the early ‘80s, and was a household name for any tech-adjacent worker.
Moreover, though much of the story focuses on SVB being a bank for startup founders, most of its victims would have been employees. Major tech companies relied on SVB for operations — namely for paying salaries, which caught attention when payroll company Rippling announced that they couldn’t make their client companies’ payroll, because their funds were with SVB. The collateral damage of a non-backstopped collapse would have had major consequences on the tech ecosystem.
On the other hand, there is the argument of, like, what does that $250k FDIC limit really mean anymore? What the Treasury announced isn’t a bailout to be sure, shareholders and debtors are going to have to eat this one on their own, but the message this sends is a somewhat risky one. Sure, the bank’s execs aren’t getting “rewarded” for their risk-taking like in the GFC interventions — they’re still losing their bank; but the precedent is set: you can probably let your risk appetite grow a little bit, because even if things go bad, the result is mostly just a few days of worry for depositors rather than sector-wide contagion. And when the trial starts, you can go ahead and say “we were maximising shareholder returns, as is our mandate, while balancing our risk given reasonable expectations — we were in the middle of shoring up the bank’s capital position, it’s not our fault Peter Thiel was feeling chatty that day.” And that would be sort of true! It would also be missing a fair bit of the story, though.
Like, for example, SVB not having a chief risk officer for most of 2022. During the period of fastest rate hikes since the late 1970s, the bank was missing a key figure in its risk management. Mike at NonGaap puts it well:
[…] I read the 2023 Proxy and see that the Risk Committee 1) met 18 times in 2022 vs. their typical 5-7 times per year, 2) appointed a new committee Chair, and 3) didn’t have a Chief Risk Officer in place for much of the year, that’s a pretty strong set of signals the company might have been aware they had a serious problem relative to what they were communicating
We’ll have to wait for the investigations to begin to find out more here, but it does seem a bit of an oversight.
Another part of the story might also be the light this collapse shines (again) on US financial regulation. Dan Davies wrote a great piece about this “very American mess”, highlighting the uniquely flexible application of the Basel III regulations, designed to help prevent this exact type of situation. The regulations enshrine a handful of key ratios around liquidity and capital risk by which banks have to abide. They mostly target large, international banks, but regulators in the EU and UK ended up applying them more broadly, to basically all banks. In the very American fashion, the community bank lobby managed to stave off much of this regulation, and SVB benefited from the flexibility. But as Davies summarises: “the fact that a risk isn’t covered by a regulatory ratio doesn’t mean it doesn’t exist.”
And we’ll wrap up on a final note which might end up being the most egregious factor which led to the confusion and panic of SVB’s last days — communications. It looks like a key mistake the bank made in handling the growing problem was announcing the sudden capital raise and assets selloff without having a clear, rational story with which to frame it. The press release picked up by the first news outlets told only the facts: “SVB is raising all this money, right now”; it had no narrative to give the rationale behind the raise. The associated 8-K was solid, illustrating the problem and proposed solution well, but depositors aren’t checking EDGAR, they’re getting their news from tech publications and VC circles, eager to share headlines as they come. Essentially, the bank made all the right moves to safeguard the stock price, but neglected to tell the story to the bank’s clients. Perhaps the bank was wagering that clients would be more worried by the sound of “nothing to worry about — just making sure we have enough money” than by a nondescript, fairly standard financial procedure.
If that was the case, at least we can strike that one from the playbook for next time.
Before last week, it had been 868 days since a bank had last failed in the US — close to the longest stretch on record. As we reset the counter to zero, what can we learn? Well, mostly, if you’re the 16th biggest bank in the country, you should probably have a chief risk officer! Because as it turns out, even small banks can prove too big to fail.