larry white is a professor of economics at New York University’s Stern School of Business. From 1986 to 1989, he was a member of Federal Home Loan Bank Board, which regulated the savings and loan industry.
Illustrations by daniel hertzberg.
Published July 24, 2023
March, as the early rounds of the NCAA basketball championships were in progress, a very different high-stakes drama was unfolding in northern California. Silicon Valley Bank — the 16th-largest commercial bank in the U.S. with $212 billion of assets and $173 billion in deposits — was circling the drain.
SVB is now history. Meanwhile, First Republic, another mid-sized bank with similar problems, fell in early May, making it the third significant bank failure this year. Here, I offer a primer on what went wrong, what the failures say about the deeper problems of the banking system — and what to do about it.
Back to SVB. After nervous depositors withdrew almost a quarter of their money from the bank, which was closely identified with the venture capital establishment, regulators swiftly closed the bank to staunch the hemorrhage. Two weeks later, the remnants were transferred to First Citizens Bank & Trust of Raleigh, North Carolina, at an estimated $20 billion loss to the FDIC.
This sort of thing isn’t supposed to happen in our tightly regulated banking system — and yet it did. How did a bank with assets larger than Hungary’s GDP, which hadn’t even been on the FDIC’s “problem bank” list three months earlier, end up on the brink? The reasons are simple; a resilient repair, alas, is not.
The Essence of a Bank
Stripped to basics, banks do two important things: they take in deposits, and they make loans. The deposits are a liability of the bank (since the depositors expect to be able to claim their deposits, usually “on demand”). The loans are assets of the bank since it expects to be paid back with interest.
Now to complicate matters, add a third important function: banks are intermediaries at key junctures of the payment system of the whole economy, facilitating the transfer of literally trillions of dollars — those deposits — each day between and among households, businesses and governments.
The balance sheet of the hypothetical ABC Bank conveys the essence (see the figures below). ABC’s assets ($1,000 in outstanding loans) exceed its liabilities ($920 in deposits) by $80. So the net worth of the bank, the owners’ equity, is $80; this is by simple subtraction. (You can add seven or eight zeros to each of the numbers if you want the example to reflect the true scale of a typical American bank.) This difference between assets and liabilities is also called the bank’s “capital.” Since ABC’s capital is positive, the bank is solvent: if all of its loans are repaid, there would be more than enough funds to satisfy the claims of the depositors.
Now, most bank deposits can be withdrawn at short notice at the tap of a smartphone button. But the loans to businesses and individuals (and governments in the form of bonds) that constitute the bank’s assets are usually made for lengthy periods — as much as 30 years for residential mortgages or U.S. Treasury bonds. And a hefty portion can’t be readily called in or sold swiftly in the event that depositors want their money immediately. So banks keep on hand some very liquid assets like short-term government securities that can be turned into cash quickly.
But what if the depositors want more cash immediately than the bank can lay its hands on? Unless the bank can borrow the needed funds, the bank must either suspend its promise to the depositors that they can withdraw their funds at will, which usually forces the closure of the bank. Or the bank must hold a fire sale of its illiquid assets, which (again by simple subtraction) is sure to mean a big loss of its capital.
Now let’s look at the hypothetical XYZ Bank, which has only $800 in assets (loans) and $920 in liabilities (deposits). This bank is insolvent: if depositors demand their money, the last ones in line will be out of luck. The depositors who suspect the worst will get there first. Soon enough, though, less-well-informed depositors will figure out that something’s wrong, and the “run” on the bank will accelerate.
How might the ABC Bank become the XYZ Bank — and how do you tell if it has? First, there is the straightforward problem of credit risk: not every loan is destined to be repaid, and the collateral backing the loan (say, a car or a house, or an office building) may be inadequate to cover it. This has been a potential problem for banks since the beginning of time. Indeed, measuring this risk and allocating loans accordingly is a central function of capital markets.
But there’s another kind of risk faced by banks: interest rate risk. Suppose our bank makes a long-term loan to a super-safe borrower, buying a 10-year U.S. Treasury bond at an annual interest rate of, say, 1.5 percent, and pays $1,000 face value for it. (You’ll see why I chose 1.5 percent in a minute.) The bank uses checking account deposits that pay customers a lot less than 1.5 percent — in many cases, no interest at all — to finance the purchase. This “borrow-short-and-lend-long” pattern is one very important way that banks make a profit. And it’s a nice boring way to make a living — provided interest rates don’t rise sharply and unexpectedly.
But that’s a big proviso. Suppose that a year after the bank has bought the bond, the interest rate for newly issued 10-year bonds rises to 4 percent. That 1.5 percent bond (with nine years still to go until maturity) is now worth only about $800 in the open market. If depositors notice and worry that the bank won’t be able to sell its bonds for enough to fully cover its deposits, the end can come very, very quickly.
Both credit risk and interest rate risk are inherent in modern banking. Well-managed banks keep both kinds of risks at modest levels. Poorly managed banks end up like the XYZ Bank.
So, What Went Wrong at Silicon Valley Bank?
Multiple things went wrong at Silicon Valley Bank, but three stand out.
First, as can be seen in the table, over just two years SVB tripled in size, whether measured by total assets or by deposits. And rapid growth can strain any organization. More important, the composition of its assets changed radically. The bank disproportionately invested in bonds (such as Treasuries and mortgage- backed securities) with distant maturity dates. Whereas SVB’s total bond holdings were less than its total loan holdings at yearend 2019, two years later the former was double the size of the latter.
From a credit-risk perspective, these bonds were bulletproof — SVB wasn’t going to become the XYZ Bank in that way. But the bank was taking on massive interest rate risk by holding a lot of bonds with very long maturities.
Second, around 90 percent of SVB’s deposits were not insured by the FDIC, which only covers deposits up to $250,000. This meant that those depositors, which included Silicon Valley tech firms as well as investors in those firms and other wealthy households, would be at risk in the event SVB became insolvent. And nowadays, worried depositors could withdraw their funds at the blink of an electron.
Third, because the standard accounting framework used by U.S. corporations (including banks) has a backward-looking “historical cost” orientation, SVB could continue to claim that its assets were worth their purchase prices — not what they could be sold for to meet withdrawal requests. So as interest rates rose starting in early 2022, Silicon Valley Bank was rapidly morphing into the XYZ Bank. Through most of 2021, anyone who bought newly issued 10-year Treasury bonds received an average interest rate of around 1.5 percent. But by year-end 2022, newly issued 10-year Treasuries carried an interest rate of around 4 percent.
Despite these interest rate increases, standard (perfectly legal) accounting allowed SVB at year-end 2022 to state that it had $211.8 billion in assets, $173.1 billion in deposit liabilities and $22.4 billion in other liabilities, which (by subtraction) meant that it had $16.3 billion in net worth (equity), or a ratio of equity to assets of 7.7 percent.
However, anyone curious enough to inspect page 95 (of 193 pages!) of SVB’s annual report for 2022 (issued on February 24, 2023) would have discovered that, of SVB’s year-end $120 billion in investment securities, $91 billion was in a category labeled “held-to-maturity” — which meant that SVB didn’t intend to sell these securities before they matured. And our financial sleuth would have also discovered that the “fair value” (which is accounting jargon for the current market value) of those held-to-maturity securities was only $76 billion. And if the sleuth had earlier (in October) looked at the bank’s report for the third quarter of 2022, she would have seen roughly similar information.
Thus, hiding in plain sight for months was SVB’s admission that its assets, if sold immediately, were worth $15 billion less than the figure on its balance sheet. So, SVB was in fact on the edge of insolvency — and had been since at least the end of September 2022.
True believers in markets like to think that more often than not, they are self-policing. However, very few of those sophisticated SVB customers with uninsured balances seem to have noticed. Even bank analysts — the nerds who study this stuff for a living — were sanguine. Although SVB did lose some deposits between June 30, 2022, and year-end, there was certainly no rush to the exits. The run on SVB occurred only after SVB’s announcement on March 8, 2023, that the bank had sold $21 billion in securities at a loss of $1.8 billion. Then, and only then, did exposed depositors wake up.
Where Were the Regulators?
For good and obvious reasons, banks are heavily regulated. And the essence of prudential regulation is the requirements that banks maintain adequate capital (so that they can absorb losses) and that they keep sufficient liquid assets on hand (so as to accommodate the ebb and flow of depositor withdrawals).
SVB’s primary safety-and-soundness regulator was the Fed. This meant that SVB’s books were periodically examined for capital and liquidity adequacy, so that the regulators could intervene in a timely fashion when necessary. The very largest banks (SVB didn’t qualify) are also subjected to “stress test” scenarios to get a sense of whether they would survive, say, a rerun of the 2008 global financial crisis.
Now, from mid-2021 onward, Fed regulators did express concerns to SVB’s management about SVB’s investments and uninsured deposits. But they didn’t follow-up with demands that SVB raise more capital (say, by selling more stock) or otherwise address the substantial risks — including liquidity risks — that the regulators clearly knew about.
It’s true that, thanks to 2018 federal legislation easing regulation on mid-sized banks, SVB was scrutinized less intensively than would otherwise have been the case. But SVB’s exemption from stress testing did not make a significant difference because — bizarrely — these stress tests didn’t include scenarios that envisioned the sharp increase in interest rates that proved fatal to the bank.
What Happened Next?
Immediately after SVB’s closure, the FDIC (together with the New York State regulators) closed another bank — Signature Bank, headquartered in New York City. Signature was half the size of SVB, but still sizable: $110 billion in assets and $89 billion in deposits. Critically, like SVB, around 90 percent of Signature’s deposits were uninsured, and thus these depositors were also easily spooked. A week later, Flagstar Bank of Hicksville, New York, took over what remained of Signature in return for $2.5 billion from the FDIC.
That’s not the end of the story, though. Federal regulators quickly became concerned about contagion — that uninsured depositors at other banks would rush to withdraw their money as well. These concerns were surely heightened by loud wails from Silicon Valley movers and shakers who predicted Armageddon if the uninsured depositors at SVB didn’t get back every penny. For whatever reason, two days after SVB’s closure the regulators decided that all deposits at the two failed banks would be retroactively insured.
Further, hoping to head off a chain reaction of runs (and demands for more ad hoc deposit guarantees), the Fed announced that other banks could obtain loans directly from the Fed by offering securities as collateral. What was exceptional here was that banks could borrow up to the face value of these securities, even if the current market value was lower. The Fed, in effect, was guaranteeing deposits at banks that were sitting on big securities losses in order to smooth over the moment of maximum risk of contagion.
Judging by what didn’t happen in the following week, the strategy worked — the runs never materialized. But bank analysts did start paying attention to those embedded (that is, unrecognized) losses in other banks’ securities holdings. And they began figuring out which banks had high percentages of depositors who were uninsured and therefore more likely to run if they became nervous.
Much more attention was given to the FDIC’s estimate that, all told, the banking system was carrying some $620 billion in unrecognized securities losses (see exhibit), which was over a quarter of the banking system’s aggregate equity capital of $2.2 trillion.
Less than two months later, the chickens came home to roost for First Republic Bank of San Francisco, which was vulnerable on both counts. The FDIC arranged a shotgun marriage with JPMorgan Chase, which is too big to fail — but that’s another story.
As of year-end 2022, there were at least 30 banks, each with assets exceeding $50 billion, where at least 45 percent of deposits were uninsured. And many of them surely have sizable unrecognized securities losses lurking in their balance sheets — that aforementioned $620 billion must be somewhere. This suggests that mid-sized banks aren’t out of the woods yet, especially if interest rates stay at their current high levels as a consequence of the Fed’s efforts to tame inflation. The only good news is that Washington’s decisive moves to tamp down contagion after the SVB debacle may deter uninsured depositors from jumping ship at the first sign of problems.
There is, however, a wild card that wasn’t a problem for those three large banks but that may add to the woes of the banking sector going forward: the uncertain state of loans for building and operating commercial real estate (CRE), which consists of office buildings, shopping centers, hotels, warehouses and the like. At year-end 2022, the banking system in aggregate was carrying $1.8 trillion in CRE loans on its books, while the system’s capital was only modestly larger, at $2.2 trillion.
It is also worth noting that CRE lending is relatively more important for smaller banks. They have $570 billion of CRE loans on their books, accounting for one-fifth of the assets of the 4,258 banks that the FDIC considers to be “community banks.” To put those numbers in context, these CRE loans were more than twice the $252 billion capital of these community banks.
Mutual funds and investment banks mark-to-market daily, which seems a worthy goal for banks, too. This would be a challenge, of course, for real estate loans or garden-variety business loans, where there is no ready market to use as a yardstick, and the deposit franchise of a bank has a value (albeit sometimes sketchy).
The big question is what portion of the loans (and the banks) are in trouble. Certainly the occupants of many commercial structures — notably, brick-and-mortar retailers — are under a cloud. And Covid-19, plus everimproving digital technologies and telecommunications that make remote work more practical, has now brought office occupancy under a similar cloud. Thus, some of that $1.8 trillion in commercial real estate loans will sooner or later need to be written down. Real estate developers and bankers clearly favor later. But in a banking system with a heightened propensity to contagion, “later” may arrive sooner than they expect.
There is a further worrisome aspect of banks’ CRE lending. The banking industry — especially small- and medium-size banks — have a dismaying record of lending on favorable terms to real estate developers and landlords, and then ruing the consequences. Excessive real estate lending contributed to the demise of over 1,000 commercial banks in the 1980s and early 1990s, as well as to the disappearance of many hundreds of savings and loans. And real estate lending contributed to the demise of a few hundred smalland medium-size banks in the years of the Great Recession.
Was Yogi Berra right? Will it be “déjà vu all over again”?
What Is to be Done?
For the near-term, the unrecognized losses, uninsured deposits and poor-quality commercial real estate loans are largely baked in; government policy will have to deal with the consequences on an ad hoc basis. But it is not too soon to consider reforms that can ameliorate the problems that otherwise will likely tip the banking system into yet another crisis a few years hence.
Perhaps the most important reform — and the one that has received the least attention — would be a mandate for market-value accounting. It is a travesty that SVB was legally able to portray itself as well-capitalized until two days before its demise.
Regulators need market-based valuations as the rock on which to build a safer system. And if regulators have this information, the investing public ought to have it as well. Moreover, in an electronic age, these data ought to be available much more frequently than once a quarter.
Mutual funds and investment banks markto- market daily, which seems a worthy goal for banks, too. This would be a challenge, of course, for real estate loans or garden-variety business loans, where there is no ready market to use as a yardstick, and the deposit franchise of a bank has a value (albeit sometimes sketchy). But these challenges are no excuse for delaying up-to-date pricing for traded securities like bonds, which would go a long way toward reducing the unrecognized losses problem.
Bankers will scream that market valuation would introduce needless volatility into their earnings and balance sheet reports. But hiding volatility is hardly a cure for volatility. And the advantages are plain.
No longer would regulators be left to analyze bank stability with information that is at best out of date, at worst grossly misleading. The Fed’s emergency lending program that began in March, which allows banks to borrow against the face value of their assets rather than their market value, was a step in the wrong direction. There were surely ways of providing emergency liquidity to banks that do not involve the Fed’s implicit endorsement of the banks’ Tinkerbell accounting.
Addressing Interest Rate Risk
Capital requirements need to be explicitly linked to the amount of interest rate risk that banks choose to take. It is truly astonishing that the matter is left to the discretion of regulators — especially since it has been over 40 years since these risks were widely acknowledged to be a central cause of the collapse of the savings and loan sector.
Higher, Better-Measured Capital Requirements
Capital adequacy is the cornerstone of a resilient banking system. Capital, of course, serves as a buffer against losses, but it is also a deterrent to excessive risk-taking since holding more capital means that bank owners have more to lose. Capital protects the FDIC directly since more capital reduces the likelihood that the insurance fund will be tapped, as well as indirectly because capital serves as a barrier to contagion.
But unfortunately, the capital shield is growing thinner, not thicker. While banks comply with minimum capital requirements, they have allowed average capital-to-asset ratios to decline from 11.3 percent in 2019 to 9.3 percent in 2022. Note, too, that the failure to use market value accounting only exacerbates the risk.
Rethinking Uninsured Deposits
Uninsured deposits, which currently represent about 40 percent of all deposits (up from only 20 percent as recently as the mid-1990s), are plainly a source of instability in the banking system. Their owners may be slow to recognize the financial problems of their banks. But when they do, their reactions are fast, massive — and contagious.
The immediate need, then, is to make sure that banks’ liquidity is deep enough to manage the problem. Since cash-like liquid assets such as short-term government bonds generate less income for a bank than less-liquid loans, this sacrifice in earnings ought to be reflected in a lower interest rate that is offered to uninsured depositors who always have one foot out the door.
Over the longer run, one way to address the problem is simply to eliminate this source of instability — to extend insurance to all deposits. That would require a change in banking laws as well as a determination on the part of the FDIC to do a much better job of linking the size of the deposit insurance premiums that it charges banks to the level of risk that the banks take on.
The primary argument against 100 percent deposit insurance is “moral hazard” — that large depositors will become indifferent to how their banks manage risk. But as the SVB experience vividly illustrates, large depositors currently don’t seem to notice until their banks are in deep trouble. Depositors — whether insured or uninsured — are simply not good monitors of bank behavior, and thus are not effective restraints.
One promising approach (whether or not 100 percent deposit insurance is adopted) would be to require banks to issue subordinated (and, of course, uninsured) debt. The bondholders, who couldn’t run because they would be holding longer-term debt rather than withdrawal-on-demand deposits, would presumably be more sophisticated and knowledgeable about banks and banking than are uninsured depositors. Ideally, the bondholders would be given a say in bank governance as part of the deal, so they would have the ability (as well as the incentive) to restrain bank excesses.
This idea of mandatory subordinated debt (or some rough equivalent) was widely dis cussed after the financial crisis of 2007 to 2009. But the discussion petered out — and at year-end 2022, the U.S. banking system had only $65 billion in subordinated debt out standing, which is less than 0.3 percent of total assets. Surely, it’s time to revisit that idea.
Another argument that is offered in opposition to 100 percent deposit insurance is that the beneficiaries will be high-income households who won’t pay their share of risk-based deposit insurance premiums. But this argument ignores the spillover costs for the banking system that any substantial amount of uninsured deposits create.
If 100 percent deposit insurance is too much to swallow politically, other measures will be needed to reduce the flight risk of uninsured deposits. The guiding idea should be that the act of withdrawing a large deposit from a bank has the potential to harm all stakeholders — including depositors at other banks. This makes it a standard “negative externality,” much like air pollution or traffic congestion. The problem needs to be addressed through measures that would slow down withdrawals of uninsured deposits and/ or internalize the cost that withdrawal-ondemand poses to third parties.
This is no place for a deep dive, though. Suffice it to say that a number of means to this end are worth serious consideration.
And in the End …
Like plumbing, when banking works well, nobody much notices. When banking has problems, however, everybody has problems. Unfortunately, unlike plumbing, banking is arcane — even savvy reporters struggle with the mechanics.
The SVB debacle is only the latest instance in which the machinery that makes the metaphoric sausage is exposed. Even if the shortrun fixes can keep the machinery going, the long-run problems, alluded to above, are ignored at our peril.