Professional Documents
Culture Documents
Metrick 2024 The Failure of Silicon Valley Bank and The Panic of 2023
Metrick 2024 The Failure of Silicon Valley Bank and The Panic of 2023
Andrew Metrick
O
n Thursday, March 9, 2023, depositors withdrew almost 25 percent of the
total deposits at Silicon Valley Bank (SVB). On the morning of Friday,
March 10, early evidence suggested that virtually all of the remaining
deposits would be withdrawn. Regulators were forced to close the bank in the
middle of the day, unable to even “get to the weekend” as had been done for every
other bank failure since the Great Depression. Silicon Valley Bank, with $209 billion
of assets, the sixteenth-largest bank in the United States, became the second-largest
bank failure in nominal terms in US history, ranking in size only behind the failure
of Washington Mutual Bank during the Great Recession of 2007–2009.
Since the founding of the Federal Deposit Insurance Corporation (FDIC) in
1934, no US bank depositor has lost any insured funds. As a result, US bank runs
have been rare, because depositors see no need to rush to the exits, even if a bank
failure is imminent. However, US deposit insurance has a limit on the amount
insured of $250,000 per account. Silicon Valley Bank was highly unusual in that
about 94 percent of its deposits are large enough to be uninsured. This makes SVB a
throwback, almost as though it was from the bank-panic frenzies of the late nine-
teenth century, before the modern safety net for banks was in place. It also makes
the failure of SVB an illustrative model for the economics of banking—how things
work can often become most clear when they break.
After the failure of Silicon Valley Bank, commentary focused on several
possible causes: poor risk management by the bank; changes in bank regulation
■ Andrew Metrick is the Janet L. Yellen Professor of Finance and Management, Yale School
of Management, New Haven, Connecticut. His email address is [email protected].
For supplementary materials such as appendices, datasets, and author disclosure statements, see the
article page at https://doi.org/10.1257/jep.38.1.133.
134 Journal of Economic Perspectives
in 2019 that had recently relaxed the rules for banks like SVB; weak oversight by
supervisors for the rules that remained; a customer base highly concentrated in one
industry (technology, particularly startups) and region (Silicon Valley, of course);
and the interaction of modern social media with the tech-savvy customer base of
SVB in accelerating the run. From this long list, one might conclude that SVB was
a special case, an outlier that reveals little about bank stress in general. This would
not, however, be correct, because the underlying problem was systemic, and when
SVB was shut down, the Panic of 2023 had begun.
On Sunday, March 12, the Federal Deposit Insurance Corporation took control
of another failed institution, Signature Bank, which then became the third-largest
bank failure in the history of the United States. This failure convinced authorities
that the US banking system faced the possibility of widespread runs. In response, the
FDIC and the Federal Reserve each invoked emergency provisions available under
the law. The “systemic-risk exception” included in a 1991 law allows the FDIC to act
immediately without taking time to seek out a least-cost method of resolving a bank
failure: in this case, the FDIC announced that even uninsured depositors would be
fully covered at both Silicon Valley Bank and Signature. In turn, the Federal Reserve
invoked its power to take action during “unusual and exigent circumstances” that
was included in a 1935 revision of the Federal Reserve Act and created a novel
emergency-lending program with generous terms. Together, these policy actions
succeeded in slowing the deposit outflows, which delayed but did not prevent
the eventual failure of First Republic Bank in May. At $213 billion in assets, First
Republic was even larger than SVB, thus leapfrogging to become the “new” second-
largest bank failure ever in the United States. Panic in the US banking sector also
spread to Europe, providing the final nail in the coffin for the long-troubled Credit
Suisse, which in April 2023 became the single largest bank failure in the world, ever.
Each of these failures had their own idiosyncratic reasons. But the US examples
all shared two things: a very high percentage of uninsured deposits and significant
“unrealized” losses on assets. That combination is a dangerous mix, and variations
on it lie at the core of most financial panics. In most modern examples, such as
the global financial crisis of 2008–2009, this dangerous mix is shrouded in layers of
complexity, making it much harder to uncover its dynamics. But the special case of
Silicon Valley Bank turns out to be a remarkably clean example, and thus can teach
us a lot about the general case.
To elicit these lessons, we focus on underlying economics of the Silicon Valley
Bank failure, leaving most of the discussion of government regulation until after
the basic story is told. In the next section, we examine the balance sheet of SVB at
the end of 2019 and thus before the pandemic, focusing on the largest categories
of assets and liabilities. Like most banks, SVB’s profits are driven mostly by “net
interest margin”—the difference in interest earned and paid for via its assets and
liabilities—and this business model has come under pressure since the onset of
the pandemic. In the following section, we illustrate this pressure using diagrams
showing the components of net interest margin and show how those components
changed during the pandemic. With this background of SVB’s balance sheet and
Andrew Metrick 135
business model, we can then make sense of what happened at SVB and the ensuing
Panic of 2023. In the penultimate section of the paper, we turn to the role of govern-
ment bank regulation and oversight in the lead up to the crisis. The final section
concludes with a discussion of the policy challenge in preventing future panics of
this type.
The top panel of Figure 1 shows the balance sheet for Silicon Valley Bank on
December 31, 2019. At this time, the bank had a total of $70 billion of assets. The
single largest item on the balance sheet was $63 billion of deposits, which comprised
the vast majority of the bank’s funding. SVB was unusual in that almost all of these
deposits—about 94 percent—were not covered by federal deposit insurance.
In the United States, individual bank accounts are insured by the Federal Deposit
Insurance Corporation up to a limit of $250,000. Balances below $250,000 are
“insured,” while all balances above $250,000 are “uninsured.” This limit is high
enough that the vast majority of all accounts are fully insured, but on a dollar basis,
about 42 percent of the total deposits in the United States were above this limit, and
thus uninsured (FDIC 2023c, p. 10). The specific choice of $250,000 as the legal
threshold is an attempt to balance the benefits (financial stability) and costs (excess
risk-taking by banks) of insurance. The Panic of 2023 has reopened the debate on
this threshold—a topic we discuss in the conclusion of the paper.
FDIC insurance is backed by a standing fund financed by insurance premiums
charged to banks, with a credit line from the US Treasury and the full faith and credit
of the United States standing as additional lines of defense. For insured depositors,
the typical bank failure is seamless. Uninsured deposits do not have this explicit
protection. Nevertheless, in most bank failures the FDIC has been able to transfer
the entire deposit base to another bank, allowing the uninsured depositors to be
made whole. Indeed, the total losses by uninsured depositors in the last 30 years
has been less than $300 million (FDIC 2023c, p. 22). With this history, it would be
reasonable for uninsured bank depositors to feel relatively safe, even if that safety
has not been legally guaranteed. This implicit safety supports the significant level of
uninsured deposits in the US banking system. Most of these uninsured deposits are
concentrated in the largest banks. Even among this group, however, Silicon Valley
Bank stood out for having the highest percentage of uninsured deposits among all
banks with assets of $50 billion or more.
On the asset side of the Silicon Valley Bank balance sheet in 2019, the largest
line item was “loans” at $33 billion, followed by “securities” at $28 billion. Most of
the loans were in the form of credit lines to SVB customers, instead of more tradi-
tional term loans for specific projects; the startup firms served by SVB often have
lumpy inflows and outflows and rely on credit lines to smooth those lumps. Very few
of the loans were residential or commercial mortgages. The securities holdings were
almost exclusively government or agency-backed bonds. With securities holdings at
136 Journal of Economic Perspectives
Figure 1
SVB Balance Sheet, 2019 and 2022
40 percent of all assets, SVB was above the national average of 25 percent, but was
not an extreme outlier. Banks hold securities for two main reasons: (1) as a higher-
yielding (compared to cash) way to store funds that they plan to later loan out; and
(2) as a source of liquidity, because safe government securities can easily be sold for
cash to meet deposit withdrawals. Plain “cash” was only at $6 billion.
Bank assets face two main risks: “credit risk” and “interest-rate risk.” Credit risk
is that a borrower will not make timely payments of interest or principal; this risk is
what usually gets banks into trouble, but was not a problem for Silicon Valley Bank.
Their loan portfolio was concentrated in the ecosystem of startup technology firms,
the venture capital firms that fund them, and the employees of both. Although
this portfolio was not well-diversified, SVB never faced serious concerns about the
quality of its loans. Furthermore, SVB’s securities portfolio was almost exclusively
The Failure of Silicon Valley Bank and the Panic of 2023 137
cushion that protects depositors and other debtholders from variations in the value
of the assets. To the extent that assets are completely riskless in all respects, any posi-
tive level of capital would be sufficient to provide safety for all depositors, insured
and uninsured. But once any uncertainty arises about the value of assets, a rational
depositor (or insurer) would require more capital to feel safe.
One useful way to conceptualize the necessary capital cushion for a bank
is to think of it as an input in the production of money. To satisfy the value of
bank deposits (like checking accounts) as a transactions medium, it must be
possible to use these deposits in exchange at par value with “no questions asked”
(Holmstrom 2015). If an agent has to worry that the money they are receiving is not
actually worth the number written on the check, then it will lose its usefulness as
money. Having a sufficient capital buffer is the most important step a bank can take
to ensure this monetary use. Conversely, if equity levels fall too low, then it would
be reasonable for counterparties to be concerned about the quality of the liabilities,
and the bank’s checks and deposits would lose their no-questions-asked property.
The link between the potential solvency of a bank (through its equity levels) and
the liquidity of the bank (through the risk of bank runs) is driven by the connection
of both to the no-questions-asked property of deposits. When bank money loses its
no-questions-asked property, the rational response for (uninsured) depositors is to
take their money out of the bank and put it someplace safer. Bank balance sheets
contain many components that are costly to evaluate. No depositor earning a small
(or zero) interest rate on their deposits should do a deep analysis of their bank to
determine whether or not it was still safe—withdrawing deposits is much easier.
At its core, banking is a simple business. Almost all of a bank’s assets and liabili-
ties are essentially just digital equivalents of pieces of paper: mortgages, commercial
loans, checking accounts, cash reserves, and so on. The profitability of a bank is
driven by earning higher interest on the assets than is paid on the liabilities, with
this difference known as the “net interest margin.” Indeed, one can think of a bank
of producing a consumer good (“deposit services”) that is itself just a liability, using
the rest of its balance sheet as the factory floor for this production.1
1
US banks earn about two-thirds of their revenue from net interest. The remaining noninterest revenue
comes from fees and service charges on deposit accounts and credit cards, investment banking, asset
management, and other activities. See Haubrich and Young (2019) for an analysis of these components
over time. The Silicon Valley Bank business model focused mainly on net interest: “Net interest income
accounts for the major portion of our earnings. It is comprised primarily of income generated from
interest rate spread differences between the interest rates received on interest-earning assets, such as
loans extended to clients and securities held in our fixed income securities portfolio, and the interest
rates paid by us on interest-bearing liabilities, such as deposits and borrowings” (Silicon Valley Bank
Financial Group 2023, p. 7).
Andrew Metrick 139
Figure 2
Silicon Valley Bank and the Business Model of Banking
4
2019 “risky loan” curve
Yield (percent)
Credit spread
3
1
Deposit spread
2019 deposit rate
0
1-mo 2-mo 3-mo 6-mo 1-yr 2-yr 3-yr 5-yr 7-yr 10-yr 20-yr 30-yr
Duration
Figure 2 illustrates the three main sources of net interest margin for Silicon
Valley Bank as of December 31, 2019. The y-axis shows interest rates (= “yield”)
measured as an annualized percentage; the x-axis shows time. Starting in the
middle of the figure, the solid blue line shows the term structure of government
debt: the market interest rates paid by the government to borrow at various maturi-
ties, ranging from one month to 30 years. (As is standard in these kinds of figures,
the x-axis is not drawn to scale, instead showing many observations for short-term
maturities before getting sparser at the long end.) The graph of these government
interest rates, represented by the blue line, is called the “yield curve.” The curve
begins at approximately 1.5 percent for Treasury bills with one month remaining
on their term, rising to about 2 percent for 10-year bonds and 2.5 percent for
30-year bonds. The “term spread” shown on the right side of the figure represents
the difference between the yields on the shortest- and longest-horizon government
debt. This term spread is positive, which is the usual condition. To the extent that
SVB—like most banks—will be borrowing at the short end of the curve and lending
at the long end of the curve, a bank will capture some component of this term
spread as part of its net interest margin.
If all the bank did was to borrow short-term at the government rate and then
lend long-term back to the government, the net interest margin would be exactly
equal to the term spread. But banks can improve upon the seemingly risk-free
140 Journal of Economic Perspectives
borrowing rates of the government. At the bottom left of the figure, the solid circle
at 0.4 percent represents the average interest rate paid by Silicon Valley Bank on
deposits at the end of 2019. Because these deposits can be demanded at any time,
their legal term-to-maturity is close to zero, and thus the solid circle lies right next
to the y-axis. The vertical difference between this solid circle and the leftmost point
on the blue government yield curve is the “deposit spread” earned by SVB. Effec-
tively, SVB’s customers are willing to accept an even lower return than they would
get in government T-bills because of the transaction services and other perqui-
sites provided by their demand deposits at the bank. Unlike Treasury bills, bank
deposits can be used to facilitate transactions; many components of the legal ability
to provide these services are unique to banks.
Figure 2 also includes a dotted line extended from the solid circle. This dotted
line represents the expectation that deposits are “sticky” in the sense that depositors
do not switch banks in the short run. In the limit, if we make an extreme (coun-
terfactual) assumption that most depositors make a lifetime commitment to their
bank at current dotted-line rates, then any further increase in the blue line would
effectively be a windfall for the bank.
Of course, in normal times we do not expect banks to invest all of their deposits
in government securities. As seen in the Silicon Valley Bank balance sheet in Figure 1,
approximately half of the assets at SVB were in securities at the end of 2019, with the
other half in more traditional loans to businesses and individuals. Unlike govern-
ment securities, these traditional loans are risky and would be expected to earn a
premium over the government yield curve. This “risky-loan” curve is shown in red,
with the difference between the red and blue curves shown on the right as the
“credit spread.” The risky-loan curve represents an estimate of expected yields on
the loans, drawn just for expositional purposes: no public data is available to show
the rates on the SVB loan book, so we just use a fixed premium here of 1.5 percent.
Note that if a bank turns out to have chosen its loans poorly, the realized return on
risky loans could easily lie beneath the government yield curve.
Taken together, the net interest margin in any given period would be the sum
of these three components: the term spread, the deposit spread, and the credit
spread. In ordinary times, when a bank is able to pay deposit rates below the govern-
ment borrowing rate (positive deposit spread), when the government yield curve
slopes up (positive term spread), and when risky loans are being paid back (posi-
tive realized credit spread), banks will have healthy net interest margins. For 2019,
Silicon Valley Bank realized a net interest margin above 3 percent.
But the monetary policy response to COVID-19 led to major changes. Figure 3
traces the evolution of the government yield curve for the next three years. The
blue line gives the starting point at the end of 2019, with the same curve as shown
in the same color on Figure 2. Even though this curve was already low by historical
standards, the response of the Federal Reserve to the pandemic drove rates even
lower. The orange line shows the government yield curve at the end of 2020. Short-
term rates were effectively zero, and Silicon Valley Bank—like all banks—was no
longer earning a positive deposit spread.
The Failure of Silicon Valley Bank and the Panic of 2023 141
Figure 3
Treasury Yield Curves, 2019–2022
5
2022
4
Yield (percent)
2
2019
1
2021
2020
0
1-mo 2-mo 3-mo 6-mo 1-yr 2-yr 3-yr 5-yr 7-yr 10-yr 20-yr 30-yr
Duration
Had the yield curve stayed at these historical lows, Silicon Valley Bank would
have been fine. Although the deposit spreads available at the end of 2019 had disap-
peared, the term spread still existed and the legacy loans from before the COVID era
were continuing to perform, preserving credit spreads from that time. But the Fed
began fighting the COVID-era inflation in earnest in early 2022, and by the end
of the year the government yield curve (in purple) was strikingly different from
its blue, orange, and green predecessors; both the level and the slope of the curve
had changed. On December 31, 2022, the yield curve for government securities
was relatively flat at about 4 percent for all maturities, eliminating the term-spread
component.
Figure 4 illustrates the main changes to Silicon Valley Bank’s end-of-year balance
sheet from 2019 to 2022, using the same color scheme as for interest rates in Figure
3. Taken together, these two figures show how the rapid growth at SVB coincided
with changes in the interest-rate environment during the COVID era. In 2020, the
powerful easing of rates was also combined with aggressive expansionary fiscal
policy. The combined effect of these twin policies was felt particularly strongly in
the tech sector of the Silicon Valley region. Younger firms were awash in cash, and a
lot of that cash made its way to SVB, where deposits more than tripled from the end
of 2019 to the end of 2021. Even in the best of times, it would have been difficult
to find enough good lending opportunities to soak up these deposits. SVB did lend
out some of these deposits, but the majority went into the purchase of securities.
The same dynamics applied in 2021: the loan book grew, but not as fast as the secu-
rities portfolio.
142 Journal of Economic Perspectives
Figure 4
SVB: Loans, Deposits, and Securities, 2019–2022
150
$ Billion
100
50
0
Deposits Loans Securities
There is no mystery about this mechanical relationship, and it was well under-
stood by management, bank supervisors, and market analysts. Indeed, most of
these losses had already been incurred by the end of the third quarter of 2022, and
the information about these losses was available in public filings by the bank well
before the end of 2022. By this measure, Silicon Valley Bank was already insolvent
by December 31, 2022, as shown in the bottom panel of Figure 1, with negative
$3 billion in equity. From these facts, one might wonder what regulators, inves-
tors, and uninsured depositors could possibly have been thinking to allow SVB to
continue operating for so long after this apparent “insolvency” was clear.
The answer here is that Silicon Valley Bank relied on the stability of their
customer relationships—the “stickiness” of deposits—as their main method to
hedge interest-rate risk. At first glance, this belief may seem counterintuitive, and
perhaps even a little magical. Most deposits are in-demand accounts and can be
withdrawn at any time. In principle, if interest rates rise, one would expect that
either deposit rates would rise or customers would leave and go to another bank.
But reality is different. Loyalty, transactions costs, and simple inertia keep many
consumers at their banks even when interest rates change—indeed, this stickiness is
the reason for the dotted line in Figure 2.
Silicon Valley Bank is not unusual in relying on the stability of its deposits. In
an influential paper, Dreschler, Savov, and Schnabl (2021) demonstrate that profits
from the deposit spread have been a remarkably good hedge for interest-rate risk
for US banks. Their analysis shows that deposit rates are quite inelastic to market
interest rates, so that an increase in market rates leads to an increase in deposit
spreads. Banks build their business plans around this relationship, using marketing,
branch networks, and personal service to maximize the stability of their deposit
base. We can think of the net present value of this deposit spread like an additional
asset for banks, the “franchise value of deposits,” but one that is never included in
any formal balance sheets. When interest rates rise, the deposit spread increases and
this franchise value goes up, but not even “mark-to-market” accounting will capture
this change.
The rise in interest rates in 2022 caused significant losses for the entire banking
system (Jiang et al. 2023; Flannery and Sorescu 2023). Hundreds of banks were
likely close to insolvency if we were to accurately mark-to-market only the securities
and loans of their balance sheets. If we try to account for the increased value of the
deposit franchise, the insolvency often disappears. If deposits lose their stickiness,
however, then the value of the deposit franchise will fall, and fear of insolvency can
cause actual insolvency. For Silicon Valley Bank, that is exactly what happened next.
this measure Silicon Valley Bank was “insolvent” on paper. So why didn’t anything
happen until March 2023?
The crucial shock arrived on March 8, 2023, when Silicon Valley Bank announced
that it had sold $24 billion of book value securities for a loss of $1.8 billion, along
with a plan to raise $2.25 billion of new equity. This announcement contained new
information. By selling the securities, SVB was now required to recognize these
losses on its income statement and balance sheet—the losses were now “realized.”
Banks would prefer not to realize losses, and it was reasonable to conclude that the
bank did not have better options. Outside onlookers could ask reasonable ques-
tions: Maybe the deposit outflow is more severe than we thought? Maybe SVB will
need to recognize even more losses to the point that they are actually insolvent?
When these kinds of questions become salient, that is all it takes for uninsured
deposits to lose their no-questions-asked property.
Once the no-questions-asked property is lost, the cost-benefit calculation for unin-
sured depositors changes. While history suggests that uninsured depositors have a good
chance of ultimately getting all of their money back, it also suggests the possibility of
significant delays before this happens. Thus, uninsured depositors knew that if Silicon
Valley Bank failed, the potential costs of delay could be material, easily surpassing the
transactions costs of switching banks. Many depositors seemed to share this thinking
on Thursday, March 9, 2023. The $42 billion withdrawn that day represented almost
one-quarter of SVB deposits. Once the news of those withdrawals spread, even more
depositors started asking questions. On Friday, March 10, the pace of withdrawals was
so rapid that the authorities did not expect SVB to make it through the day.
The Federal Deposit Insurance Corporation took the bank into receivership
in the morning—the first intraday receivership in its history. Its announcement on
Friday, March 10, 2023, showed that uninsured depositors were right to be concerned:
“All insured depositors will have full access to their insured deposits no later than
Monday morning, March 13, 2023. The FDIC will pay uninsured depositors an
advance dividend within the next week. Uninsured depositors will receive a receiver-
ship certificate for the remaining amount of their uninsured funds. As the FDIC sells
the assets of Silicon Valley Bank, future dividend payments may be made to uninsured
depositors” (FDIC 2023a).
On that Friday, uncertainty was all the Federal Deposit Insurance Corpora-
tion could promise for uninsured depositors. But the failure of Silicon Valley Bank,
along with the murky outlook for uninsured depositors, led to a violation of the
no-questions-asked property at many other banks. In particular, depositors ran from
banks that had the same dangerous combination as SVB: a high percentage of unin-
sured deposits combined with significant unrealized losses on their assets. Two large
banks were particularly hard-hit by deposit outflows that same Friday, March 10:
Signature Bank, with $110 billion in assets, had about $10 billion in outflows, and
First Republic Bank, with $213 billion in assets, had about $25 billion in outflows.
Signature was particularly unprepared for this event. Many banks have collateral
“pre-positioned” at the Federal Reserve, which involves filling out paperwork which
designates certain bank assets and assures that the Fed would have a top priority
The Failure of Silicon Valley Bank and the Panic of 2023 145
claim on those assets if the bank wants to use them as collateral for emergency
discount-window borrowing. However, Signature Bank had no pre-positioned
collateral, and thus would not have been able to access emergency borrowing from
the Fed. This unpreparedness, combined with long-standing regulatory concerns
about the management of the bank, convinced regulators to close the bank on
Sunday, March 12, and create another FDIC receivership.
As the Federal Deposit Insurance Corporation sought to reassure the unin-
sured depositors, it faced a substantial obstacle. In any bank resolution, the FDIC
is required by law to cover the insured depositors at the “least cost” to the deposit
insurance fund. In many cases, another bank is often willing to take on all of the
deposits of the failed bank in order to obtain those banking relationships for itself.
However, the Silicon Valley Bank failure happened so fast that there was no time to
arrange for such a purchase, nor any way to be certain that any such purchase would
eventually occur. The FDIC would have had a hard time making a convincing case
it was “least cost” to make noncontractual payments to uninsured depositors. The
only way around the least-cost requirement is for the FDIC to invoke a “systemic risk
exception,” enacted in 1991 legislation. This exception cannot be invoked casually:
doing so requires a positive vote from the boards of both the FDIC and Federal
Reserve, as well as approval by the US Secretary of the Treasury. Invoking the excep-
tion also brings close scrutiny to the FDIC, especially since Silicon Valley Bank had
not previously been classified as being systemically important.
However, the Federal Deposit Insurance Corporation invoked the systemic risk
exception on Sunday, March 12, and promised to protect uninsured depositors,
thus showing how much can happen in two days (FDIC 2023b):
After receiving a recommendation from the boards of the FDIC and the Federal
Reserve, and consulting with the President, Secretary Yellen approved actions
enabling the FDIC to complete its resolution of Silicon Valley Bank, Santa
Clara, California, in a manner that fully protects all depositors. Depositors will
have access to all of their money starting Monday, March 13. No losses associ-
ated with the resolution of Silicon Valley Bank will be borne by the taxpayer.
We are also announcing a similar systemic risk exception for Signature Bank,
New York, New York, which was closed today by its state chartering authority.
All depositors of this institution will be made whole. As with the resolution of
Silicon Valley Bank, no losses will be borne by the taxpayer.
What changed from Friday, March 10, to Sunday, March 12? Given the informa-
tion now publicly available, I speculate that on Friday, March 10, when Silicon Valley
Bank was closed, the authorities had hoped that its failure was idiosyncratic and that
the panic would not spread to other banks—or at least would not spread quickly.
However, by Sunday, March 12, it was clear that the panic was contagious, and that
the no-questions-asked property had been lost for unsecured deposits at banks with
large unrealized losses.
146 Journal of Economic Perspectives
2
The historical development of capital rules is described in Haubrich (2020). The changes to inter-
national standards imposed after the global financial crisis are described in McNamara, Wedow, and
Metrick (2019). A detailed discussion of the current rules and acceptable minimums in the United States
is Scott and Labonte (2023). Aggregate statistics for insured banks at the end of 2019 is in FDIC (2020).
3
A summary of the current international standards for liquidity regulation is given in Bank for Interna-
tional Settlements (2013). The specific application of these standards for Silicon Valley Bank is analyzed
in Feldberg (2023a).
148 Journal of Economic Perspectives
Bank Financial Group 2023). Because the San Francisco Fed has a much larger staff
than the California Department of Financial Protection and Innovation, and also
much more experience in the examination of large banks, the Fed took the lead
on most of the examinations. Finally, as the insurer, the Federal Deposit Insurance
Corporation is the “backup federal regulator,” and had access to all information
from the supervisory process.
In the aftermath of the Panic of 2023, the performance of these regulations—
and the supervisors that enforce them—came under close scrutiny, with several
government agencies producing reports (DFPI 2023; Federal Reserve Board 2023;
FDIC 2023d; Government Accountability Office 2023). The most comprehensive
of these reports came from the Federal Reserve, the agency with the most super-
visory resources dedicated to Silicon Valley Bank. The report concludes that SVB
was in compliance with the specific capital and liquidity rules in place at the time
of its failure, but faults itself for a set of recent reforms that kept SVB (and other
similarly-sized banks) from having more stringent standards. They conclude,
however, that there is no guarantee that these more stringent standards would
have prevented SVB’s collapse. The Fed report also criticizes its own lack of super-
visory zeal, saying that supervisors were aware of weaknesses at SVB, but were not
aggressive enough in using their discretion to insist on timely corrections. This
conclusion is well supported by evidence, but the remedy is not obvious. “Supervi-
sors should do their jobs better” is not an easy goal to achieve.
To illustrate the challenge faced by supervisors, it is helpful to examine the largest
source of concern—interest-rate risk. Many of the official reports emphasized the
mismanagement of interest-rate risk by Silicon Valley Bank and the failure of regula-
tion and supervision to correct this mismanagement. The Fed report concludes that
“we need to evaluate how we supervise and regulate a bank’s management of interest
rate risk. While interest rate risk is a core risk of banking that is not new to banks or
supervisors, SVB did not appropriately manage its interest rate risk, and supervisors
did not force the bank to fix these issues quickly enough” (Federal Reserve Board
2023, p. 3).
But on this point, the regulations themselves do not provide clear guid-
ance to supervisors. In the computation of regulatory capital measures, most
mark-to-market interest-rate losses are exempt from inclusion, allowing banks to
ignore such losses in meeting required ratios. Even the regulatory “stress tests”
performed on the largest banks, with binding implications for their dividend and
capital policy, do not include a specific stress for interest-rate risk. I think it is
asking too much of supervisors to enforce an ethos that is not already in the plain
language of the rules.
In contrast, the European Union carefully monitors “interest-rate risk in the
banking book” and uses a variety of mechanisms to discourage banks from taking on
too much interest-rate risk. Their methods work: EU banks have lower interest-rate
risk then do US banks. But these rules do not make the risk disappear—they just
shift it to a different place. For example, one main mechanism in the EU financial
system to move interest-rate risk away from banks is for more borrowers to pay
The Failure of Silicon Valley Bank and the Panic of 2023 149
variable interest rates. In that case, if interest rates go up, bank balance sheets do
not take a hit, but borrowers have to make higher payments. There, the risk of rising
interest rates sits with borrowers.4
As long as actual investments take time to mature, the provision of liquidity will
be a risky endeavor. That risk needs to sit somewhere, and no amount of supervisory
vigilance can make this risk disappear. The European Union has chosen to shift this
risk to borrowers, while the United States allows much more of that risk to sit on the
balance sheets of banks. To the extent that the deposit franchise is a natural hedge
for this interest-rate risk, there is a benefit in leaving this risk with banks. But there
is also a potential cost, as we saw in March 2023.
Conclusion
The alchemy of banking is that solvency and liquidity reinforce each other.
Bank deposits remain money-like because of the assurance of solvency. Money-like
deposits tend to be sticky within a given financial institution, because it is costly for
depositors to switch banks and replicate an earlier long-term relationship with a new
institution. This stickiness gives banks a form of monopoly power, the rents earned
from this power comprise a large portion of bank profits, and thus the capitalization
of these rents comprises a large portion of the bank’s market value. Therein lies the
alchemy. As long as the deposit base is stable, the expectation of future profits from
deposit spreads adds to the solvency strength that supports the very same deposit
stickiness. An information event that calls into question a bank’s solvency can quickly
become self-reinforcing, as a large component of the solvency disappears when the
questions are asked openly. This is the main story of Silicon Valley Bank.
If the stability of the deposit base plays such a central role, why not simply increase
the maximum level of deposit insurance? For example, during the global financial
crisis of 2008–2009, the Federal Deposit Insurance Commission introduced the
Transaction Account Guarantee (TAG) program, which provided unlimited insur-
ance for non-interest-bearing accounts. This program was temporary—expiring at
the end of 2010—and has generally been considered to be successful at calming
depositors during the worst times of that crisis.
There are several challenges for making a program like the Transaction Account
Guarantee permanent. First, when the program was deployed in 2009–2010,
interest rates were headed toward zero. Depositors gave up very little by sticking
with non-interest-bearing accounts in that environment. In the higher-interest-rate
environment that we have today—and which we should expect to recur at least
occasionally—the incentives for depositors to leave banks for other vehicles would
be significantly stronger. Second, if we went further than the Transaction Account
Guarantee and allowed unlimited deposit insurance even for interest-bearing
4
For a discussion of the US approach to interest-rate risk as compared with the rest of the world, see
Feldberg (2023b). For an analysis of EU interest-rate risk, see Dries et al. (2022).
150 Journal of Economic Perspectives
deposits, we should not be surprised to see clever bankers finding ways to sell all
kinds of financial products that are ultimately backed by an infinite safety net. For
an intermediate solution of raising the deposit limit above $250,000 but less than
infinity, the same two problems exist, just in intermediate form.
The challenge of finding the “right” level of deposit insurance is a good
example of the general problem of finding the right balance in financial regulation
writ large. We should—of course!—try to do the best job we can in enforcing the
regulations currently on the books. But in writing new regulations, we must recog-
nize the tradeoffs. In general, bank regulation cannot change the total amount of
risk in the economy, but it is highly effective at shifting that risk around. Banks, their
customers, their substitutes (“shadow banks”), government, and taxpayers—what is
our choice for who should hold what share and what kind of risk?
■ I thank Greg Feldberg, Eric Hurst, Steven Kelly, Nina Pavcnik, Timothy Taylor, Rosalind
Wiggins, and Heidi Williams for helpful comments and Vincient Arnold, Anmol Makhija,
and Carey Mott for excellent research assistance.
References
Bank for International Settlements. 2013. Basel III: The Liquidity Coverage Ratio and Liquidity Risk Moni-
toring Tools. Basel: Bank for International Settlements.
California Department of Financial Protection and Innovation (DFPI). 2023. “Review of DFPI’s Oversight
and Regulation of Silicon Valley Bank.” Sacramento: DFPI.
Dries, Jonathan, Benjamin Klaus, Francesca Lenoci, and Cosimo Pancaro. 2022. “Interest Rate Risk
Exposures and Hedging of Euro Area Banks’ Banking Books.” In Financial Stability Review, May
2022, 67–70. Frankfurt: European Central Bank.
Drechsler, Itamar, Alexi Savov, and Philipp Schnabl. 2021. “Banking on Deposits: Maturity Transforma-
tion without Interest Rate Risk.” Journal of Finance 76 (3): 1091–1143.
FDIC. 2020. “Quarterly Banking Profile: Fourth Quarter 2019.” FDIC Quarterly 14 (1).
FDIC. 2023a. “FDIC Creates a Deposit Insurance National Bank of Santa Clara to Protect Insured Deposi-
tors of Silicon Valley Bank, Santa Clara, California.” PR-16-2023. https://www.fdic.gov/news/
press-releases/2023/pr23016.html.
FDIC. 2023b. “Joint Statement by the Department of the Treasury, Federal Reserve, and FDIC.” PR-17-
2023. https://www.fdic.gov/news/press-releases/2023/pr23017.html.
FDIC. 2023c. FDIC’s Supervision of Signature Bank. Washington, DC: FDIC.
FDIC. 2023d. Options for Deposit Insurance Reform. Washington, DC: FDIC.
Federal Reserve Board. 2023. Review of the Federal Reserve’s Supervision and Regulation of Silicon
Valley Bank. Washington, DC: Federal Reserve Board.
Feldberg, Greg. 2023a. “Lessons from Applying the Liquidity Coverage Ratio to Silicon Valley Bank.” Yale
School of Management, March 27. https://som.yale.edu/story/2023/lessons-applying-liquidity-
coverage-ratio-silicon-valley-bank.
Feldberg, Greg. 2023b. “US Banks’ Interest-Rate Risk Reporting and Regulation: A Comparative
Context.” Yale School of Management, April 26. https://som.yale.edu/story/2023/us-banks-
interest-rate-risk-reporting-and-regulation-comparative-context.
Andrew Metrick 151
Flannery, Mark J., and Sorin M. Sorescu. 2023. “Partial Effects of Fed Tightening on U.S. Banks’ Capital.”
Unpublished.
Government Accountability Office. 2023. “Bank Regulation: Preliminary Review of Agency Actions
Related to March 2023 Bank Failures.” GAO-23-106834. https://www.gao.gov/products/gao-23-
106834.
Haubrich, Joseph G. 2020. “A Brief History of Bank Capital Requirements in the United States.” Federal
Reserve Bank of Cleveland Economic Commentary 2020-05.
Haubrich, Joseph G., and Tristan Young. 2019. “Trends in the Noninterest Income of Banks.” Federal
Reserve Bank of Cleveland Economic Commentary 2019-14.
Holmstrom, Bengt. 2015. “Understanding the Role of Debt in the Financial System.” BIS Working Paper
479.
Jiang, Erica Xuewei, Gregor Matvos, Tomasz Piskorski, and Amit Seru. 2023. “Monetary Tightening
and U.S. Bank Fragility in 2023: Mark-to-Market Losses and Uninsured Depositor Runs?” NBER
Working Paper 31048.
McNamara, Christian, Michael Wedow, and Andrew Metrick. 2019. “Basel III B: Basel III Overview.”
Journal of Financial Crises 1 (4): 59–69.
Scott, Andrew P., and Marc Labonte. 2023. Bank Capital Requirements: A Primer and Policy Issues. CRS
Report R47447. Washington, DC: Congressional Research Service.
Silicon Valley Bank (SVB). 2019. “Call Report at the Close of Business December 31, 2019.” RSSD-ID
802866. https://cdr.ffiec.gov/public/ManageFacsimiles.aspx#.
Silicon Valley Bank (SVB). 2020a. “Q4 and FY 2019 Financial Highlights.” https://s201.q4cdn.
com/589201576/files/doc_financials/2019/q4/Q4’19-IR-Presentation_1.31.20.pdf.
Silicon Valley Bank (SVB). 2020b. “Call Report at the Close of Business December 31, 2020.” RSSD-ID
802866. https://cdr.ffiec.gov/public/ManageFacsimiles.aspx#.
Silicon Valley Bank (SVB). 2021. “Call Report at the Close of Business December 31, 2021.” RSSD-ID
802866. https://cdr.ffiec.gov/public/ManageFacsimiles.aspx#.
Silicon Valley Bank (SVB). 2022. “Call Report at the Close of Business December 31, 2022.” RSSD-ID
802866. https://cdr.ffiec.gov/public/ManageFacsimiles.aspx#.
Silicon Valley Bank Financial Group. 2023. “Form 10-K for the Fiscal Year Ended December 31, 2022.”
Washington, DC: United State Securities and Exchange Commission https://d18rn0p25nwr6d.
cloudfront.net/CIK-0000719739/f36fc4d7-9459-41d7-9e3d-2c468971b386.pdf.
US Department of the Treasury. 2023. “Daily Treasury Par Yield Curve Rates.” https://home.
treasury.gov/resource-center/data-chart-center/interest-rates/TextView?type=daily_treasury_
yield_curve&field_tdr_date_value_month=202311.
152 Journal of Economic Perspectives