When will they ever learn? The US banking crisis of 2023

Anat Admati is Professor of Finance and Economics at the Stanford Graduate School of Business, Martin Hellwig is Director (em.) of the Max Planck Institute for Research on Collective Goods, and Richard Portes is Professor of Economics at London Business School

At a press conference on 3 May 2023, Jay Powell, the Federal Reserve Chair, suggested that “the resolution and sale of First Republic … is an important step toward drawing a line under (the) period of severe stress” that started with Silicon Valley Bank (SVB)’s 10 March failure. Mr Powell’s statement suggests that either he did not understand the nature and the extent of the problems afflicting US banks or that he understood but chose to ignore them. Markets have not accepted the view he expressed1.

The crisis in US banking is ‘systemic’, concerning a large part of the banking system not because banks are so interconnected but because they have followed similar strategies and are now in a similar bind. SVB is an extreme example, but the difficulties of SVB suggest lessons about other banks in the US and elsewhere.

On 31 December 2019, SVB’s financials showed $62 billion in deposits, $33 billion in loans, and $29 billion in securities2. On 31 March 2022, 27 months later, the bank’s disclosures included $198 billion in deposits, $68 billion in loans, and $127 billion in securities. The massive inflows consisted mostly of large uninsured corporate deposits.

With interest rates close to zero for several years, corporate treasurers saw no gains from investing in money market instruments rather than deposits.

SVB lending did not keep pace with the tripling of deposits during this period. Excess inflows were placed in safe, fixed-interest securities such as government bonds and agency debt. But ‘safety’ does not mean absence of risks to market values from interest rate changes.

In 2022, the Federal Reserve started raising interest rates to fight inflation. Investors then moved gradually from deposits to money market investments that paid higher interest. By 31 December 2022, SVB’s deposits had declined by $25 billion.

In response, it reduced its cash, borrowed from the Federal Home Loan Bank, and slowed replacements of maturing securities. In the first two months of 2023, deposits declined further. On 8 March, the bank announced that $21 billion worth of securities had been sold at a loss of $1.8 billion.

To compensate for the loss, it would raise new equity. Potential equity investors shied away and depositors became aware that SVB had incurred losses on securities and could not raise more equity. They reacted with a massive run, withdrawing $42 billion on 9 March. On 10 March, the authorities closed the bank, citing “inadequate liquidity and insolvency.”

SVB had in fact been insolvent at least since September 2022, but its accounting practices concealed that fact. Interest rate increases since March 2022 had led to a substantial decline in the fair values of SVB’s assets. These losses hardly appeared in the accounts because SVB had classified most securities and loans as ‘held to maturity’ (HTM).

For such assets, accounting rules do not require acknowledging valuation losses from interest rate increases. If the assets are held to maturity and there is no concern about whether promised payments will be made in full, there will ultimately be no losses, goes the logic.

Holding assets to maturity, however, requires the bank to roll over its short-term debt or maintain all its deposit funding. If market rates increase, doing so at previous low rates may be impossible3, 4.

SVB’s insolvency was evident from its Annual Report 2022. That report gives the accounting value of HTM securities as $91 billion while mentioning that fair value was only $76 billion. The difference of $15 billion far exceeded the tangible equity of the bank, which was given as $12 billion. If the losses had to be realised, the bank would have to default. The 8 March announcement showed that loss realisation had begun.

The Federal Reserve’s report on SVB (Federal Reserve 2023) refers to management recklessness and to the supervisors’ slowness in addressing this recklessness and raises many disturbing questions. The bank received high regulatory ratings; most of the issues supervisors raised were procedural and ignored the key issues that ultimately led to the bank’s failure.

The crisis provides an opportunity to implement long-delayed restructuring of the banking sector, with some banks leaving and some merging. It is important that solvent banks should be the ones to survive. And the too-big-to-fail problem must not be exacerbated

The preliminary review by the Government Accountability Office (GAO) notes that the Federal Reserve failed to engage in prompt corrective action as had been recommended in 2011 (GAO 2023). The supervisory failure was consistent with the ‘light touch’ approach to regulation and supervision at the Federal Reserve5.

The Fed report also treats interest rate and liquidity risks as though they had nothing to with each other. Changes in asset values and the changes in refinancing conditions were two sides of the same coin, however, both driven by the increase in interest rates in 2022.

The report also fails to discuss the bank’s solvency problem and the supervisors’ blindness towards this problem, which was masked by capital ratios based on accounting valuations and risk weights. For HTM assets, risk-based regulations are concerned only with credit risk and ignore the possibility of losses in fair value.

Supervisors should have recognised the declines in the market values of SVB’s assets as highly relevant to its viability and should have acted on this information.

SVB was special in having such extraordinary deposit growth in 2020 and 2021, catering to a very small socially connected clientele, having an extreme level of unrecognised losses on its assets, and having more than 90% uninsured deposits. These facts explain the extent and speed of the run, but the ultimate cause of the run was the underlying solvency problem.

This problem was not, and is not, unique to SVB or First Republic Bank, which failed for similar reasons6. Between early 2020 and March 2022, banks saw their deposits grow because money market investments were unattractive. Much of this growth went into fixed-income securities, which lost value when interest rates increased again. According to Jiang et al (2023), the total unrealised losses on securities in US banks amount to some $2 trillion7.

If one half of all uninsured deposits were withdrawn, about 190 banks would have to realise losses so large that they might be unable to repay insured deposits. If asset prices fell because many banks were selling simultaneously, the number of banks affected would be even larger.

The cases of SVB and First Republic have alerted markets to the systemic problem, as the Fed also recognised by invoking the ‘systemic risk exemption’ to justify its interventions.

The Fed’s policy reaction, which expanded its lending programmes to banks, neutralises the effects of deposit withdrawals temporarily, but does nothing to alleviate the banks’ solvency problems. If a bank pays off depositors by borrowing from the Fed rather than selling securities, its borrowing costs rise above the return on the securities.

Borrowing at 5% while earning less than 2% on government bonds bought in 2021 is a path to failure8. So is retaining deposits by raising rates paid to depositors if asset yields remain low.

Much of this looks like a replay of the Savings and Loans (S&L) crisis of the 1980s. When market rates of interest peaked in 1980-81, depositors moved their funds from S&Ls to money market funds. In response to the S&Ls’ complaints about the resulting ‘liquidity problems’, regulation of deposit rates was abolished so S&Ls could match money market rates. But these rates were much higher than the 6% they earned on fixed-rate mortgages made in 1965 that still had 15 years to go.

Under fair-value accounting, a large part of the S&Ls would have been insolvent in 1981 (Kane 1985). Then as now, however, there was no fair-value accounting for HTM assets. The ‘zombie’ S&Ls spent the 1980s gambling for resurrection, using funds from depositors attracted by high interest rates and by deposit insurance to taking risks that blew up when interest rates rose again in 1989. The ultimate cost to taxpayers was much higher than it would have been if insolvencies had been addressed in 1981.

Today, policymakers, lobbyists, and commentators seem to miss the obvious lesson: ignoring insolvencies while also insuring deposits can lead to disastrous outcomes. The Federal Reserve is now providing liquidity support without restoring solvency, prolonging the agony and encouraging some banks to start gambling for resurrection as the S&Ls did in the 1980s.

Expanding deposit insurance without eliminating zombies is similarly problematic. If bailouts of uninsured depositors continue, the extra levies on the banking industry to cover losses to the insurance fund may harm the viability of remaining banks.

US authorities should acknowledge the evident banking crisis and find appropriate ways to address the underlying solvency problems. The losses that have been incurred must be recognised and absorbed by appropriate parties.

The US Treasury might absorb losses by purchasing outstanding government debt at nominal values, but doing so would aggravate the government’s fiscal problems and is likely a political non-starter. Alternatively, the Federal Reserve might engage in such purchases, but this could raise monetary policy issues.

One immediate way to reduce solvency problems would be to restrict executive compensation as well as payouts to shareholders such as dividends and stock repurchases for banks whose equity capital fails to meet specified standards when fair-value accounting is applied.

This would prevent managers and shareholders, who had benefitted from the upside of risks, from continuing to pay themselves excessively while their bank exposes depositors and the deposit insurance system to significant risks of losses.

The crisis provides an opportunity to implement long-delayed restructuring of the banking sector, with some banks leaving and some merging. It is important that solvent banks should be the ones to survive. And the too-big-to-fail problem must not be exacerbated.

Mergers should involve small and medium banks rather than banks that are already too big themselves. The takeover of First Republic Bank by JPMorgan Chase in the US is a case in point. So is the shotgun acquisition of Credit Suisse by UBS that we discussed in a companion piece (Admati et al 2023).

Longer term, we need regulatory reforms to prevent a recurrence of the problems that underlay SVB’s failure. Extensions of deposit insurance that have been proposed by many observers are not a panacea and must be treated with caution if we want to avoid a repeat of the S&L debacle (Dewatripont et al 2023, Heider et al 2023, Perotti 2023).

Certain transactions such as meeting payrolls may require large transactions balances. The $3.3 billion that Circle Corporation held with SVB, however, had little to do with transactions and much with a convenient way of holding assets.

Corporate wealth management should not benefit from the levy on the banking industry that is needed to reimburse depositors of an insolvent bank. There is a case for limits to guaranteed deposits except for a grace period surrounding large transactions. The central bank might offer depository facilities for institutions that necessarily hold large balances.

The following reforms should have been introduced long ago:

  • Apply mark-to-market or fair-value accounting to all assets; abandon the fiction that changes in fair values of HTM assets are irrelevant.
  • Stop compartmentalising risks – recognise correlations among different risks that arise from common dependence on macro developments such as changes in interest rates affecting fair values of assets as well as the scope and conditions for rolling over short-term liabilities.
  • Strengthen supervision under Pillar 2 of the Basel Accord, which asks supervisors to consider the professionalism of bankers. The extent of maturity transformation at SVB was unprofessional. Supervisors should have interfered much more actively early on.
  • Raise equity requirements to enable banks to withstand fair-value losses on all assets when interest rates rise. If SVB had been subject to a 20% equity requirement, its losses would have been borne by shareholders rather than falling on the FDIC.

Narrow interests and intellectual misconceptions have blocked such reforms despite many banking crises. When will they ever learn?


1. Bank stock prices are far from recovering. The closures of SVB and First Republic have also caused substantial and persistent increases in banks’ deposit funding costs (see Slok 2023).

2. All numbers are quoted from the 10Ks and 10 Qs that SVB filed with the Securities and Exchange Commission.

3. The claim that ‘deposits are merely an accounting record’ and that banks ‘create’ money by making loans, and therefore that banks never have to worry about funding problems and only have liquidity problems, is fundamentally flawed and misleading. From the bank’s perspective, deposits are a form of debt, a legal liability that they must fulfil should the depositor want to withdraw funds, or else a third party such as the government fulfils, or the bank defaults and the depositor is not paid in full (Admati and Hellwig 2019, Claim 6).

4. The run on SVB also made it clear that deposits are no longer as ‘sticky’ as they were in the past and cannot be considered a long-duration liability that hedges long-duration assets.

5. See Wilmarth (2013) and “The Regulatory Breakdown Behind the Collapse of Silicon Valley Bank”, New Yorker, 19 March 2023, which includes an interview with Peter Conti-Brown.

6. Signature Bank in New York also failed on 12 March 2023, two days after SVB. FDIC (2023) traces this failure to the impact of the November 2022 failure of Silvergate Bank in California and to poor management and lack of response to supervisors.

7. Drechsler et al (2023) arrive at a similar figure.

8. At this writing, the Federal Reserve is charging at least 5% in its discount window (see https://www.frbdiscountwindow.org/). On 12 March, it opened a new Bank Term Funding Program that makes loans of up to a year, currently at 4.74% interest. Both programmes value bank collateral at par, which is often much higher than the standard fair market value. In addition, many US banks borrow from the Federal Home Loan Bank (FHLB) system, which generally lends with harsher terms. Both Silicon Valley Bank and First Republic Banks borrowed from the San Francisco FHLB, and many more are taking such ‘advances’ (eg. https://www.risk.net/risk-quantum/7956520/schwab-turns-to-costly-fhlb-advances-as-deposits-drop).


Admati, A, M Hellwig and R Portes (2023), “Credit Suisse: Too big to manage, too big to resolve, or simply too big?,” VoxEU.org, 8 May.

GAO – Government Accountability Office (2023), “Bank Regulation: Preliminary Review of Agency Actions Related to March 2023 Bank Failures”, April.

Dewatripont, M, P Praet and A Sapir (2023), “The Silicon Valley Bank collapse: Prudential regulation lessons for Europe and the world”, VoxEU.org, 20 March.

Drechsler, I, A Savov, P Schnabl, and O Wang (2023), “Banking on uninsured deposits”, NBER Working Paper 31138.

Heider, F, JP Krahnen, L Pelizzon, J Schlegel, and T Tröger (2023), “European lessons from Silicon Valley Bank resolution: A plea for a comprehensive demand deposit protection scheme,” SAFE Policy Letter No. 98.

FDIC – Federal Deposit Insurance Corporation (2023), “FDIC’s Supervision of Signature Bank”, 28 April.

Federal Reserve (2023), “Review of the Federal Reserve’s Supervision and Regulation of Silicon Valley Bank”, Federal Reserve Board of Governors, 28 April.

Jiang, E, G Matvos, T Pikorski, and A Seru (2023), “Monetary tightening and U.S. bank fragility in 2023: Mark-to-market losses and uninsured depositor runs”, SSRN, 10 April.

Kane, E (1985), The Gathering Crisis in Federal Deposit Insurance, MIT Press.

Perotti, E (2023), “Learning from Silicon Valley Bank’s uninsured deposit run”, VoxEU.org, 5 May.

Slok, T (2023), “Quantifying the negative effects of the banking crisis”, The Daily Spark, 14 May.

Wilmarth, A E (2013), “Turning a Blind Eye. Why Washington Keeps Giving in to Wall Street”, University of Cincinnati Law Review 81(4): 1283-1446.

Editors’ note: This column is part of the Vox debate on “Lessons from Recent Stress in the Financial System.” This article was first published on VoxEU.org.