As Prepared for Delivery
Introduction
Thank you for the invitation and opportunity to speak to you today.1 The topic of your research conference—depositor behavior, bank liquidity, and run risk -- is critically important for the business of banking and for financial stability.
The deposit runs we observed in the spring of 2023 were in many ways the most serious disruption to banks in more than a decade. The runs contributed to the failures of three relatively large banks in the U.S., stresses at other banks with similar business models, and the failure of a GSIB. More broadly, they were a stark reminder that depositor runs and contagion to other banks are not anachronistic events of the past, and what they exposed were not new, unknown vulnerabilities. Instead, the runs highlighted the buildup in some well-known vulnerabilities in the banking system associated with unrealized mark-to-market losses on the asset side of bank balance sheets, and a large share of uninsured deposits on the liability side. In addition, they focused attention on how technology may be speeding up deposit withdrawals, and the role of social media. Many of these topics are covered by the papers on the program for this conference.
Today, I will start by reviewing the events of March 2023, but then I will take a step back and frame these events within a broader view of banks’ business models over the past 40 years. I will end with some questions and considerations for potential policies around liquidity management aimed at protecting the safety and soundness of banks and financial stability.
Before I delve into these issues around liquidity, I’d like to make sure we don’t lose sight of the importance of capital for interest-rate risk in bank regulatory capital. Last year’s runs took place during an interest rate tightening cycle that led to significant unrealized mark-to-market losses on banks’ bond portfolios and threatened the solvency of some banks. In many ways, this episode is another reminder of the importance of capital and the interaction of capital and liquidity. I will also touch a bit on this later.
U.S. bank runs in March 2023
Turning to the bank runs of March 2023, I’d like to briefly remind us of the broad U.S. macroeconomic environment we were in. As the economy recovered from the pandemic and inflation reached high levels, interest rates increased significantly from very low levels, starting in late 2021 and through 2022, reflecting the monetary policy tightening by the Federal Reserve.
The tightening came on the heels of a period of rapid deposit growth during the pandemic. Total deposits at commercial banks in the U.S. rose by more than 35 percent over the two years from 2019 to 2021, to around $18 trillion. This rise in aggregate deposits was outsized compared to any period in recent history, as documented in a paper by Castro et al. (2022).2 Their paper noted four factors as largely explaining this significant deposit growth: the initial spike in commercial and industrial (C&I) credit line drawdowns at the onset of the pandemic; asset purchases by the Federal Reserve; large fiscal transfers to households held mostly in savings in the form of deposits; and a higher personal savings rate.3
Notably, as deposits grew, the share of uninsured deposits also grew, especially for regional banks. As shown in figure 1, the share of uninsured deposits for all domestic banks grew by a few percentage points to about 45 percent. But, for banks with assets between $100 and $250 billion, this share grew by 10 percentage points and reached 55 percent. This share also grew by 10 percentage points for smaller banks with assets between $10 and $100 billion, but to a lower share of roughly 45 percent.
As interest rates rose, bank deposits began to fall. The outflow of bank deposits into money market funds and other alternative short-term instruments was mostly anticipated and orderly. However, deposits at the large regional banks declined more significantly. Banks turned to Federal Home Loan Bank (FHLB) advances in many instances to make up for these lost deposits. These advances more than quadrupled from around $190 billion in late 2021 to over $800 billion in the first quarter of 2023.4
Moreover, some banks, including some regional banks, had invested in long-dated Treasury and agency securities or other long-duration, fixed-rate assets when their deposits had been growing, taking on increased duration risk. And as interest rates rose, these banks experienced sizable declines in the fair value of these securities and loans, leading to significant unrealized losses.5
A few firms had a combination of especially elevated shares of uninsured deposits and unrealized losses, as well as rapid growth in uninsured deposits, and the runs and failures that have now become history followed. You can see both patterns in figure 2—the left-hand side plots individual banks’ shares of uninsured deposits as of the fourth quarter of 2022 against their adjusted common equity tier 1 (CET1) capital ratios for the same period. This CET1 measure reflects the unrealized losses on banks’ books, as it incorporates the fair values of securities and loans on banks’ books. The right-hand side depicts individual banks’ shares of uninsured deposits in the fourth quarter of 2022 against the growth in uninsured deposits between 2019 and 2022. In both figures, the three banks that failed— Silicon Valley Bank (SVB), Signature Bank, and First Republic—really do stand out.
Perhaps what’s especially striking about these runs is that they were very large and fast by historical standards. For example, as documented in Rose (2023), SVB lost 25% of its deposits in one day and was closed before an additional 62% was scheduled to leave the next day.6 At Signature Bank, 20% of deposits were withdrawn in a matter of hours. And at First Republic, customers withdrew about 14% of deposits on the first day, 23% the next business day, and an additional 20% until it failed. In contrast, the failure of Washington Mutual in 2008, the largest bank failure ever in the U.S., was the culmination of stresses that occurred over several weeks, with total deposit outflows of 10%.
While technology that allows deposits to be transferred easier and faster likely contributed to the speed and size of these runs, and there was substantial social media coverage of SVB the day its run started, a key factor seems to have been the concentrated and highly networked depositor base of these banks. For example, at SVB, venture capital firms, portfolio companies, tech and crypto companies, and high-net-worth individuals appear to have communicated their concerns quickly with one another, leading to a massive and rapid run. There was also some uncertainty around if FHLBs would be able or willing to meet SVB’s and Signature’s requests for increased advances, which may have added to the concerns about the banks’ ability to meet redemption requests and made their liabilities “information sensitive.”7
The contagion that followed these runs was also a classic bank run, as other regional and mid-sized banks that were perceived to have similar weaknesses experienced outflows as well. Those banks that came under greater pressure tended to have large unrealized losses in their loan and securities portfolios and relied heavily on uninsured deposits.8 Importantly, during this episode, deposits at smaller banks generally remained stable, and the outflows became inflows into the largest banks.9
In response, the Federal Reserve, the FDIC, and the Treasury Department stepped in quickly by invoking the systemic risk exception, permitting the FDIC to fully pay out uninsured depositor claims for SVB and Signature, and by establishing the Bank Term Funding Program. These actions were helpful in limiting a broader contagion and further damage to the U.S. economy, even as some regional banks continued to see outflows and came under stress for a period.
Evolution of bank business models
While the events of March 2023 remind us of something we have known for a long time—the reality of bank runs—I want to place these events in a broader context, and in particular highlight some observations around how banking has evolved over the past 40 years. I think these broad trends are helpful for thinking about any policy changes that could be considered to reduce the fragility of the banking system, a point nicely articulated also by Hanson et al. (2024).10
As shown in figure 3, deposits have been growing rapidly relative to GDP over the past 40 years—rising on net by almost 60% between 1985 and 2019, before surging in 2020, for the reasons mentioned earlier. The uninsured share of deposits also has risen substantially during this period, with much of the increase happening before the Global Financial Crisis (GFC), but remaining high after the insured deposit limit was raised to $250,000. This longer-term upward trend for deposits suggests banks have significant value in their deposit franchise, continuing a significant role in providing liquidity services, facilitating transactions and payments.
At the same time, as shown in figure 4, direct C&I lending by banks has been almost flat to even a bit negative over the same 40-year period. Considering that nonfinancial credit to GDP increased significantly during this window, this figure underscores the shrinking role of banks in providing credit to nonfinancial businesses. Instead, as Hanson et al. (2024) highlight, banks seem to be shifting their asset portfolios towards securities holdings.
But on-balance sheet lending is not the full story. Notably, even as direct bank loans to businesses have fallen, banks still provide liquidity to nonfinancial businesses through revolving credit lines. We know that businesses drew significantly on these backup lines at the start of the pandemic, which can be seen in the chart and is also discussed in Bräuning and Ivashina (2024).11
In addition, banks provide credit and liquidity lines to nonbank financial intermediaries (NBFIs) to support their lending, including to nonfinancial businesses. Growth in lending by banks to NBFIs, including nonbank lenders, securitization vehicles, open-end funds, and other private funds, can be seen in figure 5. Bank loans and credit line commitments to NBFIs have reached more than $2 trillion (relative to C&I loans of $2.7 trillion), with commitments accounting currently for about 80% of the total shown here. Acharya et al. (2024) highlight this trend, emphasizing that banks via their provision of loans and credit lines to NBFIs remain exposed to credit and contingent liquidity risk.12 In other words, the growth in NBFIs vis- à-vis the drop in bank lending may not entirely be a zero-sum game for banks.
These indicators suggest important changes in how banks provide credit and liquidity to businesses and support financial activity, yet highlight how banks remain a vital part of financial intermediation despite the declining measures of their direct loans to businesses. You see this through both the continued rise in deposits over decades and the rise in credit and liquidity to NBFIs. In other words, even as some things have changed, core bank functions have not. Banks continue to play a key role in liquidity provision through both the asset and the liability sides of their portfolios, as formulated in Kashyap, Rajan, and Stein (2002), with strong synergies between the two, where the deposit franchise helps them provide the liquidity on the asset side.13
Policy considerations
A key lesson of the events of March 2023 is that depositor runs and contagion to other banks are still part of our lives, as they were in It’s a Wonderful Life, the 1946 movie. Yes, the speed and size of the recent runs were larger than in recent decades, and technology and social media may have played a role. But the vulnerabilities were not new, and bank deposit runs are not sunspots. At the same time, from a broader perspective, banks’ value is increasingly based on their ability to provide liquidity, on both the liability and the asset side. As we consider what policy changes might be appropriate, it seems critical to focus on how we can ensure that banks are able to provide that liquidity both in normal and in stress times.
Let me mention six policy areas for consideration. First, we need to ensure that our supervisory and regulatory frameworks can effectively monitor and address core vulnerabilities that were the key drivers of the recent runs. These vulnerabilities include increased shares of uninsured and concentrated deposits, as well as unrealized losses on loan and securities portfolios. In that regard, more complete coverage of interest rate risk in capital requirements seems necessary. Bank regulatory capital measures do not reflect unrealized changes in the market value of banks’ securities holdings, such as due to a rise in interest rates, except for the largest banks. This is due to “hold-to-maturity” accounting and the “AOCI opt-out” for securities that are accounted for on an “available-for-sale” basis. The banking agencies have proposed to eliminate this opt-out for banks with assets between $100 billion and $700 billion. Finalizing this change would be an important step in addressing interest rate risk. As a side but related note for researchers: it would be helpful to understand why uninsured deposits and long-dated securities grew disproportionately at these regional and mid-sized banks, and particularly if differences in regulatory requirements were a factor.
Second, banks should have the operational capacity to borrow from the discount window and to periodically test this capacity. It’s really the only true backstop for banks to quickly raise or provide assurance of needed liquidity, short of holding cash and reserves. This access is key for banks’ ability to fulfill their liquidity-provision role, both to meet the demand for deposits and committed lines, which become heavily tapped in stress times.
At the same time, the work by the Federal Reserve to improve the discount window’s operational efficacy, such as moving to online systems, or enabling a smoother transfer of collateral between FHLBs and the discount window, is needed. Of course, there is still a stigma associated with the discount window, perhaps reflecting its long history. But efforts to require regular testing, building discount window readiness into liquidity regulations and supervision, and re-thinking the two-year ex-post disclosure requirements seem like additional useful steps that can help reduce stigma.
Banks also borrow from the FHLBs, at times in great volume, as I mentioned earlier, but FHLBs are not well suited to be lenders of last resort. This point was articulated clearly in the FHFA report released late last year.14 But this practice raises a question of when does a FHLB decide to stop meeting a bank’s demand for advances. As part of the work to reform FHLB practices and bank liquidity risk management practices, I think it will be helpful to look for ways for the FHLBs to be more transparent about their practices. It is also critical that FHLB actions to stop advances are not viewed to contain private information about a borrower’s financial condition and to make a bank’s liabilities “information sensitive.” One potential idea in this regard would be to consider ex-ante concentration and growth limits for bank liabilities that would apply to FHLB advances, similar to how it may be appropriate to keep an eye on concentration of deposits.
Fifth, the proposals for pre-positioned collateral requirements at the discount window are promising, but there are important questions to consider. These include how much and what kind of collateral should be required. For example, the G30 report proposes requiring banks to pre-position enough collateral at the central bank to meet all their “runnable” obligations, including 100% of uninsured deposits.15 Similarly, Hanson et al. (2024) proposes that the pre-positioned collateral should be largely short-term government debt, to lean against the use of long-duration securities as backing for uninsured deposits, but they also give a nod to the possibility of including loans with appropriate haircuts. Another question is whether the pre-positioned collateral should count towards existing liquidity requirements like the Liquidity Coverage Ratio. As alternatives are considered, an important principle is that pre-positioning requirements are not so significant that they undermine banks’ deposit franchise value, which is tied to their ability to provide liquidity to nonfinancial corporations and NBFIs.
Finally, re-examining deposit insurance coverage could be considered, but additional research is needed here too. The FDIC issued a report following the bank failures last year on options for deposit insurance reform, including wider coverage or higher limits. One of the options in the FDIC report is to expand coverage to business transactions accounts, similar to what is done in some other countries.16 But the FDIC also highlighted unanswered questions, including how to define such deposits. Another question is whether an expansion of deposit insurance might produce significant costs in the form of reduced market discipline, increased moral hazard, or other costs.
To conclude, there are many policy responses to consider in response to the events in March 2023, and research like what is being presented at this conference is essential to deepening our understanding of these issues and questions. I look forward to reading and learning from them.
Thank you.
The figures referenced in these remarks are available here.
###
[1] I would like to thank Burcu Duygan-Bump, Eric Goldberg, and Laurie Schaffer for assistance in preparing these remarks.
[2] Castro, Andrew, Michele Cavallo, and Rebecca Zarutskie, “Understanding Bank Deposit Growth during the COVID-19 Pandemic,” FEDS Notes, Washington: Board of Governors of the Federal Reserve System, June 06, 2022. https://doi.org/10.17016/2380-7172.3133.
[3] In addition, some banks—like Silicon Valley Bank—experienced extraordinary growth in deposits related to the growth of the venture capital and technology sectors during the tech boom around this period. See Silicon Valley Bank profit squeeze in tech downturn attracts short sellers (ft.com).
[4] Federal Deposit Insurance Corporation, Balance Sheet: Total Liabilities and Capital: FHLB Advances [QBPBSTLKFHLB], retrieved from FRED, Federal Reserve Bank of St. Louis; https://fred.stlouisfed.org/series/QBPBSTLKFHLB.
[5] Federal Reserve Financial Stability Report, May 2023.
[6] Jonathan Rose, "Understanding the Speed and Size of Bank Runs in Historical Comparison," Economic Synopses, No. 12, 2023. https://doi.org/10.20955/es.2023.12.
[7] Dang, Tri Vi and Gorton, Gary B. and Holmström, Bengt R., “The Information View of Financial Crises,” Annual Review of Financial Economics, Vol. 12, November 2020. http://dx.doi.org/10.1146/annurev-financial-110118-123041.
[8] This experience is similar to runs on asset-backed commercial paper programs in 2007, which were also not random but instead were significantly more likely at riskier programs, based on observable program characteristics, including the quality of back-up liquidity. See Covitz, Daniel M., Nellie Liang, and Gustavo Suarez, “The Evolution of a Financial Crisis: Collapse of the Asset-Backed Commercial Paper Market,” Journal of Finance, Volume 68, Issue 3, June 2013. http://dx.doi.org/10.2139/ssrn.1364576.
[9] See for example Luck, Stephan, Matthew Plosser, and Josh Younger, “Bank Funding during the Current Monetary Policy Tightening Cycle,” Federal Reserve Bank of New York Liberty Street Economics, May 11, 2023, and Cipriani, Marco, Thomas M. Eisenbach, and Anna Kovner, “Tracing Bank Runs in Real Time.” Federal Reserve Bank of New York Staff Reports, no. 1104, May 2024. https://doi.org/10.59576/sr.1104.
[10] Hanson, Sam, Victoria Ivashina, Laura Nicolae, Jeremy Stein, Adi Sunderam, and Dan Tarullo, “The Evolution of Banking in the 21st Century: Evidence and Regulatory Implications,” Brookings Papers on Economic Activity, Spring 2024.
[11] Bräuning, Falk and Ivashina, Victoria, “Bank Runs and Interest Rates: A Revolving Lines Perspective,” May 14, 2024. http://dx.doi.org/10.2139/ssrn.4827005.
[12] Acharya, Viral V. and Acharya, Viral V. and Cetorelli, Nicola and Tuckman, Bruce, “Where Do Banks End and NBFIs Begin?” March 15, 2024. http://dx.doi.org/10.2139/ssrn.4760963.
[13] Kashyap, Anil K, Raghuram Rajan, and Jeremy C Stein, “Banks as Liquidity Providers: An Explanation for the Co-Existence of Lending and Deposit-Taking,” Journal of Finance, Volume 57, Issue 1, 2002.
[14] See Federal Home Loan Bank (FHLBank) System at 100: Focusing on the Future, 2023.
[15] G30 Report on “Bank Failures and Contagion: Lender Of Last Resort, Liquidity, And Risk Management,” January 2024.