“Reflections on Bank Regulatory and Resolution Issues”

Introduction 

The past year and a half has been a pivotal time for the FDIC.  With three of the four largest bank failures in U.S. history, longstanding questions have reemerged about how to address bank runs, and newer questions have emerged regarding how FDIC receiverships are funded.  Today, I will offer thoughts on both of these topics, along with views on the FDIC’s approach to brokered deposits and a few thoughts on the path forward for the Basel III Endgame proposal.

Bank Liquidity and the Discount Window 

The bank failures last spring sparked a conversation about the Federal Reserve’s discount window and bank liquidity rules, among many other topics.  Many wondered how Silicon Valley Bank (SVB), with a huge stockpile of government bonds, was largely unable to borrow from the discount window before its failure.¹ Subsequent reports from the Federal Reserve and FDIC on SVB and Signature Bank, respectively, described how both institutions were unprepared to do so.²

The Federal Reserve was originally created to mobilize reserves, provide liquidity to banks, and reduce banking panics,³ with varying success over time.⁴ One prerequisite for this objective to succeed is that banks must be able and willing to use the liquidity facilities available.  The lack of use is a multifaceted problem, but the less banks use the discount window in aggregate, the more stigmatizing it is when banks do use it, and the more reluctant supervisors are to allow banks to rely on it as part of their contingency planning.  

As an initial step to address the issue, following stories of SVB’s frantic, last-minute efforts to borrow from the discount window,⁵ the Federal Reserve has reportedly been working on operational improvements to modernize the discount window’s antiquated⁶ operations, such as by allowing banks to make funding requests electronically rather than by phone⁷ and potentially expanding the hours of operation.⁸ These types of operational improvements seem like a necessary but insufficient step.   

Another idea that has been under consideration is a discount window prepositioning rule.  This concept is a derivation of Mervyn King’s “Pawnbroker For All Seasons” proposal, which would require a bank’s cash and central bank borrowing capacity to exceed all runnable liabilities.⁹ The objective is similar to that of deposit insurance, to remove the risk of runs by removing the risk of loss, but broader (in the sense that it would apply to all short-term liabilities, rather than just a subset of deposits) and leakier (in the sense that it would not completely remove the risk of loss¹⁰).

More recently, in 2023, a number of institutions experiencing stress or negative attention repositioned their balance sheets so that total cash plus borrowing capacity at the Federal Reserve and Federal Home Loan Banks exceeded total uninsured deposits.  This was generally a helpful approach to stem deposit outflows, as it removed the “first mover advantage” that can spark a bank run, and helped restore confidence among potentially skittish depositors. However, maintaining such a ratio, while doable in the short term for some banks, would be extremely costly to maintain over the long term for most banks, and impossible to achieve even in the short term for some banks.  Moreover, moving the entire banking system to such a 100% ratio would radically reshape the role banks play in the economy. 

The proposal currently being discussed is a derivation of this concept.  Presumably in recognition that this type of 100% ratio is unfeasible, the new proposal under consideration would require banks to maintain a minimum ratio of cash plus discount window borrowing capacity to uninsured deposits set at some value below 100%,¹¹ with news reports suggesting that ratio could be set around 40%.¹²

While I think establishing some incentives or requirements for banks to establish access to, preposition collateral at, and borrow from the discount window is worth considering, setting this type of hardwired ratio seems dangerous.  Whereas a ratio above 100% may remove the first mover advantage (at least temporarily), a ratio under 100% amplifies it, by shining a magnifying glass on the fact that a bank cannot cover every depositor.  If SVB’s management said to its uninsured depositors on March 8, 2023, “don’t worry, we have 40% coverage of our deposits,” I suspect the reaction from depositors would have been something like, “we better make sure we are part of the 40%, and not part of the 60%!”

Part of the motivation cited for the proposal is destigmatizing the discount window.¹³ The Federal Reserve has been working to address stigma for many years,¹⁴ an issue that has existed for nearly its entire history.¹⁵ I tend to think that, at least for large banks, stigma will remain an issue so long as banks do not borrow from the discount window in meaningful volumes and markets can figure out when banks under stress are forced to borrow.¹⁶ A fundamental challenge is encouraging banks to use the discount window, so that it is less damaging to use when really needed, but not to make it so attractive that the central bank becomes an everyday source of liquidity across the industry. In that vein, while it is helpful for regulators to do more to consistently convey that borrowing from the discount window is acceptable, what message does it send to establish a rule requiring prepositioning, but not incorporate discount window capacity in the Liquidity Coverage Ratio (LCR)¹⁷ and Internal Liquidity Stress Tests (ILST)¹⁸?

I think it makes more sense to consider incorporating discount window capacity into the LCR, rather than creating a new liquidity metric for each stress event.¹⁹ If a bank run unfolds at supersonic speed, as occurred at SVB, the discount window is likely to be the only game in town, as only the central bank can create liquidity at scale immediately.  Even if the run at SVB had occurred over the course of days rather than hours, rapidly liquidating its huge portfolio of securities to meet outflows would have been impractical and extremely disruptive to markets – as the FDIC learned firsthand when we assumed and later sold off that same securities portfolio.²⁰

And even First Republic, which suffered a run that lasted several weeks, primarily²¹ met its deposit outflows with Federal Reserve borrowings,²² and was finally forced into receivership when it dropped to secondary credit²³ and lacked the borrowing capacity to continue to meet outflows.  Furthermore, selling large stockpiles of a security craters the value of that security and harms any other banks who own those securities.

Allowing banks to count loans that are monetizable through the discount window in the LCR could also reduce the trend of forcing large banks to buy increasingly large quantities of securities, and operate increasingly like bond funds,²⁴ particularly if the agencies are also planning on ratcheting up HQLA demands through other LCR changes.²⁵ From a stigma perspective, this would achieve a similar goal of incentivizing banks to preposition more collateral at the discount window, while delivering a more consistent message regarding supervisors’ expectations of discount window use under stress.

In addition to improvements to the discount window, we should continue to think about other ways that regulators can better manage future liquidity stresses.  One example is better access to real-time data. I have spoken in the past about the FDIC’s rapid phase prototype competition, which was initiated in 2020 and discontinued by the current Chairman, to develop technologies to provide more granular and frequent data, which could eventually have given the FDIC access to real-time deposit flow data.²⁶ Another option, on potentially a much shorter time horizon, arises from a recent paper by researchers at the Federal Reserve Bank of New York.  The paper analyzed which banks suffered runs in March 2023 by looking at interbank payments through Fedwire and ACH.²⁷ The weekend after SVB failed, I asked officials at the Federal Reserve whether this type of data was available to see deposit movements, and I was told it was not.²⁸ It would be very helpful if the Federal Reserve can find a way to make this data available in real time for key decision-makers, along with capabilities to rapidly analyze it.  Among other benefits, better visibility into real-time deposit movements across the system could significantly improve our ability to monitor for runs and contagion, and reduce the risk of unnecessarily invoking the system risk exception (and vice versa).

Receivership Funding 

Much like the bank failures last year raised questions about banks’ contingency funding plans (or lack thereof), so too did the failures raise similar questions about the FDIC’s.  Bank failures can occur suddenly and involve significant short-term liquidity demands, particularly in the event of large bank failures, or a large number of bank failures.  The FDIC faced such a situation last spring, when the failures of SVB and Signature (and subsequently, but to a lesser extent, First Republic) placed substantial liquidity demands on the FDIC.  The FDIC initially met these demands through borrowings from the Federal Reserve,²⁹ and did not pay off the borrowings in full until nearly nine months later.  The unprecedented nature of these borrowings, and the substantial cost incurred, have raised a number of questions.

An initial question is when and to what extent the FDIC should have drawn from the Deposit Insurance Fund (DIF), rather than the Federal Reserve, to fund these receiverships.  The FDIC invests the DIF’s funds in nonmarketable Treasury securities,³⁰ which earn a return for the DIF but cannot be traded in the market.  In order to access cash from the DIF, the FDIC must redeem securities³¹ with the Treasury Department in exchange for cash.³² The average yield on the DIF’s portfolio last March was slightly over 3 percent,³³ meaning borrowing at market rates (much less at a penalty rate) was a much costlier option than spending DIF funds.³⁴

On March 10, 2023, the day SVB failed, cash in the Treasury General Account (TGA) amounted to $208 billion,³⁵ several hundred billion dollars lower than usual due to the fact that the outstanding debt of the United States reached its statutory limit on January 19, 2023.³⁶ The FDIC was able to access some funds from the DIF, but there were significant challenges and frictions in doing so because of sensitivities related to the debt ceiling and Treasury’s cash position.  The FDIC ultimately withdrew approximately $52 billion³⁷ from the DIF in multiple installments between March and May, amounting to approximately 40 percent of the funds in the DIF when SVB failed. 

I will offer three observations on this point. First, it seems the FDIC could have accessed more cash than Treasury provided, given the level of the TGA in March, and perhaps the FDIC should have insisted on it.  Second, in the event Treasury could not have provided more cash (or, in a future event, the level of the TGA is smaller³⁸), the FDIC and Treasury could have devised a plan for Treasury to, at one or more upcoming auctions, (1) redeem the FDIC’s securities for cash and, on the same day, (2) issue new securities into the market.  As others have pointed out,³⁹ redeeming securities from the DIF lowers the national debt and creates more space under the debt limit for the Treasury Department to issue new securities.  Under this scenario, the FDIC would need to borrow from the Federal Reserve (or another outside source) for a few days until the cash proceeds from the securities issuance(s) arrived.

Third, and most fundamentally, the FDIC should have unfettered access to funds in the DIF. Perhaps, rather than rely on the Treasury Department’s ability to redeem the DIF’s securities for cash, another option might be for the FDIC to leave funds in a deposit account at a Federal Reserve Bank.⁴⁰ This would ensure the FDIC could access and deploy DIF liquidity on very short notice, including after business hours and over a weekend.  While many government agencies spend much larger sums than the FDIC, I am unaware of any that has the same potential for very large, unexpected liquidity needs that must be met immediately, with the one possible exception the Federal Reserve itself, which serves as the lender of last resort precisely because it can create its own liquidity.   

A related question is how far the FDIC should deplete the DIF’s liquidity to meet short-term receivership needs.  Even after the debt ceiling was resolved, the FDIC continued to pay the Federal Reserve’s penalty rate for six more months,⁴¹ before finally using the DIF to fully pay off the Federal Reserve at the end of November 2023.⁴² In the immediate aftermath of the March 2023 failures, I think it was understandable to be cautious, but the justification for caution decreased significantly over the ensuing weeks, and was mostly gone by the summer.  More generally, I suspect we may overestimate the impact that the DIF’s temporary liquidity position has on depositor confidence, and if the FDIC is reluctant to use funds in the DIF for bank failures, what would this mean for our entire assessment framework?⁴³

Even if the FDIC had been able and willing to use all the funds in the DIF, the FDIC still would have needed to access additional liquidity.  (I should underscore that this was solely a liquidity issue, as the FDIC’s assets also included all the loans, securities, and other assets assumed from the failed banks that we would subsequently sell or monetize to pay off borrowings.)  The most obvious places the FDIC could turn are the Federal Financing Bank⁴⁴ (FFB) and the Treasury Department,⁴⁵ which would have charged a cheaper rate than the Federal Reserve’s penalty rate.  Before the debt ceiling limit was raised, the Treasury Department could not have issued new debt to finance any FDIC borrowings from the FFB or Treasury – but, as noted above, the Treasury Department still had considerable cash in March, which, in addition to being used to redeem the DIF’s securities, also could have been used to lend to the FDIC.  Nonetheless, the FDIC chose to continue paying the Federal Reserve’s penalty rate for months even after the debt ceiling was raised, rather than borrow from the FFB or Treasury.

The decision of whether to borrow from the FFB or Treasury is one the FDIC has faced before.  From 1991 to 1993, the FDIC borrowed from the FFB,⁴⁶ without any adverse market reaction, while in 2009, the FDIC chose not to borrow from the FFB or Treasury, due to fears around the optics of borrowing from Treasury.⁴⁷ In 2023, the FDIC made the same choice for similar reasons as 2009.⁴⁸ Of course, every circumstance is unique, but I suspect we overestimated the negative consequences of borrowing from the FFB or Treasury in 2023, particularly after the debt ceiling was resolved in June,⁴⁹ and especially given the DIF’s net worth was nowhere near zero.⁵⁰

A related option is to securitize assets through the FFB, a process in which the FDIC, in effect, sells the right to future cash flows (with an FDIC guarantee) to the FFB in exchange for cash, while retaining ownership of the assets.  This allows the FDIC to monetize the assets faster than selling them into the market, particularly for illiquid assets, and avoids the risk of disrupting markets. The FDIC securitized approximately $50 billion in assets through the FFB in September 2023 and an additional $47 billion in January 2024.⁵¹ This could be a useful option in the future – if the FDIC and FFB can execute such transactions much more quickly, ideally over resolution weekend, or if not within a few days of failure. 

Ultimately, the Federal Reserve funded the receiverships. More specifically, the two bridge banks borrowed extensively from the Federal Reserve to satisfy deposit outflows, with those borrowings guaranteed by the FDIC, while First Republic, as mentioned above, borrowed extensively from the Federal Reserve prior to failure. The three receiverships assumed those borrowings, which, combined, amounted to $273 billion in total,⁵² which the FDIC ultimately paid off in November 2023. 

When the bridge banks were dissolved, and when First Republic was sold, the Federal Reserve viewed the advances as in default, which, under the Federal Reserve’s Operating Circular, results in a 500 basis point increase in the interest rate charged.⁵³ After some negotiation, the Federal Reserve reduced the penalty rate to 100 basis points. Yet, the FDIC fully guaranteed the borrowings, and, as banks are required by law to tell their customers,⁵⁴ the FDIC is backed by the full faith and credit of the U.S. government.⁵⁵ The 100 basis point penalty rate was extremely costly given the volumes involved and the time it took to pay it off, ultimately costing the FDIC roughly $1 billion.⁵⁶ In effect, this was a shadow tax on the banking industry, primarily charged through the special assessment, that in the short run decreases the Federal Reserve’s losses and in the long run increases the “earnings” the Federal Reserve sends to Treasury.

The FDIC could have considered additional options, such as borrowing from the private sector or from the Federal Home Loan Banks, neither of which it appears the FDIC seriously considered or had ever prepared to do.  It is clear there are a number of lessons to be learned from this experience, perhaps chiefly the importance of doing much more advance work for this type of scenario. Given that this was not the first, and is unlikely to be the last, time the Treasury Department has dealt with debt ceiling issues, and given the possibility we could have much larger banks fail in the future, we should have a plan in place to immediately access DIF funds when the debt ceiling is in play, and we should have other backup options in place.  The FDIC has an obligation to minimize losses when resolving failed banks, and it is hard to argue we did that here. 

Brokered Deposits

Next, I will offer a few thoughts on brokered deposits.  The brokered deposits framework was first put in place by Congress in 1989.⁵⁷ 35 years later, I believe that the regime is no longer fit for purpose.

Brokered deposits are deposits that are placed with a bank through, or with the involvement of, certain types of intermediaries (“deposit brokers”).⁵⁸ The principal consequences of a deposit being classified as brokered are (1) a “tax” in the form of higher deposit insurance assessments⁵⁹ and more punitive treatment under liquidity rules, (2) limitations on the ability of a bank to accept such a deposit if the bank falls below well-capitalized,⁶⁰ and (3) in some cases, a reflexive judgment by examiners that the deposit is “risky.” The first is provided by rule, the second required by statute, and the third varies by examiner and agency.

The regime suffers from a number of flaws.  First, whether a deposit comes to a bank through an intermediary is not the key test in determining what type of risks it presents.  For example, SVB suffered a deposit run in part because it relied heavily on large, concentrated, closely networked, uninsured depositors.⁶¹ Whether any such deposits came in through intermediaries was not a relevant factor. In fact, when SVB failed, not a single one of its deposit was reported as brokered.⁶²

Second, the deposit landscape now encompasses a broad range of deposit arrangements, many of which involve third parties that bear little resemblance to traditional deposit brokers.  In the 1980s, the term “brokered deposit” was commonly understood to refer to a specific market in which third-party brokers gathered and pooled funds and placed them into “brokered” certificates of deposit (CDs) issued by banks.⁶³ At the time, Congress could use a broad definition of “deposit broker,” and could broadly exclude those whose primary purpose was not the placement of deposits,⁶⁴ because “deposit brokers” represented a well-understood universe, and few others were “engaged in the business of placing and facilitating the placement of deposits” at the time.  Most deposits came into a bank when customers walked into a local branch. However, since then, the internet, smartphones, and other innovations have revolutionized how banks interact with customers, with many types of deposit arrangements now involving intermediaries.

The emergence of new deposit arrangements over the years has continually put pressure on the brokered deposit framework. Prior to the 2020 rule, the FDIC addressed each new arrangement on a one-off basis, resulting in an unwieldy, opaque regime inconsistently applied across the industry.  The 2020 rule brought structure, process, and order to a disorderly regime, but a fundamental problem still exists: it is incredibly difficult to make sweeping judgments on which side of the line every new arrangement falls in a fair and risk-sensitive way. 

Today, different types of deposits currently classified as “brokered” present, in certain respects, diametrically opposite characteristics.  Some types of brokered deposits raise safety and soundness concerns out of fears that they are volatile, unstable, or otherwise likely to be withdrawn if the bank experiences turmoil or negative attention.  On the other hand, traditional brokered deposits present the opposite concern – the deposits have no franchise value because the depositors have no relationship with the bank, earn high rates, are fully insured, generally cannot withdraw before maturity, and thus are generally indifferent to the condition of the bank, but are incredibly stable.⁶⁵

Meanwhile, as discussed above, regulators have been trying to encourage more discount window borrowing for banks experiencing funding stress, but yet the FDIC may push further to prohibit the use of brokered deposits for essentially the same purpose.  From a contingency funding perspective, each has tradeoffs: the discount window can provide funds faster and at unlimited volume, but suffers from greater stigma.  Yet both are generally high-rate liabilities for the bank, and are more or less equally costly to the DIF if the bank subsequently fails – in either case, the funding generally has no franchise value and gets paid off in full upon failure.⁶⁶

Additionally, deposits gathered through a listing service website have never been treated as brokered, meaning a bank that falls below well-capitalized has no regulatory restrictions on its ability to obtain deposits through such a website by offering the highest rate – essentially the problem the brokered deposits law was initially intended to address.  Banks are sometimes encouraged by supervisors to seek out listing service deposits once they become subject to brokered deposits restrictions, which I think demonstrates that this whole regime does not work.

That all said, we should not view the deposit franchise of a bank that funds itself entirely with traditional brokered deposits the same as that of a bank that funds itself entirely with, for example, retail deposits from a diverse branch network.  But I do think we should make an effort to modernize how we think of a bank’s funding mix, and to think more holistically about how we view a bank’s deposit franchise.  We could consider doing a new study on deposits similar to the FDIC’s 2011 study,⁶⁷ though hopefully with a more open mind about brokered deposits.⁶⁸ And I continue to be skeptical⁶⁹ of sweeping policies restricting a bank’s access to certain sources of liquidity as its condition deteriorates.  I am generally supportive of the proposal suggested by Chairman McWilliams in 2019 to replace the brokered deposits restrictions with a growth cap, which would ensure banks in poor condition could not attempt to grow out of their problems,⁷⁰ but would allow them to meet deposit outflows in an orderly way.⁷¹

I also think it would be a mistake for the FDIC to substantively reopen the brokered deposits rule.  Doubling down on the pre-2020 brokered deposits regime in 2024 is like doubling down on stone castles after the invention of cannons.⁷²

Capital 

Finally, I would like to briefly address the Basel III Endgame proposal.  It’s clear that the proposal – issued almost exactly a year ago – lacked appreciation for its real-world impacts,⁷³ and I appreciate the widespread acknowledgment that “broad and material changes” are necessary. 

I agree with calls from many others that a re-proposal is necessary.⁷⁴ Given the nature, complexity, and extent of changes needed, we should allow the public an opportunity to provide feedback on what the agencies are considering.  In particular, the proposed operational risk and market risk frameworks rest on a complicated set of formulas and metrics that would be challenging to amend in a broad, material, and rational way without receiving additional feedback.  Addressing the services component of the operational risk framework, for instance, could have a profound impact on various business lines at banks, depending on how the agencies choose to modify the formula, and seeking additional comments would help ensure the agencies understand the consequences of any such modifications. 

I am skeptical of re-proposing only parts of the rule.  While some changes may be relatively straightforward and, in isolation, may benefit less from additional feedback, everything in the framework is related.  For example, commenters might react differently to certain changes to the operational risk framework if they know the agencies plan to adopt other changes to the credit risk framework (and vice versa), given that both frameworks impact the capital charge associated with the same exposures.  Additionally, I strongly believe any re-proposal should be issued jointly by all three banking agencies.⁷⁵ For just one agency to re-propose – but with an expectation that a future final rule will be issued jointly by the three agencies – would be unprecedented, sow confusion, and lead to a number of practical and legal questions.⁷⁶

In addition to broad, material, rational changes to the Basel III Endgame piece of the proposal, I also think any re-proposal should fully reverse the proposed reversal of tailoring of the capital framework for large banks.  At a minimum, the agencies should not reverse the 2019 capital simplifications rule,⁷⁷ and Category IV banks should not be subject to the new market risk framework (unless they have material trading activities), SA-CCR, or the proposed expanded risk-based approach.  The extreme volatility of a two-stack approach, in which a bank might ricochet back and forth from one standardized framework to the other, does not make sense for Category IV banks,⁷⁸ especially given that the expected capital impact is quite small. 

Conclusion

Thank you for your time today.  I am happy to answer questions.