Clear Recognition of the Discount Window Would Improve Liquidity Rules

On Jan. 18, 2024, the OCC’s Acting Comptroller Michael Hsu delivered remarks at Columbia Law school in which he reflected on the changing nature of liquidity risk demonstrated by the failures of Silicon Valley Bank and Signature Bank in spring 2023 as well as his views on appropriate regulatory responses. He drew three main lessons from the bank failures: 1) uninsured deposits can be withdrawn quickly and en masse; 2) having liquid assets is necessary but not sufficient for banks to endure liquidity stress, as operational capacity to monetize those assets quickly is also required; and 3) contagion to healthy banks (and therefore systemic threats) can occur via “guilt by association,” i.e., shared uncertainty with regards to seemingly similar risk profiles, such as high reliance on uninsured deposits for funding.

Hsu’s principal recommended regulatory response would be to…

…explicitly give banks credit for their discount window borrowing capacity to cover ultra-short term, acute outflows like those endured by SVB and Signature. This would make clear that regulators expect banks in stress to utilize the discount window to help cover short term liquidity outflows when needed. This regulatory expectation could help de-stigmatize discount window usage. pp. 5-6

In particular, he suggested that in addition to the existing liquidity regulations, banks also be required to have cash and discount window borrowing capacity sufficient to meet five days of stress outflows. 

Hsu’s remarks reflect what appears to be a growing consensus in regulatory circles toward recognizing discount window borrowing capacity in liquidity assessments. The preceding week, the Group of Thirty issued a report that made a similar recommendation, and earlier in the year the Federal Reserve and other banking agencies issued supervisory guidance encouraging banks to be ready and willing to use the discount window. Relatedly, Hsu recommended that banks incorporate the discount window into their plans for how they monetize their liquid assets.[1] 

These are welcome developments. Recognizing the important role of the discount window in liquidity risk management has several benefits. First, it makes liquidity assessments more accurate – as was illustrated by the bank failures last spring, a bank that is prepared to borrow from the discount window, including by having prepositioned collateral, is more liquid than one that is not.  Second, it provides banks with a strong incentive to be prepared to borrow. And third, it could help reduce the stigma associated with the discount window. Significantly reducing stigma is a precondition for the proposal to work – if banks won’t borrow, then collateral at the window does not improve their liquidity – and will require a comprehensive effort, which is also acknowledged in Hsu’s speech.[2]

Moreover, if the proposal is integrated into the broader set of U.S. liquidity risk requirements, it could support the Fed’s efforts to shrink its balance sheet. As the Fed allows its securities to roll off (aka QT), its liabilities, in particular reserve balances, also decline. Barring a recession, the only thing that should stop QT is if QT causes the supply of reserve balances to fall below aggregate demand for reserve balances by banks.  While the liquidity coverage ratio treats Treasuries and reserve balances as the same (i.e., both equally count as HQLA), supervisors have conveyed a strong preference that banks satisfy requirements with reserve balances, which are immediately available, versus Treasuries, which must be converted to cash through a process (e.g., via sale or repo).  But if banks are separately required to establish their ability to meet immediate cash needs in part through discount window borrowing capacity, there should be no need for that supervisory preference. Banks could instead hold Treasuries or engage in Treasury reverse repos to satisfy the LCR. There are other, specific design elements of the U.S. liquidity requirement framework that also should be reconsidered if the requirement Hsu suggested is added to the existing liquidity regulations.[3]

Hsu also notes, correctly, that “not all uninsured deposits are created equal,” and recognizes a need for an improved understanding of the most important characteristics of deposit accounts for determining their behavior under stress as well as each categories’ outflow realizations. Indeed, the OCC is sponsoring a conference in June to help expand knowledge in this area. He specifically mentions academics and government researchers as potential sources for the needed information. We hope the agencies also plan to seek input from bankers, who are the frontline responders to bank runs and therefore have the best information and data on the extent to which different categories of uninsured deposits are (or are not) susceptible to runs. Insofar as the banking agencies develop a more nuanced view of the flightiness of different types of deposits, that information should influence the design of any new short-term liquidity requirement.

Returning to the issue of stigma, there has been a stigma associated with using the discount window since soon after the founding of the Fed, and it intensified after the Global Financial Crisis when banks that used the discount window, despite being encouraged to do so, were treated as having received a bailout. Banks remain concerned that any use will later be pointed to by their examiner as an indication that their liquidity risk management was deficient. The Acting Comptroller’s remark that “[t]he line between being a lender of last resort and providing a bailout can be a fine one, especially in times of stress” is particularly unhelpful.  Instead, the banking agencies need to take a page from the Bank of England’s playbook and emphasize that any borrowing – from the discount window or the standing repo facility – is a business decision on the part of the bank and not an indication that something has gone wrong. A fully collateralized loan from a central bank is not a bailout, because losses can be covered in full by liquidation of the collateral if necessary. Of course, for a loan to be fully collateralized, the collateral must be valued at fair market prices and haircuts should be applied to the fair values to guard against subsequent declines in fair values after the loan is extended.

As we’ve written previously, liquidity regulation is important as it comes with high benefits and high costs, and it’s complex because there are no easy tradeoffs. Given the consideration of a brand-new liquidity requirement in response to the events of last spring, the many ways such a new requirement would interact with the existing parts of the regulatory liquidity framework and the acknowledged need for additional information on deposit behavior, the agencies should as a first step invite public input through an advance notice of proposed rulemaking or a request for information.


[1] It was unclear if he was using the term “monetize” generically or if he was also referring to the monetization requirements in internal liquidity stress tests.

[2] For a further discussion of stigma and suggestions for how to reduce it, see BPI notes “Discount Window Stigma: We Have Met the Enemy, and He Is Us,” and “A Major Limit on the Fed’s Crisis Toolkit: Shame.

[3] Under the LCR and internal liquidity stress test requirements, eligible liquid assets that are pledged to the discount window but are not currently backing an extension of credit are eligible to meet the requirements. The U.S. LCR requires eligible liquid assets to be unencumbered for inclusion and states (12 CFR 249.3)  [emphasis added]:

Encumbered means, with respect to an asset, that the asset:

(2) Is pledged, explicitly or implicitly, to secure or to provide credit enhancement to any transaction, not including when the asset is pledged to a central bank or a U.S. government-sponsored enterprise where:

(i) Potential credit secured by the asset is not currently extended to the Board-regulated institution or its consolidated subsidiaries; and

(ii) The pledged asset is not required to support access to the payment services of a central bank.

The assets are not encumbered unless the institution uses them to collateralize a discount window loan, in which case they would provide cash needed to address the projected liquidity need. Alternatively, book entry securities can easily and immediately be removed from the window and used to raise cash through repo or sales. We assume that, similarly and for the same reason, high-quality liquid assets pledged to the discount window and counted toward borrowing capacity under the potential new requirements would also be eligible to meet the LCR and count toward ILSTs.