To read the second post in this series, click here.
Liquidity regulations are intended to reduce the probability that a bank will be unable to meet demands on its funds — deposit outflows, margin calls and draws on lines of credit. The U.S. banking agencies are currently considering whether those regulations should incorporate in their assessment a bank’s ability to borrow from the Fed’s discount window against prepositioned collateral.[1] Similar discussions are occurring at other central banks around the world. While U.S. banks are subject to several different liquidity requirements imposed through the examination or resolution planning processes, the discussion has focused on the treatment of discount window borrowing capacity in the liquidity coverage ratio (LCR), the Basel liquidity regulation.
One concern that is often raised about such recognition is that banks’ ability to borrow from the discount window is not completely certain — for example, in the United States, the lending decision is ultimately up to the individual Reserve Bank making the loan. This note takes that concern seriously and tests it against the facts. It first reviews what the LCR measures and explains why the discount window debate has centered on whether borrowing capacity should count as a “high-quality liquid asset” (HQLA) rather than as a cash inflow — a framing that turns out to be more historical accident than considered design. It then asks how certain the discount window actually is, drawing on the Fed’s own decades of statements and actions. It then asks the same question of every other source of funds that the LCR already recognizes: monetized HQLA, cash inflows, reserve balances and committed liquidity facilities from central banks; the answer, in each case, is that certainty is high but not complete — no different from the discount window itself.
That comparison points to a broader conclusion. Applying a “complete certainty” standard uniquely to borrowing from the central bank reflects a fundamental disconnect between central bankers, who want banks to use central bank business-as-usual lending facilities freely, and examiners, who don’t understand the role of such facilities and prefer that banks not borrow from them. Such mixed messages about using business-as-usual central bank facilities are a core source of the stigma associated with borrowing that currently hobbles the discount window in the United States. They are also ironic given the long-emphasized importance of housing both central banking and bank examination at the Federal Reserve.
The note closes with several concrete ways liquidity regulation could recognize discount window capacity.
What the LCR Measures
The liquidity coverage ratio was included in the post-crisis Basel III reforms as a first attempt to adopt a quantitative liquidity rule. The LCR is defined as a ratio of the bank’s high-quality liquid assets to estimated net cash outflow (outflow minus inflow) over 30 days in a combined idiosyncratic and market-wide stress. Each bank is required to maintain an LCR of at least 1.
Strictly speaking, the numerator is also a cash inflow — it is the cash inflow that can be expected from the sale, repo or withdrawal of the bank’s HQLA.[2] In the United States, the HQLA banks maintain to meet the LCR consists almost entirely of reserve balances (deposits at a Federal Reserve Bank), Treasury securities and agency-guaranteed mortgage-backed securities. Agency MBS are subject to a 15 percent haircut on their value and cannot exceed 40 percent of an institution’s overall HQLA.[3] No cap or haircut is applied to reserves or Treasury securities.
Net cash outflow, the denominator of the LCR, is calculated as a measure of cash outflow under stress minus cash inflow (other than those from the liquidation of HQLA) under stress. These measures are not projections — they are calculated by applying fixed coefficients established in the regulation to on-balance-sheet and off-balance-sheet items. The fixed coefficients are intended to be an estimate of how that balance sheet item would respond in the LCR’s specified stress scenario, but they do not depend on current conditions. For example, 3 percent of fully insured retail deposits are assumed to leave the bank, 10 percent of committed credit lines to nonfinancial firms are assumed to be drawn and 50 percent of the principal from small business loans scheduled to mature within 30 days is assumed to be an inflow, with the rest assumed to roll over.[4]
Under the LCR’s assumptions, cash inflows are only counted up to 75 percent of cash outflows. This cap is included to limit the extent to which a bank relies on cash inflows to meet its liquidity needs.[5]
The international standard for the LCR defines the stress scenario that the metric is intended to replicate.[6] The conditions are intended to be a combined idiosyncratic and market-wide shock resembling the Global Financial Crisis plus some specific troubles for the individual bank. They include a partial loss of short-term funding, a runoff of some deposits and a three-notch downgrade in credit ratings. Moreover, the scenario envisions “…the need for the institution to fund balance sheet growth arising from non-contractual obligations honoured in the interest of mitigating reputational risk.”[7] A bank that only experienced a three-notch credit downgrade and is still funding balance sheet growth is not a bank in the midst of a severe depositor run. That is, and critically, the scenario is not intended to cover all possible bad contingencies, nor is it premised on the bank becoming insolvent. A requirement that assumed everything that can go wrong does go wrong would achieve what George Osborne, former U.K. Chancellor of the Exchequer, called “the stability of the graveyard.”[8]
The requirement is thus
LCR = HQLA Cash outflows − Cash inflows ≥ 1
with inflows counted only up to 75 percent of outflows.
This formulation creates the misimpression that a bank’s liquidity is its holdings of liquid assets. It has been understood since the 19th century that one cannot make a bank more liquid by requiring it to hold liquid assets because the bank cannot use the liquid assets without violating the requirement. A common analogy is that such a requirement is a “concrete air bag.”
Instead, bankers and bank examiners have long judged that a bank is liquid if it has a diversified source of reliable funding including contingency funding. In a sense, the LCR recognizes that reality, but one has to rearrange the terms to see it clearly. The requirement can be rewritten as
HQLA + Cash inflow liquidity resources ≥ Cash outflow liquidity need
Banks must have assets that they can convert quickly into cash as well as cash inflows that, in combination, are sufficient to meet their cash outflows. The bank’s liquidity need is the cash outflow; its liquidity resources are its liquid assets as well as its cash inflows.
Why the Debate Has Been About HQLA Rather Than Inflows
Even though a borrowing from the central bank would clearly be a cash inflow, the discussion about the treatment of discount window borrowing capacity has almost always been about whether the capacity should be considered HQLA. That odd outcome is a consequence of how central bank borrowing capacity entered the LCR from the start.
From the beginning, the LCR has included a provision to recognize the capacity to borrow from the central bank. That provision, however, was initially limited to jurisdictions that had an insufficient supply of domestic currency HQLA (that is, countries without a lot of government debt). Specifically, those jurisdictions were able to count a committed liquidity facility (CLF) from the central bank as HQLA. The capacity to borrow from the central bank was treated as HQLA because it was meant to make up for insufficient HQLA. In January 2014, the Basel Committee went further. The Committee agreed that CLFs could be used by all jurisdictions, including those with sufficient HQLA, subject to a fee in normal times of at least 75 basis points.[9] In the preamble to the final rule implementing the LCR in the United States, the banking agencies said they were considering recognizing central bank committed liquidity facilities as HQLA and might propose including such capacity at a future date.[10] They have never done so.
The point for present purposes, though, is that casting central bank credit as HQLA rather than as an inflow was an artifact of the need to increase the quantity of HQLA for jurisdictions with insufficient amounts that the original CLF provision was written to solve. That casting has invited the question “Is the capacity to borrow from the central bank as certain a means of funding as other HQLA?” whereas the appropriate question should be “Is the capacity to borrow from the central bank as certain as other cash inflows recognized in the LCR?”
As an Absolute Matter, How Certain Would Discount Window Access Be in the LCR Scenario?
In the United States, a financially sound bank is provided “primary credit” at the discount window. Primary credit is the Fed’s no-questions-asked, business-as-usual lending facility. If a bank does not qualify for primary credit, it may receive secondary credit.
“Financially sound” means at least CAMELS-3 rated and adequately capitalized.[11] That is, to become ineligible a bank would have to either have a CAMELS rating of 4 or be undercapitalized. A CAMELS rating of 4 indicates that a bank “generally exhibits unsafe and unsound practices or conditions, with serious financial or managerial deficiencies producing unsatisfactory performance.” A bank that is “undercapitalized” is subject to a raft of dire requirements including ceasing to pay dividends, ceasing to grow and no acquisitions, new branches or new lines of business.
In practice, a bank that does not qualify for primary credit is experiencing troubles that are much more severe than the situation the LCR is designed to cover. For example, when Silvergate bank experienced a 70 percent deposit outflow in the fourth quarter of 2022, it still qualified for primary credit.[12] When First Republic Bank failed in May 2023, it still had primary credit loans outstanding.[13] Moreover, a bank in a holding company with a single-point-of-entry resolution strategy should always be adequately capitalized because it would be recapitalized if necessary by the parent, which in turn would be recapitalized by its long-term debt becoming equity. Thus, while the CAMELS and capital requirements for primary credit appear significant on paper, in practice they do not bind until a bank is at the edge of insolvency.
As discussed in Madigan and Nelson (2002), primary credit was created as a replacement for the adjustment credit program with the intent of increasing banks’ willingness to borrow.[14] Adjustment credit, which had been the main type of discount window credit until it was replaced by primary credit in 2003, was provided at a below-market rate, so there were rules about banks’ ability to borrow. The rules were subjective, and banks often cited uncertainty about whether they were allowed to borrow as a reason for their reluctance to utilize the facility. Because primary credit, by contrast, is provided at an above-market rate to financially sound banks, there is no need for any additional rules. There are no restrictions on how banks use primary credit.
Madigan and Nelson state that “The main purpose of the proposed primary credit program is to make short-term credit available as a backup source of liquidity to generally sound institutions.” In support of that purpose, the five federal agencies that then regulated depository institutions jointly issued the “Interagency Advisory on the Use of the Federal Reserve’s Primary Credit Program in Effective Liquidity Management” (Federal Reserve Supervision and Regulation letter 03-15).[15] The advisory states:
By enhancing the availability of discount window credit, the new primary credit program offers depository institutions an additional tool for managing short-term liquidity risks…Since primary credit can serve as a viable source of back-up, short-term funds, supervisors and examiners should view the occasional use of primary credit as appropriate and unexceptional.
Since 2003, the Fed has taken further steps to encourage banks to see primary credit as a reliable way to meet funding needs under stress. Whereas primary credit was initially extended overnight, with terms of up to a few weeks available only to small institutions, primary credit is now extended with an initial maturity of 90 days, and banks can repay and renew the loan on request.[16]
As the Fed explains:[17]
A term primary credit loan with a remaining maturity greater than 30 days would be outside of the liquidity coverage ratio (LCR) requirement’s 30-day stress time horizon. For example, if a bank were to borrow term primary credit for 90 days, immediately upon borrowing, the bank’s reserve balances would increase. This increase in reserve balances would increase the numerator of the LCR. At the same time, the loan with a maturity longer than 30 days would not create an outflow in the denominator of the LCR. As a result, the borrowing bank’s LCR increases. As the remaining maturity of the loan declines, the bank may choose to pre-pay the loan and request a new loan up to 90 days.
Moreover, to further instill confidence that borrowing from the discount window can help a bank meet its LCR, “…Reserve Banks have waived their rights to require repayment on demand for a term primary credit loan with a remaining maturity of more than 30 calendar days…”[18]
Thus, the Fed wants banks to see primary credit as a very reliable source of funding. For the discount window to accomplish its monetary policy and financial stability objectives, banks need to be able to lend to other institutions when money market rates rise above the discount rate, confident that they will be able to fund those loans at the end of the day if needed with primary credit.
As a Relative Matter, How Certain Are the Funds From Monetizing HQLA?
A large bank’s HQLA can easily include hundreds of billions of dollars of Treasury securities. But a bank cannot sell or repo such a large quantity of Treasury securities in a few days to meet its funding needs without disrupting markets. In an FAQ on Regulation YY published in August 2024, the Fed solved this problem for internal liquidity stress tests (ILSTs) — the stress tests imposed through the examination process — by permitting banks to incorporate discount window borrowing into their scenario analysis as a supplement to private market monetization channels.[19] That is, for ILSTs, the certainty of being able to meet outflows with Treasury securities (or agency debt or agency MBS) hinges on the reliability and availability of the discount window.[20]
As a Relative Matter, How Certain Are Cash Inflows?
As noted above, the comparison for certainty purposes for discount window borrowing should be cash inflows, not HQLA. In the LCR, cash inflows are generated by scheduled interest and principal payments on loans and securities, maturing secured lending, and similar items. Those payments are then reduced by standardized assumptions about how much of the maturing credit the bank will roll over. These assumptions were calibrated entirely using judgement and policy objectives — they were not estimated using historical experience. Note that the reductions have nothing to do with the certainty of repayment. Credit uncertainty is handled separately: the rule excludes from inflows any amount owed by a counterparty that is nonperforming, or that the bank has reason to expect will become nonperforming within 30 days.[21] What is left is a behavioral assumption about the bank. When the rule counts 50 percent of the principal on maturing small business loans, it is not saying that half the borrowers might not pay. It is saying that the bank will lend half of it back out.
The inflow assumptions are conservative, but the arrival of the inflows is not certain. Nevertheless, up to the 75 percent cap, a dollar of inflow satisfies the LCR no less than a dollar of HQLA.
How Certain Are Reserve Balances?
Quite certain, of course, but not completely certain. Under the Fed’s Operating Circular 1, a bank’s balance at a Reserve Bank is subject to a continuing lien securing any obligation the bank owes to any Reserve Bank, whether existing now or arising in the future. Upon an event of default, any Reserve Bank may set off against that balance without demand or notice. And an event of default includes not only insolvency or a failure to pay an obligation when due, but also whenever a Reserve Bank deems itself insecure with respect to the bank’s financial condition or its ability to perform its obligations. Separately, and requiring no default at all, the administrative Reserve Bank may require a bank to maintain adequate balances in such amount as it determines, and a Reserve Bank may restrict or limit a bank’s use of its master account.[22]
How Certain Must a Central Bank Line of Credit Be in the Current Standard?
As noted, the LCR standard already includes a provision to recognize the capacity of a bank to borrow from its central bank under committed liquidity facilities. How certain must the ability to tap a CLF be? Not completely certain. The standard states:
Conditional on the bank being assessed to be solvent, the [committed liquidity facility] contract must otherwise be irrevocable prior to maturity and involve no other ex-post credit decision by the central bank. The commitment period must exceed the 30-day stress period stipulated by the LCR framework. [emphasis added]
That is, the central bank can refuse to lend under the committed line of credit if the bank is not solvent.
Why the Double Standard?
Why then is the capacity to borrow from the central bank against prepositioned collateral being held to a standard of complete certainty when no current source of funding in the LCR is held to that standard? As noted, it could be in part because of the historical fluke that a cash inflow from the central bank ended up being considered HQLA. It could also, however, arise from the widespread and incorrect belief that the LCR was intended to keep banks from borrowing from their central bank’s business-as-usual lending facilities. Not only is that reason absent from the international standard and the U.S. rule, it would be a strange objective, since lending to banks is how many central banks supply reserves in the first place. As discussed in “Discount Window Stigma: We Have Met the Enemy, and He Is Us,” the century-long flip-flop by the Federal Reserve about whether it does, or does not, want banks to borrow from it is a key cause of stigma.[23]
Ways Forward
There are several ways jurisdictions could resolve any concerns about certainty when including the capacity to borrow from the central bank against prepositioned collateral in liquidity assessments.
- The capacity could be included as a cash inflow rather than HQLA, and therefore subject to the 75-percent-of-outflows cap already in place to limit overreliance on inflows.
- The capacity could be restricted to those banks that would still qualify for primary credit after weathering the LCR stress scenario. This is what SR 03-15 already directs banks to do in their contingency planning, and it is a more demanding test than contemporaneous access to primary credit, because the stress scenario includes a deterioration in the bank’s financial condition.[24]
- The central bank could provide a committed liquidity facility as recognized in the LCR (although not including the uneconomic 75 basis point fee). The U.S. rulemaking implementing the LCR indicated that the agencies could consider CLFs in the future.
- The capacity could be included as HQLA but subject to a diversification cap. A core principle of liquidity risk management is maintaining diversified sources of contingency funding, and the Basel standard already supplies the template: a CLF is included in Level 2B and is therefore subject to both the 15 percent Level 2B cap and the 40 percent Level 2 cap.
Conclusion
The Fed has spent more than 20 years telling banks that primary credit is available; that it will be granted no questions asked; that it can be borrowed for 90 days and renewed on request; that it will not be called; and that examiners should regard its use as appropriate and unexceptional. It is hard to think of another source of funds in the LCR about which so much has been promised, in advance and in public. The problem is not that the discount window is uncertain. The problem is that bank regulators have not understood business-as-usual central bank lending and think it is bad.
[1] See, for example, Bessent, Scott (as delivered by Jonathan McKernan), “Remarks: A Reset on Liquidity Regulation,” U.S. Department of the Treasury, March 3, 2026, https://home.treasury.gov/news/press-releases/sb0412.
[2] For example, the international standard states that jurisdictions may wish to apply a haircut to level 1 securities based on typical repo haircuts along with other things. Basel Committee on Banking Supervision, “Basel III: The Liquidity Coverage Ratio and liquidity risk monitoring tools,” paragraph 49, January 2013, https://www.bis.org/publications/201301-standards-basel-iii-liquidity-coverage-ratio-and-liquidity-risk-monitoring-tools.pdf.
[3] One exception is that Ginnie Mae MBS, which carry the full faith and credit of U.S. government, are treated as Treasury securities in the LCR.
[4] 12 CFR 249.32(a) (retail deposit outflows), 249.32(e) (commitments), and 249.33(c) (retail cash inflows). The LCR final rule explains “The agencies also expect covered companies to maintain ample liquidity to sustain core business lines, including continuing to extend credit to retail customers and wholesale customers and counterparties that are not financial sector entities. Indeed, one purpose of the proposed rule was to ensure that covered companies would have sufficient liquidity to sustain such business lines during a period of liquidity stress.” Not only is the explanation illogical (liquidity is not needed to renew a maturing loan), the treatment makes it more, not less, costly for a bank to extend a small business loan by requiring banks to maintain additional HQLA whenever they make one. 79 Fed. Reg. 61,440, 61,512 (Oct. 10, 2014).
[5] The cap also prevents the mathematical breakdown that would occur if the denominator of the LCR equals zero or becomes negative.
[6] Basel Committee on Banking Supervision, “Basel III: The Liquidity Coverage Ratio and liquidity risk monitoring tools,” January 2013, https://www.bis.org/publ/bcbs238.pdf.
[7] Basel Committee on Banking Supervision, “International Framework for Liquidity Risk Measurement, Standards and Monitoring,” paragraph 22, December 2009, https://www.bis.org/publications/200912-consultation-international-framework-liquidity-risk-measurement-standards-and-monitoring.pdf.
[8] Osborne, George, “Speech by the Chancellor of the Exchequer, Rt Hon George Osborne MP, at the Lord Mayor’s Dinner for Bankers and Merchants of the City of London,” HM Treasury / GOV.UK, 2012, https://www.gov.uk/government/speeches/speech-by-the-chancellor-of-the-exchequer-rt-hon-george-osborne-mp-at-the-lord-mayors-dinner-for-bankers-and-merchants-of-the-city-of-london.
[9] Basel Committee on Banking Supervision, “The Liquidity Coverage Ratio and restricted-use committed liquidity facilities,” January 2014, https://www.bis.org/publ/bcbs274.htm. The CLF conditions are set out in the annex to the accompanying press release.
[10] 79 Fed. Reg. 61,440, 61,457 (Oct. 10, 2014).
[11] The Federal Reserve’s formulation is that “in most cases, an institution must have a CAMELS composite rating of 3 or stronger and a PCA designation of ‘adequately capitalized’ or stronger,” and that Reserve Banks may also consider other information. Federal Reserve Discount Window, “Primary and Secondary Credit Programs,” https://www.frbdiscountwindow.org/pages/general-information/primary-and-secondary-credit-programs.
[12] Nelson, Bill (2025), “Silvergate Bank: A Discount Window Success Story,” Bank Policy Institute, January 9, https://bpi.com/wp-content/uploads/2025/01/Silvergate-Bank-A-Discount-Window-Success-Story.pdf.
[13] It also had loans outstanding from the Fed’s bank term funding program — the emergency facility the Fed set up to help banks experiencing outflows in spring 2023. To borrow from the BTFP, a bank had to qualify for primary credit.
[14] Madigan, Brian, and William Nelson (2002), “Proposed revision to the Federal Reserve’s discount window lending programs,” Federal Reserve Bulletin, July 2002, pp. 313–319, https://www.federalreserve.gov/pubs/bulletin/2002/0702lead.pdf.
[15] Board of Governors of the Federal Reserve System, “Interagency Advisory on the Use of the Federal Reserve’s Primary Credit Program in Effective Liquidity Management,” SR 03-15, July 23, 2003, https://www.federalreserve.gov/boarddocs/SRLETTERS/2003/sr0315a1.pdf.
[16] Board of Governors of the Federal Reserve System, press release, March 15, 2020, https://www.federalreserve.gov/newsevents/pressreleases/monetary20200315b.htm. The 90-day term, prepayable and renewable by the borrower daily, remains in effect and is restated in the Board’s current publications.
[17] Board of Governors of the Federal Reserve System, discount window and discount rate information, https://www.federalreserve.gov/monetarypolicy/discountrate.htm.
[18] Federal Reserve Discount Window, “Frequently Asked Questions,” https://www.frbdiscountwindow.org/en/Pages/General-Information/Frequently-Asked-Questions.aspx. The waiver, announced March 15, 2020, is of the rights provided in section 5.1(a) of Operating Circular 10.
[19] Board of Governors of the Federal Reserve System, “Regulation YY Frequently Asked Questions,” August 13, 2024, https://www.federalreserve.gov/supervisionreg/legalinterpretations/reg-yy-frequently-asked-questions.htm.
[20] In the same FAQ, the Fed also reiterated that banks could not anticipate a cash inflow at the overnight or 30-day horizon from borrowing from the discount window even if the bank had collateral prepositioned at its Reserve Bank. The net result is that banks can anticipate borrowing against Treasury and agency collateral but not against other types of collateral. To be clear, in its discount window operations, the Fed makes no distinction between types of collateral other than the haircut. Most collateral, in fact, is large pools of business and consumer loans. As then-Vice Chair for Supervision Michael Barr said in a December 2023 speech about loan collateral, these are the very assets that would be best pledged to the discount window, so that more liquid assets are held for other uses. Information on the composition of discount window collateral is available here: https://www.federalreserve.gov/monetarypolicy/discount-window-readiness.htm. Then-Vice Chair Barr’s speech is available here: https://www.federalreserve.gov/newsevents/speech/barr20231201a.htm.
[21] 12 CFR 249.33(a)(5).
[22] Federal Reserve Banks Operating Circular No. 1, sections 2.2, 2.11, and 5.3, https://www.frbservices.org/binaries/content/assets/crsocms/resources/rules-regulations/090123-operating-circular-1.pdf.
[23] “Discount Window Stigma: We Have Met the Enemy, and He Is Us,” Bank Policy Institute, August 10, 2021, https://bpi.com/wp-content/uploads/2021/08/Discount-Window-Stigma-We-Have-Met-the-Enemy-and-He-Is-Us.pdf.
[24] “Against What Liquidity Risks Should a Bank Self-insure?” Bank Policy Institute, December 1, 2022, https://bpi.com/against-what-liquidity-risks-should-a-bank-self-insure/.
