The failure of SVB revealed that bank deposits can run much faster than previously thought. While SVB (and Signature) had unusually concentrated deposit bases that contributed to the speed of the run, the revised view of deposit stickiness will certainly call into question the assumptions for deposit outflows in banks’ liquidity requirements. But the solution isn’t as simple as dialing outflow rates up. Taken alone, this would require banks to devote even more of their balance sheets to lending to the government in the form of Treasury securities and deposits at the Federal Reserve, rather than lending to businesses and households. Unless we begin to think outside the current regulatory box, bank liquidity regulations may choke off banks’ contribution to economic growth and employment.
For this reason, the Bank Policy Institute has called for any recalibration of liquidity outflow assumptions to occur as part of a comprehensive redesign of the liquidity requirement framework.[1] An important part of that redesign could be the offering of committed liquidity facilities by the Federal Reserve and the recognition of such facilities as sources of funding in regulatory and supervisory assessments of banks’ liquidity. CLFs are collateralized, committed lines of credit that would be provided by the Federal Reserve to financially sound banks for a fee. Just as they do now for the discount window, banks would be allowed to pledge their loans as collateral for a CLF. As a result, banks with CLFs could still provide credit to the economy while also internalizing and responsibly managing their liquidity risk.
CLFs would sit alongside and be a sensible complement to other Federal Reserve liquidity programs, like the discount window and the standing repo facility, and thereby enhance and strengthen the Federal Reserve’s traditional and central role as liquidity provider to the banking system during periods of stress.[2] Because a bank would pay a fee for its CLF, and pay an above-market rate on any draws under the CLF, it would be self-insuring against its liquidity risk and have a financial incentive to manage that risk.
As discussed, CLFs are a component of the international Basel standard for liquidity regulations. To make CLFs functional, the United States would need to design CLFs slightly differently, but because the U.S. requirement is more stringent than the Basel standard, and likely to become more so, CLFs could be incorporated while remaining Basel-compliant.
This note explains what CLFs are, how they currently fit into the international standards for liquidity requirements, and the design choices that the Fed would need to make if providing them.
What Are the Basic Characteristics of a CLF?
CLFs are defined in the flagship international standard for liquidity regulation: the liquidity coverage ratio. The LCR is designed to ensure that a bank would have sufficient funding resources to meet its obligations during a 30-day period of bank and market liquidity stress. The international LCR standard defines what it calls “restricted-use committed liquidity facilities” that can be counted as a source of liquidity. A CLF must have at least three defining characteristics in order to meet the international standard.
- It must be a firm commitment by the central bank to provide funds on demand for at least the subsequent 30 days unless the bank becomes insolvent.
- It must be collateralized.
- The central bank must charge a fee for the line.
There are additional requirements for when a jurisdiction can allow banks to count a CLF as a source of liquidity, and additional more detailed rules about CLF design.
Table 1 is a term sheet for possible CLFs that could be provided by the Federal Reserve. The terms and conditions, which are just hypothetical, are discussed in subsequent sections, including how they compare to the Basel LCR’s requirements.
CLFs as Part of the Basel LCR Design
The LCR is defined as the ratio of high-quality liquid assets (HQLA), such as Treasury securities and deposits at the central bank (reserve balances), to estimated net cash outflows (projected outflows minus projected inflows) over a 30-day period of stress.[3] Logically, the CLF would enter into projected net cash outflows as an inflow because draws on the CLF would be a cash inflow. But in the Basel standard, CLFs are included in HQLA because CLFs were originally included as replacement HQLA for countries that had insufficient regular HQLA.
HQLA mostly consists of government securities and deposits at the central bank. In some jurisdictions, most notably Australia, there was insufficient government debt available to provide the HQLA the banking system needed, and the Reserve Bank of Australia didn’t want to expand its balance sheet sufficiently to provide HQLA in the form of reserve balances. Such jurisdictions were allowed to treat CLFs as HQLA. (See paragraph 58 of the standard here.) In 2014, the Basel Committee adjusted the LCR standard so that all jurisdictions can count CLFs as HQLA. (See the Annex here.)
A critical parameter in the LCR standard’s definition is the fee that must be charged for the CLF (the fee for the line itself, not the interest rate on draws under the line). For countries that qualify based on a lack of government debt, the fee must be set no lower than the difference between the average yield on the assets eligible to secure the CLF and the yield on other HQLA after adjusting for differences in credit risk and the haircuts applied to the collateral. The fee was intended to ensure that firms had the same financial incentive to manage their liquidity risk using a CLF as they did when using other HQLA. For example, and ignoring the adjustment for risk and any haircut, suppose a bank could pledge corporate bonds as collateral for the CLF and the interest rate on corporate bonds was 4.5 percent. Suppose further that the interest rate on Treasury securities, which count as regular HQLA, was 4 percent. Then the fee for the CLF would be 0.5 percent. After taking into account the haircuts applied to eligible collateral and the differences in credit risk, the Reserve Bank of Australia concluded that the appropriate fee was 15 basis points, later raising it to 20 basis points.[4]
For countries that do not qualify for CLFs based on a lack of government debt, the fee is required to be much higher. Specifically, it must be the greater of
- 75 basis points
- 25 basis points above the difference in yield on the assets used to secure the CLF and the yield on a representative portfolio of HQLA, after adjusting for any material differences in credit risk. (That is, 25 basis points above the minimum required fee for countries that qualify because of insufficient HQLA.)
During periods of market-wide stress, the fee can be lowered to the level applicable to countries with insufficient HQLA. Needless to say, the required fee is so high that no country has chosen to offer CLFs other than those qualifying for the lower fee. For further discussion, see the appendix to this note.
How Might CLFs Look in the United States?
In the final U.S. LCR rule, the banking agencies noted that they were considering including CLFs as a recognized source of liquidity.
The agencies are considering the merits of including central bank restricted committed facility capacity as HQLA for purposes of the U.S. LCR requirement and may propose at a future date to include such capacity as HQLA.[5]
If the Fed were to offer banks CLFs that would count for liquidity requirements in the United States, it would need to make several design choices.
HQLA or projected inflow?
As noted, even though a CLF is not an asset and provides liquidity in a contingency as a cash inflow for the bank, it is treated as a HQLA in the LCR.
Arithmetically, it doesn’t make much difference whether the CLF is treated as HQLA or a cash inflow. The LCR requirement is:
which can be rewritten as
The second version illustrates the equivalence between HQLA and a projected cash inflow, which makes sense because both provide the bank dollars to meet demands on its funding.
The only reason to treat CLFs as HQLA in the United States would be to adhere as closely as possible to the international standard. But since, in order to work, the CLFs would have to charge a lower fee than required by the Basel standard anyway, there is no particular reason to stick to the illogical treatment as HQLA. Thus, if adopted in the United States, CLFs should enter into the LCR as a projected cash inflow.
CLFs would presumably also be recognized as a source of liquidity in any internal liquidity stress tests that banks were required to conduct and report to their examiners. The distinction between HQLA or cash inflow is essentially irrelevant for the internal liquidity stress tests banks are required to conduct because ILSTs simply determine if the bank has enough liquidity to make it over the interval being tested, regardless of whether the liquidity comes from monetizing an asset or a cash inflow. Banks conduct ILSTs at the overnight, 30-day, 90-day, and one-year horizon. A CLF designed to provide guaranteed funding for 31 days would only be relevant for the horizons of 30 days or less and for the first 30 days of longer-horizon tests.
How much CLF should count toward liquidity needs?
In the LCR, projected cash inflows cannot be more than 75 percent of the projected cash outflows. If that requirement were maintained after including potential future draws on the CLF as a cash inflow, the result would be that at least a quarter of the projected gross cash need would always be met with HQLA.
Another way to incorporate but limit CLFs would be to only allow them to satisfy projected funding needs that are in excess of the Basel standard. Insofar as the U.S. banking agencies raise the assumed deposit outflow rates above the Basel standard in light of the experience of banks recently, that opens up space to count CLF capacity while remaining Basel compliant.
For example, the Basel standard requires banks to have sufficient liquidity at 30 days while the U.S. implementation requires banks to have enough liquidity to make it to 30 days. Because the largest projected liquidity needs tends to occur after just a few days, U.S. banks are held to a higher standard.[6] Four years ago, BPI recommended that banks should be allowed to cover the difference between the “deepest hole” requirement and the “end of 30 days” requirement with borrowing capacity at the discount window (see “Give Banks Credit For Robust Contingent Liquidity Arrangements”), on the view that the bank could borrow from the Fed to cover the within-30-day need. At the time, we noted that such a requirement would “… provide banks an incentive to be prepared to borrow from the discount window if necessary, enhancing financial stability.”
Yet another way to incorporate CLFs would be to allow them to be used to meet HQLA requirements in excess of 10 percent of a bank’s balance sheet. Such an approach would put a cap on the amount to which banks are required to lend to the government directly in the form of holding Treasury securities or indirectly in the form of deposits at the Federal Reserve as opposed to lending to businesses and households. At the same time, it would put a floor on the amount through which banks are meeting their liquidity requirements with assets rather than a CLF.
As of Feb. 26, 2020, banks had collateral pledged to the Fed that provided borrowing capacity of $1.6 trillion.[7] If the Fed were to actively encourage or require banks to pledge more collateral, there probably could be sufficient CLF capacity to provide the additional liquidity required by any recalibration of deposit outflows. CLFs that approximately equaled the increase in banks’ assessed liquidity needs would accomplish the objective of preventing the recalibration from reducing the supply of bank credit.
Which institutions?
The Fed could provide a CLF to a commercial bank or other depository institution under Section 10B of the Federal Reserve Act. That authority, which is the authority used for regular discount window lending, allows the Fed to lend for terms of up to four months against any collateral and at any rate. Lending is always at the discretion of the Reserve Bank, but the Fed could nevertheless indicate that it intended to lend under the circumstances spelled out in the terms of the facility.
The Fed could only provide a CLF to a bank holding company in two circumstances. First, it could provide the CLF if the facility were collateralized using only Treasury or agency obligations under Section 13(13) of the FRA. Second, it could provide the CLF in unusual and exigent circumstances under Section 13(3) of the FRA.
As a result, CLFs are not especially useful for bank holding companies. Treasury debt, agency MBS and agency debt are all highly liquid and already count as HQLA, so a CLF collateralized by them might not be helpful.[8] And liquidity requirements need to be met all the time, so CLFs that are available only in emergencies would also not be especially helpful.
However, upward recalibration of deposit outflows would boost projected liquidity needs at the commercial bank, so corresponding projected cash inflows at the commercial bank would be beneficial. Consequently, the inability to provide CLFs to BHCs may not be a significant limitation. After all, even though only banks can maintain deposits at the Fed (not bank holding companies), such reserve balances are a critical component of how banking institutions satisfy liquidity requirements.
What financial soundness criteria?
Under the Basel standard, banks that have been provided a CLF must be extended a loan on request as long as they are solvent. In practice, “solvent” in any U.S. implementation would probably be defined as “not critically undercapitalized.” If the institution is critically undercapitalized, the Fed is legally subject to financial penalties if it lends for more than five days and the lending increases costs to the FDIC if the institution fails.[9]
The Fed could and should apply tighter financial soundness criteria for qualifying for a CLF. The natural choice for the Fed would be to only provide CLFs to banks that qualify for primary credit – banks that are at least CAMELS 3 rated and adequately capitalized.[10]
Maturity of the lines and maturity of loans?
CLFs would have a perpetual maturity unless a bank with a facility ceased being financially sound or the bank decided that it no longer wanted the facility. Facilities could only be canceled by the Fed with a minimum of 31 days’ notice.[11] To reduce cliff effects, the CLF of a bank that became unqualified but remained solvent could be gradually wound down.
Draws under the CLF would have overnight maturities, renewable for as long as the institution had access to the CLF.[12] Because a bank that had a CLF but became ineligible would retain access for at least 31 days, at any point when the CLF was counted as HQLA it would provide credit that was available for at least one day longer than the LCR’s 30-day horizon.
Memo item: Reliability
The OCC recently revised its examination manual on liquidity and added a new section on the discount window. The manual states “…Federal Reserve Banks are not required to lend through the discount window and may turn banks away.”[13] CLFs would address the OCC’s concerns about reliability. An institution that had a CLF would be assured funding for at least the subsequent 31 days unless it became insolvent.
What collateral?
The Federal Reserve accepts nearly all bank assets as collateral for discount window loans. The same policy would be appropriate for CLFs. To each type of asset, the Fed applies a haircut to the fair value of the instrument to determine lendable value. The haircut is selected to provide a high likelihood that the Fed could liquidate the collateral and recover at least the value of the loan.
What fee and rate?
As discussed in the appendix, one could make an argument for an above-market fee or a below-market fee, or a market-based fee. An above-market fee would only make sense if CLFs were seen as undesirable, in which case it would be simpler to just not offer them. Charging a market-based fee, which would be roughly 15 to 20 basis points, would be consistent with how the Fed prices its other services.
A fee set at this level would not be compliant with the Basel standard. If the fee were set in line with the Basel standard, that is, 75 basis points, then there would be no point in providing CLFs. No countries have adopted CLFs because, at 75 basis points, roughly five times the market price for a similar line, CLFs are uneconomic. That said, as discussed, non-Basel-compliant CLFs can be incorporated into the U.S. liquidity regulation framework while keeping the framework Basel compliant.
The interest rate on a loan drawn under the facility should be set at an above-market rate to discourage banks from using the facility as an ongoing source of funding as opposed to a contingency source of funding, and to encourage banks to repay their loans quickly once their funding pressures abate. The primary credit rate (the discount rate) is normally also priced at a premium over money market rates, so the term sheet in table 1 proposes that the interest rate on draws on the CLF be set equal to the primary credit rate. Currently, however, the primary credit rate is only slightly above money market rates, so an appropriate interest rate on CLF draws may be higher. An important reason to have a meaningful spread above money market rates is so that it will not be necessary to discourage or limit use of the facility in ways other than the interest rate alone.
Resemblance to Recent Proposals by Mervyn King and Paul Tucker
In The End of Alchemy, Mervyn King, former governor of the Bank of England, proposed that banks be required to maintain collateral at the central bank with lendable value equal to all the bank’s deposits and short-term borrowing.[14] As discussed in a recent interview published in the Financial Times, Paul Tucker, former deputy governor of the BoE, made essentially the same proposal.
The King and Tucker proposal is similar to allowing banks to count CLFs as HQLA but taken to an extreme. If CLFs count as a fraction of HQLA and banks are required to maintain HQLA equal to a fraction of deposits and short-term debt, then banks would be required, or at least given the option, to maintain collateral at the central bank equal to a fraction of those liabilities. King and Tucker propose banks maintain collateral with a lendable value equal to 100 percent of the deposits and other shorter-term liabilities.
Under the King and Tucker proposal, there would be no need for additional liquidity or even capital requirements or deposit guarantees. There would be no run the bank could face that could not be covered by borrowing from the central bank. Because a bank would essentially have to fund the central bank’s haircuts on its assets with equity and long-term debt, the collateral haircuts would become TLAC (Total Loss Absorbing Capital (equity and debt)) requirements. In the case of the Federal Reserve, haircuts on loans vary from 5 percent to 82 percent. TLAC requirements range from 7 to about 20 percent, so the King-Tucker proposal would require a massive increase in the percent of banks’ assets that are funded with equity and longer-term debt.
Benefits and Challenges[15]
In addition to allowing banks to continue to lend to businesses and households, with the attendant gain for economic activity, CLFs would have several other benefits. CLFs could encourage banks to make use of their HQLA when under stress because the HQLA could be replenished by increasing the size of the CLFs. Conversely, because banks could draw on their CLFs, it may be less necessary for banks to sell or repo their HQLA when markets for the assets are stressed. CLFs would also provide the Fed additional tools with which to respond to a financial crisis, in particular by easing the terms on and increasing the capacity of the CLFs. In addition, CLFs would reduce the demand for reserve balances, allowing the Federal Reserve to become smaller. Lastly, CLFs would allow the Fed to charge banks for contingency funding support, income that would accrue to Treasury.
There would also be challenges associated with successful CLF implementation. For one, for CLFs to work, banks would have to be willing to borrow from the facility when under stress, and there is a substantial stigma associated with using the discount window. If banks were unwilling to borrow, it would be counterproductive to count the CLF as a source of liquidity. That said, there may be less stigma associated with drawing on a CLF because banks would be paying for the line and could therefore see it as their right to borrow. Another challenge is that the Fed would need to be willing to transition any bank that became financially unsound out of the facility, an action that could contribute to the failure of the institution and may be difficult to take during periods of system-wide stress.
Self-Insurance?
Opposition to counting CLFs as a projected cash inflow or HQLA could be based on the view that banks should be required to self-insure against their liquidity risks. If “self-insurance” is defined as maintaining sources of liquidity other than the central bank (except for high-yielding deposits at the central bank, which for some reason count as self-insurance), then it is tautologically true that CLFs would reduce self-insurance. However, if “self-insurance” means you have purchased the resources necessary to meet liquidity stress situations so that you are not counting on the government to bail you out for free when you run out of funds, then CLFs increase self-insurance. Banks would pay for CLFs, they would be purchasing liquidity insurance, and therefore self-insuring in the sense that matters.
Only the central bank can provide immediate, completely reliable liquidity insurance and it is therefore the preferred entity to provide it from a public policy perspective.[16] Similarly, although deposits at commercial banks do not count as HQLA or a projected cash inflow, deposits at the Fed do count as HQLA, presumably because there is no chance that the reserve balances will not be available when needed and because deposits at the Fed do not increase interconnectedness among financial institutions.[17] CLFs would also be reliably available and would not increase interconnectedness.
Put simply, if the banks are purchasing the lines of credit, they are self-insuring by any reasonable definition.
Conclusion
From a liquidity perspective, the world of a deposit-taker appears to be a riskier place than it did at the beginning of March. To manage those risks while allowing banks to continue their core function of accepting deposits and making loans, it is necessary to recognize and formalize the important role played by the Federal Reserve in providing liquidity insurance. In particular, the Fed should offer banks committed liquidity facilities at a market-based fee, and the CLFs should count as sources of liquidity in the banks’ LCRs and ILSTs. Doing so will support economic activity as well as the liquidity of the banking system.
[1] See “BPI Policy Recommendations Related to the 2023 Banking Stress” and “Is It Time For a Holistic Review of Liquidity Requirements?”
[2] See “Improving The Government’s Lender of Last Resort Function: Lessons From SVB and Signature Bank.”
[3] To be clear, the outflows and inflows are not really “projected.” They are determined by multiplying the bank’s on- and off-balance sheet items by fixed parameters established in the regulation. The parameters do not depend on the current condition of the bank or the economy.
[4] RBA stopped offering CLFs on January 1, 2023 because the Australian Prudential Regulatory Authority concluded that there was now sufficient regular HQLA available.
[5] Federal Register, Vol. 79 No. 197, p. 61457.
[6] The U.S. LCR is “gold-plated” to use the term coined by Mark Van Der Weide, the Board General Counsel.
[7] See the footnote to table 5 on page 13 of the Fed’s quarterly balance sheet report available here. The Fed has not provided information on the total amount of collateral pledged to the discount window for more than 3 years.
[8] CLFs for primary dealers authorized under 13(13) (not 13(3)) that accepted Treasuries, agency debt, and agency MBS as collateral would serve a similar purpose to the Standing Repo Facility, possibly relieving pressure on the cash and repo markets for these securities in times of stress and providing primary dealers confidence that funding would be available at the end of the day at a known price.
[9] See Federal Reserve Act Section 10B (b).
[10] Just as the Fed is restricted from making loans to critically undercapitalized institutions for more than 5 days, the Fed is also restricted from making loans to undercapitalized institutions for more than 60 days out of any 120-day period (it is a little more complicated than that, but that is the gist). A bank that became undercapitalized would lose its eligibility for the CLF after 31 days, which is less than 60 days.
[11] The notice period could, of course, be set longer than 31 days to allow a longer period to address the bank’s financial troubles or reserve the bank.
[12] Alternatively, loans could be for 31 days, but in that case, a bank that lost its eligibility for the CLF should receive no new loans starting immediately.
[13] Comptroller’s Handbook, Safety and Soundness, Liquidity. Version 1.2, May 25, 2023, p. 14.
[14] King, M. (2016). The end of alchemy: Money, banking, and the future of the global economy. WW Norton & Company.
[15] For a more extensive discussion of the costs and benefits see William Nelson, “Recognizing the value of the central bank as a liquidity backstop,” Staff Working Paper 2017-1, The Clearing House, January 2017.
[16] Put another way, because the Fed can create any funding it needs, it does not face liquidity risk. When the Fed provides a line of credit to a bank, the amount of liquidity risk in the system as a whole declines. When a bank provides another bank a line of credit, the liquidity risk is simply moved from the bank purchase the line to the bank providing the line.
[17] See also “Against what liquidity risks should a bank self-insure?”
[18] Friedman, M. (2017). The optimum quantity of money. Routledge.
Appendix: What is the right fee?
What fee for a CLF would be appropriate? Many in the Basel Committee community saw the whole point of the LCR to be to prevent banks from borrowing from the central bank and so judged CLFs to be undesirable. For example, in January 2013, when the LCR was approved by the oversight body of the Basel Committee, Stefan Ingves, the head of the Committee, stated, “Today’s agreement is a clear commitment to ensure that banks hold sufficient liquid assets to prevent central banks becoming the ‘lender of first resort’.” Thus, CLFs were only begrudgingly offered, and offered at terms that made them undesirable.
Jeremy Stein of Harvard, when he was a member of the Federal Reserve Board, noted in a 2013 speech that “…if one is going to make an argument in favor of adding preventative liquidity regulation such as the LCR on top of capital regulation, a central premise must be that the use of LOLR capacity in a crisis scenario is socially costly, so that it is an explicit objective of policy to economize on its use in such circumstances.” Stein, however, was just arguing against central banks providing liquidity insurance for free; he supported the use of CLFs in his speech and was not arguing that the fee should be prohibitive.
In another context, Stein and other economists have argued that it is socially desirable for central banks to provide liquidity to the financial system generously. Gorton (2010), Gorton and Metrick (2012), and Stein (2012) have argued that investors accept a low yield on instruments they perceive as money-like, and financial instability is often the result of private firms using such money-like claims to fund investments in longer-term illiquid assets. In a 2018 paper, Stein and I and several colleagues argued that the Fed should oversupply money-like instruments, in particular reserve balances, to boost their yield and reduce the attractiveness of other money-like instruments as sources of funding for shadow banks.
These economists’ arguments are reminiscent of Milton Friedman’s that it is socially optimal for the central bank to create money until individuals are indifferent between holding money or holding other short-term investments because it is costless to produce money.[18] Similarly, because illiquidity is a market failure, if it is essentially costless for central banks to create CLFs, the socially optimal fee would be for the CLFs would be zero.
Stein argues that the social costs of central bank lending are the risks of loss and moral hazard. Both of these costs should be addressable by the central bank choosing conservative haircuts and charging an above-market interest rate on draws. Against these costs, moreover, is the fact that the central bank, and only the central bank, can provide liquidity against illiquid collateral without incurring liquidity risk.
Economically, a reserve balance at a Federal Reserve Bank and a CLF from a Federal Reserve Bank are extremely similar. Both are promises by the central bank to provide funds on demand in the future. The difference is that a reserve balance is ultimately an investment in the government securities owned by the Fed whereas a CLF is ultimately an investment in the bank loans pledged as collateral.
On balance, it may be appropriate for the fee to be set approximately equal to the market price for a similar line. As it happens, the market fee for such a line is about 15-20 basis points, consistent with the fee the RBA chose for its CLFs.
