More Taxis Sitting Idle

There is a timeworn parable told by economists that study bank liquidity requirements concerning the last taxi at the train station.

An economist’s train pulls into the station in a remote college town late in the evening and she is glad to see that there is a taxi at the station.  She jumps out of the train, goes to the taxi, jumps in and asks to be taken to her hotel.  The taxi driver says, “I’m sorry, I can’t take you.  Recognizing the importance of a taxi always being available for arriving train passengers, there is a town ordinance requiring that there always be at least one taxi waiting at the station.”

The parable was used by Dennis Robertson in his book Money from 1922 to describe the peculiarly inflexible U.S. reserve requirement system:

…it resembles the procedure of a certain municipality which tried to guard against a shortage of cabs by ordaining that there should always be at least one cab on the ranks.[1]

But 100 years later, the parable better illustrates the problem of requiring banks to always maintain a specific quantity of liquidity reserves (“High Quality Liquid Assets” or “HQLA”) to meet contingencies.[2]  Consistent with the parable, in response to interviews that BPI conducted last fall (see here), all bank treasurers we spoke with observed that they would not be willing to use the liquidity buffers that regulators mandate that they hold even though the regulation recognizes that use of the buffer may be necessary during periods of liquidity stress. 

The parable may need to be amended, however.  Nearly half of the treasurers indicated that they had raised liquidity during the financial turmoil in March and April 2020 by borrowing at term or selling illiquid assets at a discount specifically to reduce the risk of falling out of compliance with regulatory requirements rather than because they actually needed the funds.  Note that selling illiquid assets at a discount and borrowing at term under stress are precisely the actions that liquidity standards are designed to make less likely. 

A year later, the economist arrives again in the small town by train, except this time it is not only late, but also pouring rain.  Remembering her experience a year earlier, she is relieved to see that there are two taxis waiting at the taxi stand.  When she jumps in the second, the driver says “I’m sorry, I can’t take you.  Recognizing that taxi rides from the station are especially necessary when it is raining, my company decided that it should have two taxis stationed here in such circumstances.”

As discussed in “Unlocking the Liquidity Coverage Ratio,” the reasons why banks won’t use their HQLA are closely related to the reasons why banks won’t borrow from the discount window.   Using HQLA to an extent that would put a bank below a regulatory standard is allowed in principle, but it requires that the bank immediately provide an explanation to examiners.  Similarly, bank treasurers report that borrowing from the discount window is allowed but would likewise require immediate explanation to their examiner (and in some cases, outreach to the examiner in advance of the borrowing).  Moreover, in both cases (using HQLA or borrowing from the discount window), treasurers believed that even if examiners encouraged the bank to take either action during a period of financial stress, examiners would later reference the need to tap an emergency source of liquidity as an indication that the bank’s liquidity management was flawed.

The usability and stigma problems do not just have the same cause, they have the same toxic effect.  As described in “A Major Limit on the Fed’s Crisis Toolkit: Shame,” Pat Parkinson and I explain how stigma associated with the discount window kneecaps the Fed’s foremost tool for addressing financial strains.  Even though banks keep over a trillion dollars in collateral pre-positioned at their Federal Reserve Bank to back discount window loans, all the bank treasurers BPI interviewed stated that they would not borrow from the discount window except to address a temporary payment system disruption – that is, an operational issue.  They pointed to entrenched resistance by senior management at their bank (owing in part to past depictions of borrowing as a “bailout”) as well as examiner disapproval as reasons for their unwillingness to borrow.

Moreover, both the lack of HQLA usability and the stigma associated with borrowing from the discount window boost banks’ demand for reserve balances (bank deposits at the Fed), requiring the Fed to be larger than it would be otherwise.  During the repo turmoil in September 2019, repo rates spiked to 10 percent, even as banks held massive stockpiles of reserve balances.  Banks increased lending into the repo market but were quite limited in their appetite for lending out that cash in exchange for Treasuries.  Banks provided two main explanations for their reluctance:  1) They were unwilling to fall below regulatory expectations because doing so would have extreme consequences such as triggering a bank’s recovery plan in some instances.  2) They maintained an absolute minimum level of intraday reserve balances designed to ensure they would never have to borrow from the discount window.

It’s ironic that banks would hoard liquidity to prevent having to go to the discount window.  In 2003, the Federal Reserve, in an effort to reduce stigma, revamped the discount window.  The System’s discount window website indicates that banks can borrow for any purpose, and that Reserve Banks generally do not ask banks why they are borrowing.  Among other changes, in SR letter 03-15 the banking agencies praised the discount window as a reliable source of liquidity funding and encouraged banks to incorporate it into their liquidity contingency plans.  Even though the SR letter is still in force, banks are in reality discouraged from borrowing or planning on borrowing from the discount window.  A recently released staff memo to the FOMC acknowledged:

Indeed, responses to a survey of depository institutions and bank supervisors conducted by the Discount Window Working Group (DWWG) in 2014 indicated that some banking supervisors (especially of very large institutions) view any borrowing from the central bank as cause for follow-up with the institution that borrowed.[3]

Furthermore, the main bank liquidity requirement, the liquidity coverage ratio, does not recognize the discount window as a potential source of liquidity; nor do bank resolution liquidity requirements, which unbeknownst to most have been the binding constraint for the largest banks.  Indeed, in May 2019 the Fed proposed that foreign banking organizations be subject to toughened liquidity regulations precisely because “analysis using Federal Reserve Board data on [Term Auction Facility (TAF)] usage in 2008 and 2009 finds that approximately 40 percent of foreign banking organizations borrowed from the facility during the financial crisis.”[4]  Doubly ironically, the TAF, through which the Federal Reserve auctioned term discount window loans, was designed to avoid stigma so that banks would be willing to use it.

Consequently, the parable may need to be further amended.

The economist at the train station observed to the cabbie that the process wasn’t working very well, noting that she was being left without a ride to her hotel on a dark and rainy night.  The cabbie responded that the cab company contributes a couple of cars to a municipal backup maintained for such situations and keeps two drivers on call.  The economist asks whether the cabbie could call one of those backup cabs for her, noting that it would make even more sense to let the two cabs at the station give rides to customers, knowing that the reserve cabs stood ready to take their place.  The cabbie explained that if he did so, the municipality would criticize his company both for leaving the station without cabs and for having to resort to an emergency backup – a clear indication that they weren’t managing their cab resources wisely.

The Fed recently took a step intended to address both the buffers-on-top-of HQLA requirements and the discount window stigma problem.  In July 2021, it opened a new lending facility, the Standing Repo Facility (SRF).  Each afternoon, the New York Fed auctions at a fixed interest rate a large quantity of repo funding against Treasury securities, agency debt and agency MBS (Open Market Operations or “OMO” securities).  Because the interest rate is high, the primary dealers and banks that have access are expected to only borrow in exceptional circumstances, and borrowing has been zero to date.  Because the credit is extended under the Fed’s open market authority (Section 14 of the Federal Reserve Act) rather than the Fed’s discount window authority (Section 10B), and because the collateral is limited to OMO securities, the Fed hopes that the facility will be free of stigma.  Indeed the bank treasurers BPI interviewed judged that there was currently no stigma attached to the facility.

The FOMC discussed a prospective SRF in June 2019.  The minutes of the meeting indicate that the facility could serve two purposes – it could help limit spikes in money market rates, and it could encourage banks to shift the compositions of their liquidity reserves away from reserve balances and toward other assets, enabling the Fed to operate with a smaller balance sheet.  The Committee discussed the SRF again in the April, June and July 2021 FOMC meetings and opened the facility in July.  The minutes to the April 2021 meeting again list as potential advantages of the facility limiting pressures in money markets and reducing the demand for reserves. 

Bank treasurers all judged that for the SRF to remain stigma free, examiners would need to view the repos as a normal course of business borrowing by the bank, not tapping a contingency source of funding.  But, to date, the Federal Reserve has not indicated that banks can anticipate using the SRF freely in their internal liquidity stress tests or their resolution liquidity requirements.  If use of the facility is a normal course of business event, then borrowing from it should be no different from tapping any other source of liquidity except for the fact that the SRF should be completely reliable.  In particular, if the Fed wants banks and primary dealers to extend repo credit willingly during a period of financial turmoil, the institutions need to be sure that they could tap the SRF later in the day without subsequent examiner retribution.  If the Fed wants banks to substitute SRF collateral for reserve balances in their liquidity stockpiles so that it can reduce the size of its balance sheet, banks need to be able to build their liquidity contingency plans around using the facility.

Steps to Reduce Discount Window Stigma and Increase HQLA Usability

Insanity is continuing to do the same thing and expecting different results.  With the discount window, with the LCR and now with the SRF, the banking agencies have announced plans to provide liquidity to banks under stress, only to see their own examiners dissuade banks from ever doing so.  If the agencies were seeking a different result, there is a way forward.

At the cusp of its shift to contracting rather than expanding its balance sheet, there are many steps the Fed could take to address the buffers-on-buffers and stigma problems that could stymie its efforts to get smaller and leave it ill-equipped to address the next September 2019-type event.


  1. According to bank treasurers, the most important step the Fed could take to reduce stigma is for Fed leadership to educate the public and Congress that a loan from the Federal Reserve – SRF loan or discount window loan – is not a bailout.  All loans are fully collateralized and are provided at a market or above-market interest rate.  The Federal Reserve has not made a loss on a regular discount window loan since the 1930s.  Stigmatizing Fed lending means the Fed will have to be bigger and that the Fed will be less able to address financial crises.
  2. To keep the SRF stigma free, use must be seen by bank management, bank investors, bank examiners, the press, Congress and the public as an ordinary business decision, not recourse to an emergency backup.  To build and maintain that image, the Fed must publicly, quickly and repeatedly state that banks are of course not only allowed, they are encouraged to plan on using the facility when under liquidity stress.  Moreover, adjustments to such plans are precisely the mechanism whereby bank demand for reserve balances would be reduced.
  3. The most likely way that using SRF will continue to be seen as a normal-course-of-business event is if borrowing for business purposes is frequent.  Use will only be frequent if the SRF interest rate is only slightly over normal market repo rates, and if the Fed allows some volatility in those rates. But the minimum rate on SRF borrowings has thus far been well above market repo rates.
  4. Unequivocal, written guidance must be provided to examiners stating that borrowing from the SRF or discount window is something to be encouraged, not discouraged, and that they are to take no adverse action against a bank that does so.
  5. Resolution planning guidance should be amended to take into account banks’ ability to use these liquidity facilities.
  6. While the Basel Standard for the LCR requires banks to hold sufficient HQLA to meet cumulative net liquidity outflows at 30 days, the U.S. standard requires banks to have sufficient resources to make it to 30 days – that is, banks have to be able to fill the deepest liquidity hole over the month.  Instead, each bank could be required to hold HQLA equal to net cash outflow at 30 days if it had collateral pledged to the discount window with lendable value greater than its “deepest hole” add-on. The revised requirement would be compliant with the Basel standard, would result in a more accurate assessment of the bank’s liquidity situation, and would provide banks an incentive to be prepared to borrow from the discount window if necessary, enhancing financial stability.  For more information see “Give Banks Credit for Robust Contingent Liquidity Arrangements.”

HQLA Usability

  1. To address the usability problem, bank treasurers recommended that the LCR include both a minimum and a buffer, similar to the existing capital framework.  For example, the requirement would remain 100 percent in ordinary times, but in times of stress there would only be a supervisory response if a bank’s LCR dropped below 80 percent.  The 80-100 percent zone (or some similar zone) would be designed to allow the bank to use some of its liquidity under stress without adverse supervisory consequences including immediate notification or the filing of a remediation plan.
  2. Given the resemblance between reluctance to use HQLA and reluctance to borrow from the discount window, the Fed could review steps it has taken to reduce stigma for application to the buffer usability problem.  Just as the Fed has kept borrowing secret, banks could no longer be required to report publicly their LCR (except to any extent required by the securities laws).  Just as the Fed (in principle) extends discount window loans on a no-questions-asked basis, banks could no longer be required to provide an immediate explanation to their examiner for a small and transitory LCR breach.  Just as the Fed established financial soundness criteria for the discount window, supervisors could allow greater HQLA usability for better capitalized banks.  In general, usability could probably best be achieved if it became a regular and unexceptional occurrence that a bank would temporarily use its HQLA to meet an obligation.  For further discussion, see “Unlocking the Liquidity Coverage Ratio.”
  3. Because of the way the LCR is designed, if a bank uses its HQLA to meet a projected cash outflow, its LCR will decline.  The metric could be revised so that the LCR would remain unchanged in such circumstances, encouraging banks to use their HQLA as intended.  Specifically, the LCR assumes any pool of liabilities has a constant outflow rate.  If a bank has, for example, $100 billion of deposits with an assumed outflow under stress of 5 percent, then if $5 billion in fact, flee, the bank now must project a deposit outflow of $4.75 billion ($95 billion X 5 percent).  Because HQLA would fall by $5 billion but required HQLA would only fall $250 million, the bank would have to raise $4.75 billion, during a period of financial turmoil, to keep its LCR unchanged.  Instead, the LCR could assume that the flighty deposits that were projected to be withdrawn were withdrawn and that the remaining deposits are sticky.  If the LCR were so adjusted, the amount of HQLA that banks would be required to hold would be unchanged in normal times, but the perverse and procyclical result that a bank under stress would have to raise HQLA to remain compliant would be eliminated.  See “A Modest Change to the LCR That Could Substantially Improve Financial Stability.”

A change that could both reduce stigma and encourage usability

Lastly, the Fed could both reduce stigma and increase usability by making a change to both how it provides discount window credit and how it assesses bank liquidity.  Specifically, it could offer a Committed Liquidity Facility (CLF).[5]  The Basel Standard for the LCR includes an option for jurisdictions with insufficient HQLA to count a committed line of credit from the central bank for which the bank provides collateral and pays a fee as HQLA.  Although the standard also allows other jurisdictions to provide CLFs, they must charge an unworkably high fee and the CLF can only make up a small part of each bank’s HQLA, so unsurprisingly, no jurisdictions have elected to do so.  If the Fed were to offer a market-priced CLF, it would raise money for taxpayers by charging banks for an option they now get for free, banks may be more willing to use their CLF than the discount window because they pay for it, regulatory and examiner assessments of a bank’s liquidity condition would be more accurate because they would include the liquidity a bank can rely on from the Fed, and banks should be more willing to use their HQLA because they could boost their LCR by adding to their CLF when needed.  For a more extensive discussion of CLFs, see “Recognizing the value of the central bank as a liquidity backstop.


In discussing this topic with economists and policymakers, it is difficult to find a defense of the status quo.  It is difficult to defend a regime that is producing outcomes precisely the opposite of its intended outcome.  The two main obstacles to reform are inertia – especially at a time when many other rules are under examination – and politics – as there is now a lobby devoted to opposing any reform that could be characterized as making life more efficient for banks.  But the stakes here are very high, and the path very clear.  Otherwise, much of the HQLA currently held by banks will be no more useful than the last taxicab at the train station. 

The status quo should be especially unwelcome to the Fed now.  In a few months the FOMC will begin shrinking its portfolio of securities as part of its effort to pare back the extraordinary amount of monetary stimulus it is providing even when inflation has risen to the highest level in 40 years.  Indeed, some FOMC participants have observed that reducing the balance sheet quickly and substantially would allow them to raise interest rates more slowly, lessening the chance of causing turmoil in financial markets.  The Fed cannot shrink its assets any lower than the amount necessary to satisfy banks’ demand for reserve balances.  If the Fed wants to get leaner, it should take actions immediately to further policy objectives it purportedly supports – encouraging banks to both use their liquidity reserves when under stress and see the discount window and the SRF as viable sources of contingent liquidity.

[1] I thank Charles Goodhart for providing me this reference.  Robertson also observed about U.S. reserve requirements: “An iron ration which you must not touch even in the throes of starvation is something of mockery.”

[2] Goodhart (2008) may be the first use of the taxi parable to illustrate the HQLA usability problem.  Goodhart, C. (2008). Liquidity risk management. Banque de France Financial Stability Review, 11, 39-44.

[3] Standing Lending Facilities,” October 14, 2016, p. 7.


[5] See “The Liquidity Coverage Ratio and Restricted-Use Committed Liquidity Facilities,” Basel Committee on Banking Supervision, 12 January 2014.