As prepared for delivery:

I joined the banking section at the Federal Reserve Board in 1993 right out of grad school where I had studied finance and game theory but not banking. To catch up, my boss sent me to examiner school at the then-new but now notorious FDIC Seidman Center. Among many other useful things, I learned how to evaluate a bank's liquidity. I was taught that the fundamental principle was that a bank needed to have well-diversified and reliable regular and contingency funding sources. Relying on assets as a source of liquidity was viewed negatively and primarily something that smaller banks did. It was viewed negatively because the assets could be difficult to liquidate and had thin profit margins.

As discussed in the 1971 Reappraisal of the Federal Reserve Discount Mechanism, even 50 years ago, relying on excess reserves rather than the discount window and lines of credit from correspondent banks to manage liquidity was considered lazy and unhealthy. The Reappraisal was prepared as background to the 1972 revamp of the discount window, an effort to reduce stigma by providing banks a clearer understanding of their borrowing privilege. The revamp maintained the below-market rate on the discount window and limited overuse of the window by administering lending according to several rules, such as "you have to first seek market funding before coming to the window."

As years passed, however, that administration was itself considered a source of stigma and in 2003 the Fed revamped the discount window again. The System removed all rules on use, making the discount window a "no-questions-asked" facility. The Fed wanted banks to use the window freely but not as an ongoing source of funding. To encourage banks to use the window as only a contingency source of funding, the discount rate, renamed the primary credit rate, was raised from 50 basis points below the fed funds rate to 100 basis points above. Because a financially troubled bank wouldn't necessarily be deterred by an above-market rate, and because lending to a troubled institution requires extra care, the System established modest financial soundness criteria for access to credit on a no-questions-asked basis.

After the revamp, I met with bankers and encouraged them to use the new discount window. Among other things, I assured banks that the discount window was an appropriate and indeed ideal component of their liquidity contingency plans. I learned that despite the new discount window policy, bank examiners, including Federal Reserve System examiners, were uncomfortable with banks using or planning on using the discount window. I worked with the banking agencies to prepare official guidance for examiners that explained the new discount window framework and its value as a source of liquidity in contingencies. Despite the guidance, I heard back from banks that examiners continued to view the window negatively.

Even so, earlier this year a bank treasurer with a long memory assured me that the 2003 revisions and the outreach had been slowly working. Banks were becoming accustomed to using the window and including it in their liquidity plans. It was easier back then to argue that borrowing didn't necessarily mean that the bank was under stress because most borrowing by large banks was an unavoidable outcome of monetary policy. When the Fed inadvertently left the banking system short of reserves on any particular day, the fed funds rate would rise up above the discount rate and some banks would borrow, adding reserves and equilibrating the market. Indeed, by 2005 or so, at least one GSIB was regularly borrowing in such circumstances and then lending the funds to other banks that remained reluctant to use the window.

But that progress was more than completely reversed during and after the Global Financial Crisis. Borrowing from the discount window was viewed as having received a bailout; even though the loans were well collateralized, the rate was above market, and the loans were all fully repaid. Bank CEOs were hauled before Congress and excoriated for having borrowed. Whereas before, the Fed kept borrowing information secret, the Dodd-Frank Act required the Fed to publish loan-level discount window information, including the identity of the borrower, with a two-year lag. Despite the lag, bankers tell me that the disclosures are a significant source of stigma. Moreover, examiners were not informed of borrowing unless there was evidence of an underlying problem, something that the System's discount window website says is still true, but now bankers tell me their examiners would be furious if they did not tell them if they were going to borrow.

There was a related change during and after the GFC in how regulators and examiners assess a bank's liquidity. In 2008, the Basel Committee issued principles for sound liquidity risk management and supervision and stated that the new "fundamental principle" was that banks maintain a large cushion of high-quality liquid assets. Newly created liquidity metrics not only did not recognize the discount window as a source of liquidity, the regulations were touted as vehicles to keep banks from borrowing. For example, when the LCR was finalized, Stefan Ingves, the Chair of the Basel Committee, stated that the requirement would "ensure that banks hold sufficient liquid assets to prevent central banks becoming the 'lender of first resort." Stephen Cecchetti, then head of research at the BIS wrote at the time that "[t]he goal [of the LCR] is to ensure that banks can meet their obligations without relying on fire sales of their assets…or borrowing from the central bank." Similarly, the Fed does not allow banks to anticipate using the discount window in their internal liquidity stress tests, the most binding liquidity requirement for most large banks. Moreover, when the Fed contemplated proposing tighter liquidity requirements for U.S. branches of foreign banks in 2019, one of the stated reasons was that the branches had borrowed a lot from the Fed during the crisis, and one of the anticipated benefits was that the branches would borrow less.

The sea change in how a bank's liquidity should be assessed - from having reliable, diversified funding to having a large stockpile of liquid assets - occurred for five reasons. First, interbank funding, especially term funding, had not been reliable in the crisis. Second, the designers of the regulations wanted to decrease rather than increase interconnectedness between financial institutions. Third, as noted, borrowing from the central bank had come to be viewed negatively. Fourth, the designers of the LCR wanted a metric that could be uniformly measured across jurisdictions, something similar to a capital requirement; HQLA divided by projected net cash outflows fit the need while having diversified and reliable funding did not. And fifth, central banks had shifted to monetary policy implementation regimes in which reserve balances were abundant and cheap. Before the crisis, the Fed had not been allowed to pay interest on reserves; after the crisis, the interest rate on reserves was roughly equal to or even a bit above other money market rates.

The shift by the Fed to oversupplying reserves, and the shift in liquidity assessments toward an emphasis on reserves, became a self-reinforcing cycle. With reserves superabundant, trading in the federal funds market withered, reducing the reliability of the market as a source of funding. With banks supersaturated with reserves, discount window borrowing and even daylight overdrafts became rare and seen as a sign that something had gone wrong. Bankers tell me that a key driver for their high demand for reserves is to ensure that they will not run a daylight overdraft or, god forbid, have to borrow from the window.

When reserves are cheap and abundant, banks change their business models to use the reserves and bank examiners change their expectations, both of which lock in the high demand for reserves. For example, the CIO of a GSIB explained that before 2018, when the interest rate on reserves was above the repo rate, the bank shifted to holding reserves equal to 5-days of potential outflows under stress even though they were only required to have 2 days' worth. In 2018, when the repo rate rose above the IORB rate, the bank considered switching to 2-day's supply of reserves using reverse repos to cover the other 3 days, but decided against the change simply because they did not want to have to explain to their examiner why they were reducing their holding of reserves. The consequence of this ratchet effect is the steadily rising estimate of Federal Reserve staff of the quantity of reserves necessary to implement monetary policy using a floor system, shown in the table.

At the end of February 2023, I wrote about this sea change in how bank's liquidity was assessed - from reliable, diversified funding and discount window access to large stockpiles of HQLA with the discount window shunned - and called for a comprehensive review of liquidity regulations including a recognition of discount window borrowing capacity. Two weeks later SVB failed, awash in HQLA but with unreliable, concentrated funding and unable to access the discount window in sufficient size. Worse still, about a year before SVB failed it had been interested in signing up for the Fed's standing repo facility. As we now know, SVB had been failing its internal liquidity stress tests in part because it could not demonstrate the ability to monetize its HQLA. It had hoped that it would be able to point to the SRF as its planned means of monetization. However, when SVB learned that the Fed did not recognize the SRF as a legitimate means of monetization, it did not sign up.

I began my remarks by relating a story from the beginning of my time at the Fed; I will wrap up with a story from just before I left in 2016. At that point I was splitting my time between monetary policy issues including the discount window and bank regulation and supervision. In the latter capacity I learned that examiners were encouraging banks to increase their reserve balances. At the same time, as you likely recall, the FOMC was engaged in QT, reducing reserve balances. I shuttled between the two parts of the Fed, informing each side of the conflict. Vice Chair Quarles later remarked that bank examiners' preference for reserves contributed to the abrupt and disorderly end of that round of QT in September 2019.

The pernicious spiral between reserve supply and liquidity requirements can be reversed. Rather than oversupplying reserves, the Fed can follow the BoE, BoC, ECB, and RBA by reducing the supply of reserves until money market rates rise a bit above the IORB rate and borrowing at the discount window picks up. Of course, at that point Monetary Affairs and the Open Market Desk will have to proactively prevent large, foreseeable, disruptive swings in the supply of reserves, something that they forgot to do in September 2019. The slightly below-market rate on reserves will give banks a steady incentive to find alternatives. The ideal choice would be increased capacity to borrow from the discount window, which, like reserves, provides immediate funding but allows banks to devote more of their balance sheets to lending to businesses and households.

Public link

Please use the above public link if you want to share this noodl on another website.