To read the first post in this series, click here.
The Federal Reserve is heading toward a smaller balance sheet. Chairman Warsh made his view plain before taking office: “The Fed’s bloated balance sheet, designed to support the biggest firms in a bygone crisis era, can be reduced significantly” (Warsh 2025). In June, he charged a task force with reviewing “the benefits and risks of the current ample-reserves regime and the composition of the Fed’s balance sheet” and assessing “alternative frameworks for the conduct and operation of monetary policy,” with a report due by year-end. Former Governor Miran and Board staff have catalogued the changes to liquidity regulation, supervision and Fed operations that could open the door to $1.2 trillion to $2.1 trillion of reduction without abandoning ample reserves (Anderson, Barbarino, Diercks, and Miran 2026).
Because assets must equal liabilities, the Federal Reserve’s balance sheet can get no smaller than the amount needed to supply the reserve balances banks want to hold. But how many reserves banks want depends on what it costs them to hold reserves, and right now it costs them nothing: reserves pay roughly the same as the alternatives. The $3 trillion banks hold today is what they want at a price of zero. The way to find out how much they want at a price above zero is to supply somewhat fewer, until money market rates settle modestly above the interest rate the Fed pays on reserve balances (the IORB rate). That is not starving banks of reserves; it is charging a small price for something that has been free. Once holding reserves has a cost, banks will look for ways to hold fewer, and, over time, bank treasurers, supervisors, and regulators will adjust their practices and rules accordingly. Reforms to liquidity requirements and steps to reduce discount window stigma can speed that adjustment, but the first and necessary step is putting a price on reserves.
The Fed seems increasingly likely to adopt what is sometimes called a demand-driven framework, in which the Fed leaves the banking system short of reserves and banks, in aggregate, borrow from the Fed routinely to make up the difference. The fed funds and repo markets redistribute reserves among banks, but they cannot create reserves; when the system as a whole is short, the marginal dollar comes from additional Fed lending, and the interest rate on those loans, rather than the IORB rate, becomes the anchor for money market rates. A demand-driven framework does not by itself create an opportunity cost for holding reserves. The Bank of England is adopting one in which it lends at the same rate it pays on reserves, so that money market rates settle at or close to the deposit rate. If the Fed wants banks to economize on reserves, it must set its lending rate above the IORB rate, so that money market rates settle above the IORB rate as well. When reserves earn less than the alternatives, banks economize on them over time, and the Fed’s securities portfolio will decline dollar for dollar.
Even skeptics of a smaller balance sheet approach to monetary policy implementation accept the mechanism. Governor Barr, who opposes the objective, observes that “if rates are consistently above the IORB rate, banks have an incentive to economize on reserves” (Barr 2026). The New York Fed’s Roberto Perli, who manages the Fed’s portfolio, describes the same path: if the FOMC simply supplied fewer reserves, “money market rates would move persistently above IORB, and the opportunity costs of holding reserves would increase” (Perli 2026).
But reserve balances are not just a monetary policy instrument. Before 2008, banks held as few reserve balances as they could, using them to meet reserve requirements and settle payments; for contingent liquidity, they relied on their ability to borrow in the federal funds market and under stress from the discount window. Since 2008, bank regulators have made reserve balances the core asset for meeting liquidity regulations, supervisory guidance and examiner expectations. If reserves are no longer abundant, and no longer pay roughly what alternative assets pay, something must take their place as banks’ source of immediate, same-day liquidity.
The choice on offer is usually described as one between two different things: discount window borrowing on the one hand, and holdings of Treasury securities on the other. Described that way, the first sounds like reliance on the central bank and the second like self-insurance. But that characterization is wrong. Under the framework the Fed has in place, the way a bank turns Treasuries into cash late in the day is to pledge them at the discount window and borrow, and the Fed’s own supervisory guidance now assumes that is what banks will do rather than sell or repo the Treasuries in the market (Federal Reserve Board 2024a). Both options are discount window borrowing. The only difference between them is what sits in the collateral pool: loans the bank has made, securities the government has issued or both. Banks can pledge any of these today, subject to haircuts; the question is which of them the liquidity rules will credit.
This note argues that liquidity regulation and examination mandates should recognize borrowing capacity backed by loans, and that a system in which they do is better than one in which the capacity is backed only by Treasuries.
First, that has been the Fed’s own position for as long as it has had a discount window, and it is how the window actually works: a Reserve Bank lends against the whole pool of collateral a bank has pledged, not against one asset or another, so the distinction the liquidity coverage ratio (LCR) and internal liquidity stress test (ILST) requirements draw between monetizing Treasuries at the window and borrowing against loans has no operational meaning.
Second, liquidity rules that recognize loan collateral are better for growth, because they let banks fund their contingent liquidity with the loans they make rather than with claims on the government.
Third, they are better for financial stability, because lending against illiquid loans adds liquidity to the financial system, which is what the Fed was created to do, and because recognizing the full range of collateral will reduce the stigma that has made the window less effective than it should be.
Fourth, they are not materially riskier for the Fed, whose considerable recent losses have come from interest rate risk in its Treasury portfolio not credit risk from lending to banks. The Fed has not lost a dime on a discount window loan in 100 years.
Fifth, liquidity rules that can be satisfied only with reserves and Treasuries are a form of what has been labeled “financial repression,” a policy of forcing a nation’s banks to fund government spending and thus have their fortunes correlated with the nation’s fiscal success – a policy with a poor record wherever it has been tried and a bad look for any government that needs to employ it.
A Smaller Fed Requires a Replacement for Reserve Balances
The size of the Fed’s balance sheet is determined by the liabilities the Fed must supply. Currency and the Treasury’s deposit at the Fed (the Treasury General Account or “TGA”) are outside the Fed’s control; reserve balances are the residual. Reserve balances currently stand at about $3 trillion, roughly 12 percent of commercial bank assets, up from about 8 percent of bank assets in 2019. When the FOMC adopted a floor system in January 2019, staff told the Committee that about $1 trillion would be needed; in October 2019, after repo market volatility, the estimate rose to $1.45 trillion; the Committee’s decision last December to resume asset purchases revealed that the estimate is now $3 trillion (Nelson 2025c, 2026a). The estimate has never gone down. Each time the Fed floods the system with reserves, banks build them into their liquidity plans, regulators write rules around them and examiners come to expect them; because reserves are riskless, perfectly liquid and pay about what the alternatives pay, nothing pushes back. Demand ratchets upward.
The role of regulation in that ratchet was foreseeable. A note published in September 2016 by The Clearing House warned that liquidity and leverage regulation, together with the structure of money markets, would determine the Fed’s balance sheet in the longer run, and that regulations built around reserve balances would lock in a large Fed (Nelson 2016). That is precisely what happened. In BPI’s January 2026 survey of bank treasurers, banks pointed to their own liquidity risk assessments and internal liquidity stress tests as the main determinants of their reserve demand and identified recognition of discount window capacity in liquidity regulations as the single change that would do most to reduce it (Bristow, Nelson and Waxman 2026).
The reason is simple. There are only two guaranteed ways for a bank to meet a large same-day demand for cash late in the business day: reserve balances, and collateral prepositioned at the discount window. Private repo markets do most of their business early in the morning; Federal Home Loan Banks sometimes make advances later in the day but are limited by the funds that they can raise in the market; and a Treasury security sold today settles tomorrow. The Fed’s standing repo facility, which operates in the morning and again at 1:30 p.m., is itself borrowing from the Fed against Treasuries. A bank that experiences a run in the afternoon can tap reserve balances and the discount window, but has little else it can count on. A framework that recognizes only the first will necessarily produce a demand for reserves that the Fed must meet with a correspondingly large balance sheet.
Two Roads That End at the Same Window
If the Fed supplies fewer reserves, so that money market rates settle above the IORB rate, banks will economize on reserves and demand less. What they replace them with depends on the incentives provided by the liquidity framework and supervision. Two versions are on the table.
Treasuries, monetized at the discount window. Under current rules, reserve balances and Treasury securities are treated identically in the liquidity coverage ratio, and Treasuries are the dominant alternative to reserves in banks’ internal liquidity stress tests. If nothing else changes, banks will convert reserves into Treasury bills and notes. But a security is not cash, and a Treasury security sold today does not settle until tomorrow. To meet a same-day outflow, a bank must borrow against it, and once the morning repo market has passed there is one reliable way to do that: pledge the security to its Reserve Bank and borrow, either at the discount window or, during its brief afternoon operation, at the standing repo facility. The Fed’s August 2024 FAQ on Regulation YY says exactly this. Banks may assume in their internal stress tests at the overnight and 30-day horizon that they can use the discount window to monetize Treasuries, agency debt and agency MBS; they may not assume a cash inflow from borrowing against any other prepositioned collateral (Federal Reserve Board 2024a).[1] Vice Chair Jefferson described the same arrangement at the Treasury Market Conference on Sept. 22. Institutions holding Treasuries in their Fedwire Securities accounts, he noted, “can quickly transfer them to their discount window pledge account even late in the day,” and “the discount window’s acceptance of Treasury securities as collateral with same-day value helps make this stability possible” (Jefferson 2026). Under this version the Fed gets smaller, the roughly $4 trillion of high-quality liquid assets that large banks hold against the LCR migrates from one claim on the government to another, and banks’ same-day liquidity under stress is a discount window loan secured by Treasury securities.
Borrowing capacity against eligible collateral of the bank’s choosing. The alternative is to recognize in liquidity regulation a bank’s capacity to borrow from the discount window against whatever eligible collateral it has prepositioned, valued at the Fed’s standard haircuts. That is what Treasury Secretary Bessent proposed in March. The LCR and other liquidity rules, he said, “should give appropriate capped recognition of borrowing capacity associated with collateral prepositioned at the discount window,” a reform that “simply recognizes that prepositioned, collateralized borrowing capacity is real, monetizable liquidity.” Vice Chair for Supervision Bowman, speaking at the same roundtable, made the complementary point that the current framework leads banks to “over-allocate to HQLAs because they must demonstrate that liquidity needs can be met with their own balance sheet resources,” which “reduces a bank’s capacity to lend and support its communities” (Bowman 2026). Under this version the Fed also gets smaller, and banks’ same-day liquidity under stress is, again, a discount window loan. The bank simply gets to decide what secures the loan from its pool of eligible discount window collateral.
Set side by side, the two versions are the same transaction. In each, the bank calls its Reserve Bank, or logs in to Discount Window Direct, and takes primary credit at the primary credit rate with no questions asked, for up to 90 days, renewable on request. It is an ordinary, business-as-usual loan, the kind the Fed has been telling banks for more than 20 years it wants them to use. The regulators have already accepted that a discount window loan is what stands behind a bank’s securities in a stress; the Regulation YY guidance, which lets banks assume they will monetize their securities at the window, would make no sense otherwise. The debate about “self-insurance” versus reliance on the central bank is therefore beside the point.
Both roads run through the discount window. The only question left open is what collateral the Reserve Bank should hold.
The usual objection to recognizing borrowing capacity, that discount window access is not certain, applies with equal force to both versions, since both depend on the same loan from the same Reserve Bank. A companion BPI note takes the objection seriously in any case, tests it against the Fed’s statements and actions over two decades and against the certainty of every other source of funds the LCR already recognizes and concludes that primary credit is at least as reliable as the cash inflows, monetized securities, reserve balances and committed central bank facilities the rule already counts (Nelson 2026b).
One Loan, One Pool of Collateral
There is a further reason the distinction between monetizing Treasuries and borrowing against loans cannot bear the weight the rules place on it: it does not correspond to anything that happens at the discount window. Banks do not pledge a particular bond to secure a particular loan. Under the Reserve Banks’ lending agreement, Operating Circular 10, an institution pledges collateral into a single account, and everything in that account secures every obligation the institution has to its Reserve Bank (Federal Reserve Banks 2025). The Reserve Bank computes the lendable value of the pool, after haircuts, and the bank may borrow up to that amount. When a bank borrows, the loan is secured by the pool, collectively. The Fed’s own disclosures reflect this: for each discount window loan it publishes, the Fed reports the lendable value of all the collateral the borrower had pledged at the time, by type, not a list of assets assigned to that loan (Plassman and Rosa 2023).[2]
Consider a bank that has both loans and Treasuries pledged to the discount window, each with a lendable value of $2 billion after haircuts, and that borrows $1 billion during a run. There is no fact of the matter about whether it has “monetized its Treasuries” or “borrowed against its loans.” It has borrowed against everything. If, the next morning, it sells the Treasuries and the Reserve Bank releases them from the account, the discount window loan continues, now secured by the loans alone. The Reserve Bank is indifferent, because the pool of available collateral never fell below the outstanding value of the loan. Yet under the rules as written, the bank was initially exercising a recognized liquidity option, but after selling its Treasuries it was doing something the rules refuse to count.
The point is not merely a technicality. It means that the question the liquidity framework should ask is the one the Reserve Bank already answers every day: what is the lendable value of this institution’s collateral pool, net of what it has already borrowed? That number is capacity, computed by the Fed using the Fed’s prescribed haircuts and available until the close of Fedwire. Whether the pool is Treasuries, consumer loans, commercial loans or a mix is a question about the pool, not about whether the capacity exists. And on the question of what banks should pledge to the discount window, the Fed has had a settled view for a very long time: loans.
The Fed Has Always Said the Answer Is Loans
The Federal Reserve was created to lend against loans. The Federal Reserve Act of 1913 states its purpose in its preamble: “to furnish an elastic currency, [and] to afford means of rediscounting commercial paper,” where commercial paper meant a bank’s short-term loans to businesses (Nelson 2023a). Advances secured by Treasury securities were added by amendment in 1916, and during the banking contraction of the early 1930s, when the supply of eligible paper collapsed, Congress broadened the collateral the Reserve Banks could accept to any asset satisfactory to the lending Bank (Mengle 1986). In the early 1920s, more than four-fifths of the Fed’s assets were discount window loans, and when the Fed confronted the prospect in 2001 that the federal debt might disappear, its study of alternative instruments concluded that such loans could once again become its principal asset (Federal Reserve System Study Group 2002).
The practice has never changed. The Fed today accepts as collateral essentially any bank asset that meets regulatory standards for sound asset quality, and most of what banks pledge is loans. At the end of 2023, the last year for which the Fed has published figures, banks and credit unions had collateral with a lendable value of $2.76 trillion prepositioned at the discount window, of which $1.81 trillion, about two-thirds, was loans and $950 billion was securities (Federal Reserve Board 2024b). Board economists have documented that when institutions pre-pledge, “they most often pre-pledge their less liquid loan collateral,” and that having done so materially raises the probability that they actually borrow when they suffer a liquidity shock (Carlson and Styczynski 2025).
Fed officials have been explicit that this is the right collateral. Speaking as Vice Chair for Supervision in December 2023, Michael Barr acknowledged that less liquid collateral takes longer to value and pledge, and then drew the conclusion: “But these are the very assets that would be best pledged to the discount window, so that more liquid assets are held for other uses” (Barr 2023). In May of this year, now as a governor arguing against a smaller balance sheet, he nonetheless reported approvingly that “after the events [of 2023], we saw banks increase their collateral pledged at the discount window by over $1 trillion,” adding that “this improved level of operational readiness will be important to retain as memories of this episode fade, which history shows tends to happen” (Barr 2026).[3] Vice Chair for Supervision Bowman observed in March that banks “have strong incentives to convert less liquid assets, like loans, into cash to meet withdrawal and repayment demands,” and that the discount window accepts “an even broader range of collateral than the FHLBs” (Bowman 2026). And Vice Chair Jefferson reported on September 22 that the Reserve Banks have just simplified and streamlined the borrower-in-custody program precisely so that institutions “can pledge several different loan types to maximize liquidity while maintaining possession of the collateral” (Jefferson 2026; Federal Reserve Discount Window 2026a). The Fed has spent a century building, and the past three years modernizing, a facility designed to turn loans into reserves. It would be odd, having done so, to maintain liquidity rules under which these pre-pledged loans do not count.
Loan Collateral is Better for Growth
Consider what a bank is doing when it holds an asset to satisfy a liquidity requirement. If it holds reserve balances, it is lending to the Federal Reserve. If it holds Treasury securities, it is lending to the Treasury. Either way, a dollar of the bank’s balance sheet devoted to meeting the requirement is a dollar not lent to a household or a business. Secretary Bessent put the cost in perspective: about 25 percent of large banks’ balance sheets is now allocated to safe assets, up from roughly 10 percent before the crisis, and “that necessarily means less lending for mortgages, small businesses, AI hyperscalers, and critical infrastructure” (Bessent 2026).
If instead the bank meets the requirement with borrowing capacity against prepositioned loans, the same dollar supports a loan to Main Street, and the loan itself becomes the bank’s contingent liquidity. The bank does not have to choose between lending and being liquid; the lending is what makes it liquid.
Loan Collateral Is Better for Financial Stability
The financial stability case for loan collateral rests on two points, both of which go back to why the Fed exists.
Lending against loans creates liquidity; lending against Treasuries rearranges it. A loan to a business or a household is an illiquid asset. When the Fed stands ready to lend against it, the loan becomes a source of cash on demand for the bank that holds it. The bank is more liquid, the system is more liquid, and nothing has been sold. That transformation is the reason the Federal Reserve was founded. Paul Warburg, one of the System’s architects and its second vice chairman, described the Reserve Banks’ purpose in 1916 as “to provide the machinery for turning, whenever desired, the member banks’ commercial assets into available credit balances, or cash” (Warburg 1916). Charles Hamlin, the first chairman of the Federal Reserve Board, explained in 1914 that “the turning of commercial paper into a liquid investment, will enable every bank to draw down its reserves with confidence that it can replace them at will if it has proper commercial paper at its disposal” (Hamlin 1914). The founders did not regard this as a bailout or as emergency lending; they regarded it as the ordinary business of a central bank, the thing that let banks lend to their communities without hoarding cash against a run (Nelson 2023a).
A Treasury security is the opposite case. It is already the most liquid asset in the world. Pledging it to a Reserve Bank does not make it more liquid; it makes it less useful, because a Treasury security in a dealer’s inventory supports market making, repo lending, and the settlement of trades, and serves as a close substitute for cash. Pledging a liquid asset to borrow another liquid asset creates no liquidity; it only takes Treasuries out of a market that uses them (Singh and Stella 2012; Singh 2013). While lending against loans adds liquidity to the financial system, lending against Treasuries moves it from one place to another, and under stress moves it out of the market at the moment the market needs it most.
Recognizing the full range of collateral will reduce stigma and make the window more effective. The 2003 redesign of the discount window that created primary credit was undertaken specifically to reduce stigma so that the window could function as a monetary policy tool and a reliable source of liquidity (Madigan and Nelson 2002). It has not succeeded, and a principal reason is that regulation and supervision have continually sent the opposite message. As Nelson and Parkinson (2022) document, examiners have treated borrowing as a mark against a bank, liquidity rules have been written as if borrowing were a failure of the bank’s risk management, and banks have concluded that the safest course is never to borrow at all. The result is a facility that is most stigmatized precisely when it is most needed. Recognizing borrowing capacity against the full range of collateral a bank has prepositioned reverses that message. A bank that borrows against its loans under a rule that expects it to do so is not signaling distress; it is doing what the rule was written to encourage. And a bank that has prepositioned its loans in order to have that capacity recognized has, as a by-product, done exactly what the Fed has asked of every bank since 2023: put its collateral in the right place and test its access. Borrowing that is expected, and preparedness that is rewarded, are the conditions under which stigma fades, increasing the resilience of the financial system. The window becomes more effective for every bank, whatever its collateral, because the act of borrowing stops carrying a signal.
Is Lending Against Loans Riskier for the Fed?
The natural question is whether taking loan collateral exposes the Fed, and therefore taxpayers, to more risk than taking Treasuries. The structure of discount window lending makes the answer no.
Discount window loans are extended with full recourse to the borrowing bank; the collateral is protection on top of the bank’s general obligation (Federal Reserve Board 2012). Collateral is valued at fair value and subjected to conservative haircuts, which are substantially larger for loans than for securities (Federal Reserve Discount Window 2026b). Primary credit is available only to institutions that are at least adequately capitalized and CAMELS-3 rated, and the FDIC Improvement Act of 1991 limits lending to undercapitalized and critically undercapitalized banks. Discount window loans are short, with a maximum maturity of 90 days, and carry a floating rate, so the Fed bears essentially no interest rate risk on them. A Reserve Bank lending against a haircut pool of loans to a solvent, supervised bank with recourse is about as well protected as a creditor can be. Indeed, over the maximum horizon that matters for a discount window loan, a few months, loans are the safer collateral. Suppose a bank pledges only long-dated Treasuries and borrows the full lendable value. If yields rise the next day, the collateral, which the Reserve Bank marks to market daily, may no longer cover the loan, and the bank must post more collateral or repay the loan. A pool of performing loans valued at conservative haircuts will not change value so quickly.
The record confirms it. The Fed has not lost money on a discount window loan in a hundred years. Every loan the Fed made (including discount window loans) during the global financial crisis was repaid in full, on time, with interest. Silvergate Bank, which lost nearly 70 percent of its deposits in the fourth quarter of 2022, borrowed $4.5 billion from the San Francisco Fed, repaid it within a month, and wound itself down without cost to the FDIC; the discount window worked exactly as intended (Nelson 2025a). The primary credit extended to First Republic in 2023 was likewise repaid in full (Congressional Research Service 2024).
Now consider the alternative collateral. The Fed’s own losses have come entirely from Treasury securities and agency MBS. Beginning in September 2022, the interest the Fed paid on reserves and reverse repos exceeded the interest it earned on its portfolio; cumulative operating losses, booked as a deferred asset, peaked in January 2026 at more than $240 billion. Unrealized losses on the securities portfolio stood at $878 billion at the end of the second quarter of this year (Federal Reserve Board 2026). Banks fared no better on the same assets: unrealized losses on the securities held by FDIC-insured institutions reached $690 billion in the third quarter of 2022 (FDIC 2022), and it was those losses, not credit losses on loans, that set off the runs of March 2023. The riskiest asset on the Fed’s balance sheet has been the one the liquidity framework treats as riskless. Discount window loans backed by haircut loan collateral have not cost the Fed anything.
Stigma Will Diminish, Because the New Regime Requires Borrowing
A remaining objection is that recognizing discount window capacity is pointless if banks will not borrow. Stigma is real and old, rooted in the failures of habitual borrowers in the 1920s and in a century of mixed messages from the Fed about whether it wants banks to borrow (Nelson 2021). As John Williams, then president of the San Francisco Fed, told the FOMC in 2016: “It’s not about who’s causing the stigma. It’s us, right? We’re the central bank, and the policies that we set and the decisions we make are what cause banks not to want to come to the discount window” (FOMC 2016).
Two features of the regime now being contemplated will reduce stigma more than any communications campaign could. First, when money market rates sit persistently above the IORB rate, banks will borrow from the Fed as a matter of course, because it will be cheaper than the alternatives on the days when reserves are tight. That is the design. Vice Chair Jefferson noted on Sept. 22 that the Fed has “recently seen more use of the discount window amid periods associated with temporary upward pressure on money market rates, such as those observed at quarter-ends” (Jefferson 2026). Governor Barr noted that if the Fed’s ceiling tools are more frequently in the money and firms understand they are expected to tap them, “these tools would not atrophy from lack of use” (Barr 2026). Borrowing that is common and unremarkable is not stigmatized. Dealers behaved the same way after September 2019: when the Fed’s repo operations were priced close to market rates, they borrowed routinely and on-lent the proceeds (Carlson, Saravay, and Tian 2025).
Second, recognizing discount window capacity in liquidity regulation ends the contradiction at the heart of stigma. For more than 20 years the Fed has told banks that primary credit is available no questions asked, for up to 90 days, renewable on request, and that examiners should regard its use as “appropriate and unexceptional” (Federal Reserve Board 2003). At the same time, its liquidity rules have been written as if borrowing were a failure of risk management. Secretary Bessent identified the resulting irony: “by driving banks to exhaust regulatory buffers before accessing the discount window, we have entrenched discount window stigma. If you only go to the window when things are really bad, then going to the window signals that things are really bad” (Bessent 2026). The Bank of England understood this when it introduced its short-term repo facility in 2022 and had the Prudential Regulation Authority announce simultaneously that it “would judge use of the STR as routine participation in sterling money markets” (Bank of England 2022). The Fed and the other prudential banking agencies can do the same.
Liquidity Regulation that Counts only Treasuries is Financial Repression
There is one argument for insisting on Treasury collateral that is rarely stated outright but is never far from the surface: it maintains demand for Treasury securities. As the Fed sheds Treasuries, banks would be required to absorb them. Governor Barr comes closest to saying so. Counting loan collateral in the LCR, he argues, would lead banks to substitute similar-duration assets for Treasuries, so that the proposal “would lead to reduced demand for Treasury securities rather than reduced demand for reserves” (Barr 2026, note 18). While Barr’s comments are offered as an objection to the proposal, they are better read as an admission.
Designing regulations so that banks must fund the government has a name. Economists have called it financial repression since McKinnon (1973) and Shaw (1973) described it, and Reinhart and Sbrancia (2011) catalogued its use by advanced economies after the Second World War to liquidate their war debts: captive domestic institutions required to hold government paper at yields below what a free market would demand. The practice is alive. India’s statutory liquidity ratio still requires banks to hold a fixed share of their liabilities in government securities. Turkey in 2022 and 2023 imposed “securities maintenance” rules forcing banks to hold lira government bonds against deposits and loans. Argentina’s banks have for years held a large share of their assets in central bank and Treasury paper. And in the European sovereign debt crisis, Greek and other periphery banks were leaned on to absorb their governments’ debt, with consequences that became apparent when that debt was restructured.
The consensus of the empirical literature is that repression crowds out private lending. Becker and Ivashina (2018), studying the euro area between 2007 and 2013, found that the doubling of domestic banks’ government bond holdings “generated a crowding out of corporate lending,” with firms less likely to obtain loans where local banks had absorbed more of the sovereign’s debt. The mechanism is not subtle. A bank’s balance sheet is finite; what is devoted to the sovereign is not devoted to anyone else.
The United States does not do this by decree. It does it by writing liquidity rules that can be satisfied only with reserves and Treasuries and then declining to recognize the one central bank facility that would let banks meet those rules with the loans they make. The result is the same: some $4 trillion of large-bank balance sheets committed to claims on the government, and a banking system whose capacity to lend is smaller than it needs to be. If the Treasury wants demand for its debt, the answer is a well-functioning Treasury market, not a regulatory requirement that banks hold it. It is, moreover, hard to see why the Fed should be concerned about preserving bank demand for Treasuries at the very moment it is contemplating shedding a trillion or two of them itself. As Governor Barr rightly says, the maturity and composition of the debt held by the public is the Treasury’s business, not the Fed’s (Barr 2026).
Conclusion
A smaller Fed means fewer reserve balances, and fewer reserve balances mean that something else must serve as banks’ source of immediate liquidity. Whatever the specifics of the changes to the liquidity rules, that something will be the capacity to borrow from the Federal Reserve: under the Fed’s own guidance, Treasuries become same-day cash by being pledged at the discount window and borrowed against, just as loans do. The choice before the Fed and the banking agencies is therefore not whether banks may count on the discount window. That has been decided. The choice is what the Reserve Banks should hold as collateral, and, since a Reserve Bank lends against a bank’s collateral pool as a whole, the choice is really about what banks are permitted to put in the pool and have it count.
If the rules continue to recognize only reserves and Treasuries, banks will replace one claim on the government with another, the Treasury market will bear the full weight of every bank’s contingency plan, and the balance sheet capacity freed by a smaller Fed will go to Washington rather than to Main Street. If instead the rules recognize borrowing capacity against prepositioned collateral of the bank’s choosing, haircut as the Fed has always haircut it, banks can fund their contingent liquidity with the loans they make, the Fed can shrink, and stigma will recede because borrowing becomes what the rules expect. The banking agencies should make that change whatever implementation regime the Fed ultimately adopts, because it makes the regulations more accurate, gives banks a reason to be prepared to borrow, supports growth and reduces stigma. A smaller Fed only makes it more urgent.
The Fed has said for a century that loans are the best collateral for the discount window. It has spent the last three years making it easier to pledge them. It has never lost money lending against them, while it and the banks it supervises have lost hundreds of billions of dollars on Treasury securities. The choice should not be difficult.
[1] Governor Barr made the same point in May: “right now banks can use term primary credit to improve their LCR, if needed,” and “HQLA securities pre-positioned at the discount window fully count toward meeting liquidity requirements” (Barr 2026).
[2] As Plassman and Rosa (2023) explain, a bank that borrows $1 million on a test loan while maintaining $2 billion of pledged collateral appears in the Fed’s loan-level data as having $2 billion of collateral behind the loan.
[3] In the same speech, Governor Barr opposed allowing banks to count non-HQLA collateral pledged to the discount window toward the LCR, on the ground that doing so would reduce self-insurance. As discussed above, the self-insurance framing does not survive the Fed’s own guidance that banks may count Treasuries monetized at the discount window: in both cases the bank is borrowing from the Fed against its collateral pool, and the only difference is what is in the pool. The objection is also addressed in Nelson (2026b) and in the discussion of financial repression below.
References
Anderson, Alyssa G., Alessandro Barbarino, Anthony M. Diercks, and Stephen Miran. 2026. “A User’s Guide to Reducing the Federal Reserve’s Balance Sheet.” Finance and Economics Discussion Series 2026-019. Board of Governors of the Federal Reserve System, March. https://doi.org/10.17016/FEDS.2026.019.
Bank of England. 2022. “Short Term Repo Facility – Provisional Market Notice 4 August 2022.” August 4. https://www.bankofengland.co.uk/markets/market-notices/2022/august/short-term-repo-facility-provisional-market-notice-4-august-2022.
Barr, Michael S. 2023. “The Importance of Effective Liquidity Risk Management.” Speech at the ECB Forum on Banking Supervision, Frankfurt, December 1. https://www.federalreserve.gov/newsevents/speech/barr20231201a.htm.
Barr, Michael S. 2026. “Efficient and Effective Central Banking: Beyond the Balance Sheet.” Speech at the Money Marketeers of New York University, May 14. https://www.federalreserve.gov/newsevents/speech/barr20260514a.htm.
Becker, Bo, and Victoria Ivashina. 2018. “Financial Repression in the European Sovereign Debt Crisis.” Review of Finance 22 (1): 83–115.
Bessent, Scott. 2026. “Remarks: A Reset on Liquidity Regulation.” As delivered by Under Secretary for Domestic Finance Jonathan McKernan at the Roundtable on Liquidity and Lender of Last Resort, Committee on Capital Markets Regulation, March 3. U.S. Department of the Treasury. https://home.treasury.gov/news/press-releases/sb0412.
Bowman, Michelle W. 2026. “Liquidity Resiliency, Financial Stability, and the Role of the Federal Reserve.” Speech at the Roundtable on Liquidity and Lender of Last Resort, Committee on Capital Markets Regulation, March 3. https://www.federalreserve.gov/newsevents/speech/bowman20260303a.htm.
Bristow, Laurence, Bill Nelson and Brett Waxman (2026). “What Determines Banks’ Demand for Reserve Balances?” Bank Policy Institute, January 14. https://bpi.com/what-determines-banks-demand-for-reserve-balances/
Carlson, Mark, and Mary-Frances Styczynski. 2025. “Pre-Pledged Collateral and Likelihood of Discount Window Use.” FEDS Notes. Board of Governors of the Federal Reserve System, August 29. https://www.federalreserve.gov/econres/notes/feds-notes/pre-pledged-collateral-and-likelihood-of-discount-window-use-20250829.html.
Carlson, Mark, Zack Saravay, and Mary Tian. 2025. “Fed Repo Operations and Dealer Intermediation.” Finance and Economics Discussion Series 2025-052. Board of Governors of the Federal Reserve System, July. https://doi.org/10.17016/FEDS.2025.052.
Congressional Research Service. 2024. “Federal Reserve’s Discount Window: Policy Issues.” In Focus IF12655. https://www.congress.gov/crs-product/IF12655.
Federal Deposit Insurance Corporation (FDIC). 2022. Quarterly Banking Profile, Third Quarter 2022. https://www.fdic.gov/analysis/quarterly-banking-profile/qbp/2022sep/qbp.pdf.
Federal Open Market Committee (FOMC). 2016. Transcript of the Meeting of July 26–27, 2016. Board of Governors of the Federal Reserve System.
Federal Reserve Banks. n.d. Operating Circular No. 10: Lending. https://www.frbdiscountwindow.org/pages/agreements/oc10_agreements.
Federal Reserve Board. 2003. “Interagency Advisory on the Use of the Federal Reserve’s Primary Credit Program in Effective Liquidity Management.” SR 03-15, July 23.
Federal Reserve Board. 2012. “Discount Window Lending.” Credit and Liquidity Programs and the Balance Sheet. https://www.federalreserve.gov/monetarypolicy/clbs-lending-facilities-to-support-overall-market-liquidity-201206.htm.
Federal Reserve Board. 2024a. “Regulation YY Frequently Asked Questions: Internal Liquidity Stress Testing and Buffer Requirements,” Q&A 1. Posted August 13. https://www.federalreserve.gov/supervisionreg/legalinterpretations/reg-yy-frequently-asked-questions.htm#64.
Federal Reserve Board. 2024b. “Discount Window Readiness.” https://www.federalreserve.gov/monetarypolicy/discount-window-readiness.htm.
Federal Reserve Board. 2026. Federal Reserve Banks Combined Quarterly Financial Report (Unaudited), June 30, 2026. August 28. https://www.federalreserve.gov/aboutthefed/files/quarterly-report-20260828.pdf.
Federal Reserve Discount Window. 2026a. “The Federal Reserve Implements Improvements and Streamlines the Borrower-in-Custody Program.” September 8. https://www.frbdiscountwindow.org/.
Federal Reserve Discount Window. 2026b. “Collateral Margins Table.” https://www.frbdiscountwindow.org/pages/collateral/collateral_valuation.
Federal Reserve System Study Group on Alternative Instruments for System Operations. 2002. Alternative Instruments for Open Market and Discount Window Operations. Board of Governors of the Federal Reserve System, December. https://www.federalreserve.gov/boarddocs/surveys/soma/alt_instrmnts.pdf.
Hamlin, Charles S. 1914. “The Federal Reserve System as Established and in Operation.” Address before the New York Chamber of Commerce, December 3.
Jefferson, Philip N. 2026. “Discount Window Modernization and Treasury Market Functioning.” Speech at the 2026 U.S. Treasury Market Conference, Federal Reserve Bank of New York, September 22. https://www.federalreserve.gov/newsevents/speech/jefferson20260922a.htm.
Madigan, Brian F., and William R. Nelson. 2002. “Proposed Revision to the Federal Reserve’s Discount Window Lending Programs.” Federal Reserve Bulletin 88 (July): 313–319. https://www.federalreserve.gov/pubs/bulletin/2002/0702lead.pdf.
McKinnon, Ronald I. 1973. Money and Capital in Economic Development. Washington: Brookings Institution.
Mengle, David L. 1986. “The Discount Window.” Federal Reserve Bank of Richmond Economic Review 72 (3): 2–10.
Nelson, William. 2016. “Liquidity and Leverage Regulation, Money Market Structure, and the Federal Reserve’s Monetary Policy Framework in the Longer Run.” The Clearing House, September 21. https://www.theclearinghouse.org/-/media/Files/Association-Documents-2/20160921_tch_note_liquidity_and_leverage_regulation.pdf.
Nelson, Bill. 2021. “Discount Window Stigma: We Have Met the Enemy, and He Is Us.” Bank Policy Institute, August 10. https://bpi.com/wp-content/uploads/2021/08/Discount-Window-Stigma-We-Have-Met-the-Enemy-and-He-Is-Us.pdf.
Nelson, Bill. 2023a. “The Middle Course: What Fed History Teaches Us About Liquidity Requirements.” Bank Policy Institute, August 22. https://bpi.com/wp-content/uploads/2023/09/The-middle-course.-What-Fed-history-teaches-us-about-liquidity-requirements.pdf.
Nelson, Bill. 2025a. “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.
Nelson, Bill. 2025c. “Forward Guidance: The Fed Needs to Be Seen to Be Working Hard to Get Smaller.” December 22. Available on LinkedIn.
Nelson, Bill. 2026b. “Banks Can Count on the Discount Window…if Regulators Will Let Them.” Bank Policy Institute, September 29. https://bpi.com/banks-can-count-on-the-discount-window-if-regulators-will-let-them/.
Nelson, Bill, and Pat Parkinson. 2022. “A Major Limit on the Fed’s Crisis Toolkit: Shame.” Bank Policy Institute, January 25. https://bpi.com/wp-content/uploads/2022/01/A-Major-Limit-on-the-Feds-Crisis-Toolkit-Shame.pdf.
Perli, Roberto. 2026. “Reserve Management and the Fed’s System Open Market Account: Recent Experience and Insights from Surveys.” Federal Reserve Bank of New York, May 19. https://www.newyorkfed.org/newsevents/speeches/2026/per260519.
Plassman, Laura Suhr, and Felipe Garcia Rosa. 2023. “Statistics on Collateral Pledged to the Discount Window.” Bank Policy Institute, November.
Reinhart, Carmen M., and M. Belen Sbrancia. 2011. “The Liquidation of Government Debt.” NBER Working Paper 16893, March.
Shaw, Edward S. 1973. Financial Deepening in Economic Development. New York: Oxford University Press.
Singh, Manmohan. 2013. “Collateral and Monetary Policy.” IMF Working Paper 13/186. Washington: International Monetary Fund, August.
Singh, Manmohan, and Peter Stella. 2012. “Money and Collateral.” IMF Working Paper 12/95. Washington: International Monetary Fund, April.
Warburg, Paul M. 1916. “The Reserve Problem and the Future of the Federal Reserve System.” Address before the Convention of the American Bankers Association, Kansas City, September 29.
Warsh, Kevin. 2025. “The Federal Reserve’s Broken Leadership.” Wall Street Journal, November 16.
