The Window

On March 3, 2026, Vice Chair for Supervision Michelle Bowman opened a roundtable on bank liquidity at the Committee on Capital Markets Regulation in Washington. Her remarks proposed what she framed as a long-overdue reassessment of the post-crisis liquidity framework — the Liquidity Coverage Ratio, the Net Stable Funding Ratio, and the discount window that sits beneath them both. Her name was not on a nomination list. She was not testifying under oath. She was the person who will write the rules she described.

The speech was precise where policy speeches are often vague. Bowman argued that the LCR has become “an isolated, unusable buffer” — that banks maintain liquid assets above regulatory minimums but refuse to draw them down during stress, creating the fragility the framework was designed to prevent. She argued that “liquidity hoarding” imposes “unnecessary costs on the banking system and the broader U.S. economy,” reducing credit to the communities banks are meant to serve. She argued that the discount window is stigmatized, fragmented across twelve Reserve Banks, and operationally unreliable — and that reforming it would allow the Federal Reserve to “maintain a smaller balance sheet.” Fifteen years after the post-crisis framework was built, she said, it was time to ask whether “compliance actually translates into resilience.” These are testable claims. The Federal Reserve’s own archive — internal deliberations, staff analyses, Tealbook projections, committee transcripts, and public statements spanning 1970 to 2024 — contains the documentary record against which every element of what she described can be measured.

The Single Contradiction

Bowman’s speech rests on a structural proposition: the post-crisis liquidity framework is too costly, and a reformed discount window can substitute for the on-balance-sheet buffers it currently requires. If the discount window can be standardized, destigmatized, and made operationally reliable, then banks no longer need to hoard high-quality liquid assets at the levels the LCR demands. Lending capacity returns to the economy. The Fed’s balance sheet shrinks. The costs disappear.

The contradiction is that Bowman herself has already provided the definitive assessment of this proposition — and it is the opposite of the one her speech implies. On April 3, 2024, eleven months before this roundtable, she said:

In the same address, she concluded that “the Federal Reserve cannot entirely eliminate discount window borrowing stigma through regulatory fiat.” The market’s inference is an external force the Fed cannot control. Administrative reforms — standardization, harmonization, reduced pricing — operate on the supply side of the facility. Stigma operates on the demand side, in the rational calculations of bank treasurers who know that detection carries reputational costs no regulator can offset.

The speech delivered today proposes exactly the kind of regulatory reforms that the speech delivered twenty-three months earlier classified as insufficient. The question is not whether Bowman’s diagnosis is correct. It is whether the remedy she now prescribes can succeed where her own prior analysis said it could not — and where the Federal Reserve’s internal record, across five decades and six distinct crisis episodes, has documented failure after failure after failure.

We went into the archives.

The Backstop That Wasn’t

The discount window’s stigma problem is not a recent discovery. It is the oldest recurring finding in the Federal Reserve’s internal deliberative record on liquidity provision, documented continuously from the Penn Central crisis of 1970 through the bank failures of 2023.

The pattern established itself early. After Continental Illinois collapsed in 1984, Robert P. Black, President of the Federal Reserve Bank of Richmond, observed that Chicago banks “wanted to make darn sure nobody felt they were in the same situation as Continental and they might be more reluctant to borrow.” Peter Sternlight, Manager for Domestic Operations, confirmed that larger banks avoided the window “lest they be tarred with the same brush.” The behavioral consequence was immediate: Frank Morris of the Federal Reserve Bank of Boston reported that banks developed a “propensity to bid up the federal funds rate rather than come into the discount window,” producing interest rates higher than the Committee intended. Stigma was not merely a reputational inconvenience. It was degrading monetary policy transmission in real time.

The Fed attempted to fix the problem in 2003, redesigning the discount window into a “Primary Credit” facility with simplified terms and an explicit “no questions asked” policy. The redesign assumed that administrative changes could override the market’s reputational calculus. By 2007, this assumption had been tested and had failed. Brian Madigan, Director of the Division of Monetary Affairs, identified the “fundamental problem” in an April 2008 deliberation: banks feared their use of the window would be “detected in the market,” and that cost was “just not worth incurring.” The Fed was forced to create an entirely new facility — the Term Auction Facility — not because the discount window lacked liquidity, but because no solvent bank would use it.

Dudley’s assessment — delivered nineteen years before Bowman’s speech — is the analytical verdict the institution has never successfully overturned. The 2003 reforms did not eliminate stigma. The TAF circumvented it temporarily but was terminated in 2010, forcing the system back to the stigmatized primary credit window. When the 2023 bank failures arrived, the discount window failed again, and the Fed was forced to create yet another workaround — the Bank Term Funding Program.

Throughout this history, the institution’s internal analysis and its public communications have told different stories. Internally, staff were blunt: the window does not work when it is needed. But publicly, officials defended the discount window as “largely successful” because loans were “repaid in full, on time, and with interest.” This framing misses the point entirely. The discount window’s failure mode is not default. It is non-use. An emergency facility that no solvent institution will access during an emergency is not successful because the loans it did not make were not defaulted on. As recently as 2024, Governor Philip Jefferson described the 2003 Primary Credit redesign as having been “effective in reducing banks’ reluctance to borrow” — a claim contradicted within months by the 2023 bank failures, in which the window once again proved unusable under stress. The outcome was predictable from the evidence available at the time: thirty years of evidence that banks fear market detection was available when the “no questions asked” policy was implemented. The institution chose not to weigh it.

By 2019, the institution’s self-diagnosis had become explicit. Neel Kashkari, President of the Federal Reserve Bank of Minneapolis, identified the source of the problem not in market psychology but in the Fed’s own institutional posture:

Kashkari’s warning identifies a deeper problem than the one Bowman’s speech addresses. Bowman proposes to standardize the window’s operations — harmonize rules across twelve Reserve Banks, reduce pricing penalties, reform disclosure practices. These are supply-side interventions. Kashkari’s diagnosis is that the demand-side problem is generated by the Fed’s own supervisory culture — the rhetorical framing of the window as emergency infrastructure, the implicit message that any institution that uses it must be in distress. Standardizing procedures does not change the signal. The 2016 staff memo that used the word “plagued” to describe the window’s dysfunction was never reflected in the institution’s public communications. Internally, the language was clinical and blunt. Externally, the facility was described as available and effective. The gap between these two registers is the space in which Bowman’s reform proposal operates — and it is the space in which every prior reform has failed.

If the discount window has never functioned as a reliable backstop — if stigma has survived every reform from 1984 to 2024 — then the question becomes: why was the current liquidity framework built the way it was? The answer is not bureaucratic inertia. The answer is that the architects of the post-crisis regime knew exactly what the archive shows, and they built accordingly.

The Insurance They Chose

Bowman describes the LCR’s liquidity hoarding incentive as an unintended cost — an artifact of compliance that “imposes unnecessary costs on the banking system.” The documentary record shows the opposite. The hoarding incentive was a deliberate design choice, made by officials who had lived through the discount window’s failures and concluded that the only reliable liquidity was liquidity a bank already held on its own balance sheet.

Governor Elizabeth Duke provided the most direct institutional assessment of the discount window’s status during the design phase of the post-crisis framework:

“Broken.” Not inadequate. Not suboptimal. Broken. This was the institutional verdict that authorized the entire LCR project. If the central bank’s own liquidity facility could not be trusted in a crisis, then banks had to be required to hold their own buffers — buffers they controlled, buffers that did not depend on a facility the market would punish them for using.

Governor Jeremy Stein made the design logic explicit three years later, framing the LCR not as an additional regulatory layer but as a structural replacement for the lender of last resort function:

The phrase “not available at all” is the key. Stein was not describing a theoretical possibility. He was describing the observed reality of 2007–2008, when the discount window existed, was funded, was staffed — and was not used, because no rational bank would accept the reputational cost of detection. The LCR was the institution’s insurance policy against its own backstop’s failure.

The costs were not hidden. Staff analysis in December 2010 projected that the LCR would increase borrowing costs for the nonfinancial business sector, “implying a small drag on real GDP.” The Committee reviewed this analysis and proceeded. Governor Daniel Tarullo, the primary architect of the post-crisis supervisory framework, noted that even without regulation, banks tended to “hord liquidity” during instability because they could not rely on the backstop. The LCR did not create the hoarding impulse. It codified an impulse that the discount window’s unreliability had already generated.

In a striking internal moment, some officials went further. During a December 2009 deliberation, President Charles Evans observed that certain officials believed “stigma is our friend” — that the discount window’s dysfunction usefully prevented banks from arbitraging the facility. The institution was not merely aware of the window’s failure. It was, on occasion, exploiting it.

Bowman’s speech frames the LCR’s costs as evidence that the framework needs reform. The archive shows those costs were the price of admission to a system that no longer depended on a facility the Fed’s own governors called “broken.” The question is whether the balance sheet consequences Bowman identifies — the structural enlargement she argues is unnecessary — reflect the same deliberate calculation, or whether the framework’s costs have exceeded the designers’ projections.

The Balance Sheet They Built

Bowman’s most technically precise claim is that the liquidity framework requires the Federal Reserve to “maintain a larger balance sheet” than monetary policy alone would demand. On this point, the archive is unambiguous: she is correct, and the institution has known it since at least 2017.

When the Fed began balance sheet normalization after the post-crisis expansion, staff projected a “baseline level of longer-run reserve balances” of approximately $500 billion — still far above the $10 billion pre-crisis average, but a dramatic reduction from the trillions accumulated during quantitative easing. Within eighteen months, the Tealbook projection had doubled to $1 trillion, after surveys revealed that banks’ “lowest comfortable reserve level” was far higher than models predicted. Kashkari was skeptical of these numbers, noting that when the Fed asked commercial banks how many reserves they needed, “they’re putting up big numbers, and we don’t really know what the actual demand is going to be.” The banks’ answers reflected not just regulatory requirements but supervisor expectations and risk appetite — variables that would not change simply because a rule was rewritten. The LCR, internal stress metrics, and resolution planning requirements had shifted the demand curve for reserves outward — permanently.

The September 2019 repo market crisis exposed the true scale of the miscalculation. Three days before the federal funds rate spiked outside its target range, staff analysis stated that results “do not suggest that reserves are currently close to the ‘steep’ part of the reserve demand curve.” This was a catastrophic forecast failure. By 2020, the projected reserve floor had been revised to $2 trillion — a 400 percent increase from the 2017 baseline. The post-crisis regulatory framework had not merely enlarged the balance sheet. It had made the balance sheet’s equilibrium size fundamentally unpredictable.

The critical finding for Bowman’s reform proposal, however, is not the diagnosis but the response. The Fed’s internal deliberations during this period explicitly evaluated the option of reforming the discount window or the liquidity framework to reduce reserve demand — the exact proposal Bowman is now making. Jerome Powell himself argued in June 2019 that a standing repo facility would make firms “less eager to hold reserves as a liquidity buffer,” which would “enable us to run a smaller balance sheet.” Staff modeled a discount window ceiling. They designed a transitional credit facility to reduce stigma. They considered allowing banks to monetize Treasury securities rapidly through a dedicated facility.

Every alternative was evaluated. Every alternative was rejected or deferred — not out of inertia, but because the institution did not trust the discount window to perform the function these reforms would require it to serve. A December 2018 staff analysis proposed temporarily lowering the primary credit rate to use the window as a rate ceiling. It was rejected due to “persistent stigma.” The standing repo facility was discussed in June 2019 but was not operationalized in time to prevent the September crisis. The transitional credit facility was modeled but never adopted.

One instrument in this history requires a distinction. The standing repo facility that Powell endorsed in 2019 is not architecturally equivalent to a reformed discount window. It is market-facing, operates through Treasury securities rather than emergency lending, does not carry supervisory stigma, and is priced as a rate ceiling rather than a penalty. The signal extraction problem that has destroyed every discount window reform since 1984 — borrowing equals a Bayesian update on distress probability — does not apply in the same way to a facility whose counterparties are primary dealers, not supervised banks seeking emergency credit. Whether the SRF can materially reduce precautionary reserve demand without triggering the stigma dynamics that have defeated every prior reform is an empirical question the archive cannot yet answer. The facility was not fully institutionalized before the September 2019 crisis, and the 2023 failures tested the discount window, not the SRF. If Bowman’s reform agenda were built around expanding this market-based channel rather than rehabilitating the stigmatized one, the archive’s verdict would be less definitive. But her speech centers the discount window — harmonization, pricing, disclosure — and on that instrument, the record is deeply unfavorable.

The institution chose what the archive calls the “ample path of least resistance” — supplying more reserves rather than reforming the rules that generated demand for them. This was not ignorance. It was a judgment, grounded in the same evidence documented above: the discount window has never worked when needed, and the architects of the current framework built it that way on purpose. Bowman is proposing to reopen a door her institution examined, tested, and chose to leave closed — not once, but repeatedly, across multiple leadership teams and crisis episodes.

Quarles identified a further dimension that Bowman’s speech does not address. A permanently enlarged balance sheet, he warned, is an “attractive nuisance” — a political target and a standing temptation for fiscal accommodation. The balance sheet’s size is not merely a question of monetary policy efficiency. It is a question of institutional independence. Every dollar of excess reserves represents a dollar of interest the Fed pays to the banking system, a cost that becomes visible to Congress in ways that invite political intervention. Bowman frames the enlarged balance sheet as a technical problem with a technical solution. The archive frames it as a governance problem whose risks extend beyond the liquidity framework she proposes to reform. As Vice Chair for Supervision, Bowman occupies the seat from which these governance risks are most visible — and the seat from which the institutional pressure to resolve them is most acute. The speech may be about liquidity rules. The institutional stakes are about who controls the balance sheet and what it is used for.

The balance sheet is structurally enlarged. The costs are real. But the costs exist because the alternative — trusting the discount window — has been tried and has failed. If Bowman’s reform agenda rests on the premise that these costs are unnecessary, the archive poses a direct challenge: unnecessary compared to what? The communities she invokes are bearing the cost of a framework designed to protect them from something worse. The question is whether her proposed remedy would actually reach those communities — or whether the transmission mechanism is broken at a point her reforms do not touch.

The Remedy That Doesn’t Reach

Bowman’s strongest argument — the one where the reader’s instinct is to agree — is about communities. The liquidity framework reduces “credit availability to the economy.” Banks that over-allocate to high-quality liquid assets have less capacity to lend. The costs fall hardest on the borrowers who can least afford alternatives — small businesses, community institutions, the parts of the economy that depend on bank-intermediated credit because they lack access to capital markets. This is the argument that gives the reform agenda its moral weight. If the framework is hurting the people it was built to protect, something must change.

The diagnosis is real. Randal Quarles, Bowman’s predecessor as Vice Chair for Supervision, confirmed in October 2019 that the problem was “one of allocation rather than scarcity” — that regulatory “rigidities” and informal supervisory “instructions” to prefer reserves over Treasuries prevented liquidity from circulating efficiently. Bowman herself identified the structural dimension in an April 2019 speech, noting that transaction costs for a $100,000 loan are comparable to those for a $1 million loan, creating a built-in economic bias against the small-scale lending that communities need most. Governor Daniel Tarullo warned as early as 2016 that regulation and monetary policy had to be “thought of at the same time,” because the liquid markets that transmitted policy were the same markets that froze when stress arrived.

The reader nods. The framework is costly. The costs are regressive. Something should be done.

Then the archive delivers the test case. Between 2020 and 2022, the Federal Reserve created the most abundant reserve environment in its history. If Bowman’s theory is correct — that freeing up bank liquidity would transmit to small business credit — this was the natural experiment. Reserves were not merely ample; they were superabundant. Every liquidity constraint the LCR could impose was overwhelmed by the sheer volume of reserves the Fed injected into the system.

The results are documented in the Fed’s own reporting. Community bank small business lending did increase — by 39 percent between the end of 2019 and mid-2020. But this increase was driven almost entirely by the Paycheck Protection Program and other emergency facilities created in coordination with the Treasury. Outside these targeted interventions, business lending remained what Esther George, President of the Federal Reserve Bank of Kansas City, described as “tepid.” The abundant liquidity did not flow naturally through the transmission mechanism to the borrowers who needed it. It required a dedicated pipe — built by Congress, funded by the Treasury, administered through specific facilities — to reach small businesses at all.

A fair reading of this period requires acknowledging its limits as a natural experiment. Banks in 2020–2022 were risk-constrained, not merely liquidity-constrained. Pandemic-era uncertainty suppressed lending through risk scoring and borrower demand channels that operated independently of reserve levels. The PPP dominated credit flows so thoroughly that the underlying transmission mechanism was never cleanly tested. The period does not prove that abundant liquidity cannot improve small business credit. It proves that abundant liquidity did not, under the conditions that prevailed — and that a crisis severe enough to generate superabundant reserves is precisely the environment in which the confounders are most powerful.

But the structural evidence does not depend on the 2020–2022 episode. It predates the pandemic entirely.

The longer-term data is more damning. By 2022, the share of small businesses relying on personal sources for capital had risen to two-thirds, up from half in 2019. In the most liquid banking environment in American history, small businesses moved further away from bank credit, not closer to it. The bottleneck was not the LCR. It was not the quantity of reserves. It was the structural economics of small-scale lending — the transaction costs Bowman herself identified in 2019, the declining community bank footprint, the allocation rigidities that persist regardless of how much liquidity the system holds.

This finding echoes a pattern the archive has surfaced before. In The Assay, the same structural question was tested for rate cuts: does monetary accommodation transmit to small business credit? The answer, documented across four decades of easing cycles by Robert Kaplan, Elizabeth Duke, Charles Evans, and Loretta Mester, was no — low rates primarily benefit large corporations with capital market access while small businesses continue to face credit constraints through tightened lending standards. The mechanism Bowman proposes to activate — relaxing liquidity requirements to free lending capacity — operates through the same channel that rate cuts operate through. And the archive shows that channel is blocked.

The path dependence is now visible. The discount window cannot be trusted because stigma is structural. The LCR was built as self-insurance because the window was broken. The balance sheet is enlarged because the LCR demands reserves. And the communities Bowman invokes cannot be reached through the liquidity channel because the transmission mechanism is broken at a point her reforms do not touch. The diagnosis is valid at every stage. The prescription fails at every stage. The disease is real. The medicine does not reach the patient.

The Window