Déjà Vu at the Federal Reserve

How liquidity dependence keeps ratcheting up the Fed’s balance sheet and what can be done about it.

The FOMC brought quantitative tightening (QT) to an end at its December 2025 meeting. In doing so, it reaffirmed something that has become increasingly clear over the past decade: the structural size of the Fed’s balance sheet keeps ratcheting higher after each round of QE. It all feels a bit like déjà vu.

As Bill Nelson noted, the Fed now believes it needs roughly $3 trillion in reserve balances to operate its floor system, implying a minimum securities portfolio of about $6½ trillion once currency in circulation and the Treasury General Account are taken into account. What stands out is not just the size of that number, but its direction. Over time, the Fed’s estimate of what constitutes “ample” reserves has moved in only one direction: up. The table below, drawn from Nelson’s analysis, reveals this pattern:

So why does reserve demand keep rising? Nelson’s explanation is that under an ample-reserves system, banks and supervisors adapt to whatever reserve level the Fed supplies—embedding those levels into stress tests and expectations—so reserve demand ratchets upward over time and is never revised back down.

A complementary explanation is that QE encourages the expansion of runnable liquidity claims, such as demandable deposits, thereby reducing net aggregate liquidity. This, in turn, makes QT increasingly difficult to implement without generating financial stress. In this newsletter, I explain this second mechanism as articulated by Viral V. Acharya, Rahul S. Chauhan, Raghuram Rajan, and Sascha Steffen in a series of papers. I then argue that the Fed can mitigate this challenge by using its term deposit facility in a manner similar to the approach now employed by the Swedish central bank, the Riksbank.

I learned of the Riksbank use of term deposits to manage liquidity and it potential application for the Fed in my recent conversation with Per Åsberg-Sommar on the Macro Musings podcast. That discussion was the genesis of the ideas developed in this newsletter. Be sure to check out the episode; a short video excerpt from the interview appears below.

Ratcheting Up the Size of the Fed’s Balance Sheet through Liquidity Dependence

So why has QE proven so difficult to reverse? Acharya et al. offer a compelling answer: QE creates liquidity dependence.

Their argument begins with a simple but underappreciated observation: QE alters the liability structure of banks, not just mechanically but behaviorally. As QE expands reserve balances, banks grow increasingly confident that they can backstop withdrawals and drawdowns. This confidence, in turn, encourages banks to rely more heavily on relatively cheap, highly liquid funding sources—namely demandable deposits and credit lines.

Over time, this shift toward runnable liabilities creates what the authors call liquidity dependence. Driven by public demand for liquidity, institutional momentum in liquid liability creation, moral hazard arising from expected central bank intervention, and the structure of liquidity regulations, these liquid liabilities prove very difficult to unwind during QT.

The result is a fundamental asymmetry: QE expands banks’ liquid liability claims, while QT fails to reverse them—even as it reduces the reserves that backstop those claims. The system is thus left with a growing mismatch between liquidity promises and available liquidity, one that can ultimately be resolved only by renewed Fed balance-sheet expansion.

This is the ratchet mechanism at the heart of liquidity dependence. As the authors put it, “the supply of reserves creates its own demand for reserves over time, ratcheting up the required size of the Fed’s balance sheet” (p. 349).

The September 2019 repo market stress provides a concrete illustration. The standard narrative points to large inflows into the Treasury General Account (TGA) and reserve hoarding by a subset of banks. While not incorrect, Acharya et al. argue this explanation misses the deeper cause. Even as QT reduced the stock of reserves—the backstop for banks’ runnable liquidity claims—banks did not meaningfully shrink those claims. As a result, net aggregate liquidity was far smaller than headline reserve levels suggested. In that context, the TGA inflows were merely the final trigger, the “straw that broke the financial system’s liquidity back because the financial system was already overloaded with potential liquidity claims relative to available liquidity” (p. 414).

Acharya et al. argue that a similar dynamic also amplified the March 2020 “dash for cash.” After years of QE-driven expansion in runnable liquidity claims and a subsequent drawdown of reserves under QT, firms’ rapid credit-line drawdowns and investors’ scramble for cash exposed a system in which claims on liquidity far exceeded available backstops. In this sense, the pandemic shock laid bare the underlying liquidity dependence.

The liquidity dependence story be illustrated in the two figures below. The first one shows demandable deposits as a percent of total liabilities for private depository institutions with the time frames of the four QE programs shaded in grey.

The figure reveals that demandable deposits average between 40-50 percent of liabilities in the decades leading up to 2008. Since then, U.S. depository institutions have steadily shifted toward funding themselves more with demandable deposits, a trend that accelerates during QE episodes and in most cases does not reverse afterward. While post-crisis liquidity regulations such as the LCR shape banks’ asset holdings, they do not explain the persistent shift toward demandable funding shown here. Time deposits are treated more favorably than demandable deposits under the LCR, yet banks have continued to shorten liabilities during QE and failed to reverse this behavior during QT.

The rising level of demandable deposits occurs even as reserves as a percent of total bank assets temporarily declines as seen below:

In their latest paper, Acharya et al. show that this shift in funding structure is also highly uneven across the banking system, with liquidity risk becoming increasingly concentrated at small and mid-sized banks. Using bank-level data, they document that institutions less constrained by post-crisis liquidity regulations expanded uninsured demandable deposits and credit lines most aggressively during QE, while failing to unwind those claims during QT. As reserves drained from the system, liquidity tended to migrate toward the larger systemically important banks, leaving smaller ones with a growing mismatch between runnable liabilities and available liquidity.

The experience of Silicon Valley Bank illustrates this dynamic in an extreme form: although SVB’s failure also reflected idiosyncratic risk-management failures and an unusually concentrated depositor base, those weaknesses became fatal only once QT and rising rates exposed a broader reallocation of liquidity toward larger banks. In this sense, liquidity stress tends to surface first at the periphery—even when aggregate reserve measures still appear ample—ultimately forcing the Fed to respond at the system level.

If liquidity dependence is the problem, the fix must lie in changing how central bank liquidity is supplied.

A Cure for Liquidity Dependence: A Lesson from the Riksbank

Simply supplying ever-larger quantities of reserves cannot be that fix. It may relieve stress in the moment, but it also encourages banks to expand runnable liquidity claims that do not unwind symmetrically, leaving the financial system just as fragile—and often more so—when the next financial shock hits.

What is needed is a way to conduct QE that, first, does not incentivize the growth of runnable liquidity claims and, second, makes QT the natural, automatic default once large-scale asset purchases end. The solution, then, lies not in increasing the quantity of central bank liquidity, but in changing the form and pricing of Fed liabilities.

One promising approach, used by the Riksbank and discussed with Per Åsberg-Sommar on the podcast, is to make term deposits a core monetary policy instrument. The Riksbank operates a narrow corridor around its policy rate by offering banks a menu of standing facilities with different maturities and prices. Short-term central bank certificates (effectively term deposits) are offered at the targeted policy rate, while overnight deposits earn a rate set 10 basis points below the policy rate and overnight lending is available at a rate 10 basis points above it. This operating system can be seen in the figure below:

This approach creates a demand-driven operating system: banks themselves choose how much liquidity to hold as overnight reserves versus term deposits. Liquidity is always available—banks can sell certificates or borrow against them at a modest cost if reserves are needed—but they are simultaneously incentivized to minimize idle reserve holdings because term deposits earn more. In this way, the Riksbank can absorb excess liquidity during QE without creating fears of shortage, while maintaining tight control over short-term interest rates.

Ending Liquidity Dependence in the United States: Using and Properly Pricing the Term Deposit Facility

Imagine now that the Federal Reserve adopted a similar operating framework by offering term deposits at the targeted policy rate (i.e. federal fund rate) while setting interest on reserves 10 bps below it and and placing the discount window (DW) and standing repo operations (SRP) rates at 10 bps above the policy rate.

By introducing maturity into the Fed’s liabilities, this structure would reduce banks’ incentive to rely on runnable funding while preserving interest-rate control and lender-of-last-resort capacity. That is, since banks would earn a higher return by holding term deposits rather than overnight reserves, they would no longer treat reserves as a costless, open-ended backstop for short-term liabilities. This would dampen the incentive to expand highly liquid, demandable funding and instead encourage banks to rely more on term funding. At the same time, liquidity would remain readily available: banks could convert term deposits into reserves at a modest cost or borrow from the DW or SRP if needed, eliminating fears of sudden reserve shortages.

Just as importantly, this framework would make balance-sheet normalization much easier. As asset purchases ended, term deposits would mechanically roll off at maturity unless renewed, reducing central bank liabilities without discretionary action by the Fed. QT, in other words, would become the default outcome once QE ended, with liquidity contracting endogenously as banks adjusted their demand for term versus overnight central bank liabilities. In this way, the Fed could shrink its balance sheet without repeatedly triggering the liquidity stresses that have come to characterize recent episodes of QT.

Fortunately, the Federal Reserve already has much of this machinery in place. Since 2010, it has maintained a Term Deposit Facility (TDF) that allows banks to convert reserves into term deposits at the Fed. In practice, however, the facility has been rarely used and only on a limited scale. Recasting the TDF as a central instrument—priced at the target policy rate and embedded in a narrow corridor system—would fundamentally change its role. The TDF would become the primary way excess liquidity is absorbed and released over time. In this sense, ending liquidity dependence would not require inventing a new tool, but rather elevating an existing one to play the role it was always capable of playing.

Conclusion: From Liquidity Dependence to a Demand-Driven Fed

In an earlier newsletter, From Floors to Ceilings, I outlined three-steps needed to make quantitative tightening easier and more predictable. First, reduce duration mismatch through a Treasury–Fed asset swap. Second, neutralize volatility in the TGA so that fiscal cash management does not destabilize reserves. Third, normalize the use of ceiling facilities—the DW and SRP—to support a genuinely demand-driven system. .

Taken together, these proposals point toward a state-contingent operating framework. In normal times, the system would function as a demand-driven corridor system in which banks are incentivized to move out of overnight reserves and into term deposits, while retaining the freedom to choose the maturity of their liquidity holdings. In periods of financial stress, the same framework would naturally shift into a ceiling system. When needed, it could also temporarily accommodate large-scale asset purchases and an ample-reserves configuration, without embedding those conditions as the permanent operating regime, because the pricing of term deposits would naturally pull the system back toward scarcity as conditions normalize. In this way, the Fed could preserve interest-rate control and financial stability without locking itself into ever-larger balance sheets.

In short, ending liquidity dependence does not require abandoning QE. It requires rethinking how central bank liquidity is supplied, priced, and allowed to mature over time so that discipline prevails in calm periods and elasticity is available when it is truly needed.

Update: A recent paper by R. Matthew Darst, Sotirios Kokas, Alexandros Kontonikas, José-Luis Peydró, and Alexandros P. Vardoulakis (2026) builds on the liquidity-dependence framework and shows that the problem is not just one of bank balance sheets, but of liquidity insurance in the broader economy. Using granular supervisory data, the authors document that QE-induced inflows of uninsured deposits from non-bank financial institutions—such as money market funds, hedge funds, broker-dealers, and other asset managers—lead banks to actively manage liquidity risk by reshaping both sides of their balance sheets. While banks shift toward safer funding by attracting more insured deposits, they simultaneously cut back on undrawn credit lines to firms, reducing the economy’s supply of contingent liquidity insurance. Importantly, QT does not reverse these fragile funding structures; instead, it exacerbates them as reserves decline, further reducing banks’ willingness to provide liquidity backstops. Together, these findings reinforce the view that QE under the current framework can reduce, rather than expand, effective liquidity—and strengthen the case for rethinking how central bank liquidity is supplied and priced. Check it out!