How the Federal Reserve Got So Huge, and Why and How It Can Shrink

Introduction

Over the six weeks following the failure of Lehman Brothers on Sept. 15, 2008, the massive expansion of credit provision by the Federal Reserve shifted the way it conducted monetary policy to a form it had considered and rejected only five months earlier in April. Commercial banks keep money on deposit at the Fed in accounts that are effectively the banks’ checking accounts. The deposits are referred to as “reserve balances”. The deposits are loans from banks to the Fed, just as ordinary retail deposits are loans from households to their banks. The Fed had been conducting policy for decades by borrowing from banks only the amount of reserve balances the banks considered necessary for payment purposes and to satisfy reserve requirements. Because the Fed had run out of options to fund its growing balance sheet, it shifted to borrowing excessive amounts of reserves from banks—that is, more than the banks needed to meet reserve requirements and satisfy payment clearing needs. In April, staff had recommended against using an excessive-reserves approach to conduct monetary policy, because “it would represent a radical departure from the basic elements of our own current framework and from those of almost every other central bank . . . ” [emphasis added].[1]

Although the Fed’s intention in late 2008 was to return to its previous method of implementing policy, over the next decade, a sequence of developments and decisions (including but not limited to three waves of quantitative easing) made it more expedient to conduct policy using an excessive-reserves approach. Staff also increased the roadblocks to returning to the necessary-reserves framework. On Jan. 29, 2019, after extensive discussion over several meetings, the Federal Open Market Committee (FOMC), the arm of the Federal Reserve that oversees monetary policy, officially decided to permanently shift to conducting policy using the excessive-reserves framework.

That decision was a mistake. The FOMC should return to conducting monetary policy using a necessary-reserves approach. The prior system had given the Fed easy, excellent control of overnight interest rates while allowing it to operate with a vastly smaller balance sheet and minimal involvement in the financial system. The excessive-reserves system has not produced its purported benefits. Policy has not been simpler to conduct, and although the federal funds rate has been stable, that’s in part because the federal funds market has essentially disappeared.

Advocates of an excessive-reserves approach have also argued that it is superior to the necessary-reserves approach because it satisfies something akin to the Friedman rule: that the inflation rate should be the negative of the Treasury bill rate, so people are indifferent between holding cash and investing in short-term, safe, interest-earning investments.[2] The logic is that the government can offer reserves, which serve as a source of liquidity to commercial banks, for free, so it is socially optimal to provide them in abundance. As we will discuss, however, the Friedman rule can be satisfied in a corridor system using the approach staff had recommended that the Fed adopt in its April 2008 study—a voluntary-reserve-requirement regime that pays a market rate on the self-required reserves.[3]

exhibit 1 assets of fed reserve as percent of gdp

When evaluating the framework in early 2008 and again in 2018, the Fed recognized that many of the costs of the system are associated with the Federal Reserve’s necessarily large balance sheet. Indeed, as shown in Exhibit 1, the assets of the Federal Reserve System have grown from 5 percent of GDP before the Global Financial Crisis (GFC) to 28 percent at the end of 2023. The quantity of reserves that currently appears necessary to implement an excessive-reserves system is three times higher than the level the Fed judged necessary when it officially adopted the framework in January 2019, and nearly 100 times higher than estimated when the Fed considered the advisability of the framework in April 2008. That extraordinary expansion has occurred because there is a ratchet in the reserve demand process. When the Fed borrows more reserves, the system adjusts to those reserves; as a result, there is resistance when the Fed attempts to normalize its balance sheet and borrow less.[4]

One of many components of the ratchet, and the other main cost anticipated by the Fed at the outset, is that by borrowing an excessive amount of reserve balances, the Fed has caused the interbank market and the associated institutional infrastructure to evaporate. Because banks have been able to earn an above-market rate by keeping elevated balances in their reserve accounts, they have less incentive to manage their liquidity tightly. As a result, the federal funds market—the market where banks with excess reserves on any given day lent to banks that were short of reserves—is essentially gone. With no interbank market, the need for each bank to maintain a high level of reserve balances gets locked in. The fact that the excessive-reserve framework leads banks to rely on deposits at the central bank as a liquid asset is why the Norges Bank decided to stop using such a system, and why some central banks are now reconsidering this approach.

A more subtle cost of the excessive-reserve system is that it imposes no natural limit on the size of the Federal Reserve’s balance sheet. Advocates of the approach believe that the Fed can borrow as much as it would like from banks without material consequences, which we will show is incorrect. By contrast, under the necessary-reserve system, if the Fed borrowed more reserves than banks wanted to keep in their accounts for clearing needs and to meet reserve requirements, the central bank would be unable to control interest rates. This perception that increased size has no cost has contributed to internal and external assumptions that the Federal Reserve’s balance sheet can be used to solve all problems, including monetizing the federal debt.

For reasons we explain in the next section, an excessive-reserves system is also called a floor system, while a necessary-reserves system is called a corridor system. This paper makes the case that the Fed should return to a necessary-reserves system and conduct policy using a corridor framework. To do so, we first present some background on the Fed’s balance sheet and the two methods of implementing policy. We then discuss why the Federal Reserve should use a corridor system.

The final section describes how the Federal Reserve can shift back to a necessary-reserves regime. If the Fed reduces the size of its balance sheet (and therefore the quantity of reserves available to the banking system), money market rates will rise above the interest rate the Fed pays banks on their deposits. This will make it increasingly expensive for banks to lend the Fed large amounts of excessive reserves. That shift will gradually overcome the ratchet effect, and the necessary amount of reserves will shrink over time. Note that once the level of reserves approaches the necessary level, the Federal Reserve will need to control large swings in reserves, as it did for most of its history. As we will discuss, it was a failure to control such swings that led to the bout of volatility in money markets in September 2019, the last time the Fed attempted to shrink.

As part of the transition back to a necessary-reserves framework, the Fed could adopt a voluntary-reserve targeting system. Each bank would set the level of reserves it wanted to hold over a maintenance period, and it would be compensated at the FOMC’s target federal funds rate for those reserves. Excess reserves would be compensated at a lower rate, and reserve shortfalls would be penalized in a way that replicates the cost of borrowing the shortfall amount from the discount window.

Such an approach separates the issue of the quantity of the reserves provided from whether the central bank implements policy using a floor or corridor system. The necessary amount is no doubt now much larger than it was before the Global Financial Crisis (GFC), because banks are both required and choose to maintain larger deposits at the Fed for liquidity management purposes. The amount banks judge necessary can be lowered through changes to bank liquidity regulations and by reducing the stigma associated with borrowing from the discount window.

If the Fed were to manage its balance sheet so that the quantity of reserves is slightly below the aggregate amount of the voluntary reserve targets, banks would need to turn to the discount window at the end of the maintenance period. The need to borrow, and the modest upward bias to money market rates, would offer an incentive for banks to economize on their use of reserves as a liquidity management tool, while the increased borrowing—especially as an intended component of the regime—would help lessen stigma. The resulting small average spread between money market rates and the rate paid on voluntary reserves would capture the social costs associated with operating the central bank with a large balance sheet.

Background

Understanding monetary policy implementation requires understanding central bank balance sheets. Exhibit 2 presents a stylized version of the Federal Reserve’s balance sheet.

exhibit 2 simp fed reserve sheet

The Federal Reserve’s assets consist primarily of securities—Treasury securities and, during some periods, agency debt and agency mortgage-backed securities. The Fed is required to purchase these securities in the open market, and they are collectively referred to as open market operation or “OMO” securities. The other main Fed assets are loans to banks (discount window loans); repurchase agreements against OMO securities; and, during crises, loans to nonbanks. We will follow the Fed’s convention of referring to repos and reverse repos from the perspective of the Fed’s counterparty. The Fed is therefore actually engaged in a reverse repo (lending cash against OMO collateral) when it engages in a “repo,” which is why repos are listed as an asset, not a liability.

The liabilities the Fed uses to fund its asset holdings consist primarily of currency, reserve balances, the Treasury’s deposit at the Fed (the “Treasury General Account” or “TGA”) and reverse repos. Currency is a zero-interest loan from the public (especially foreigners) to the Federal Reserve. Until the Lehman default in September 2008, the Fed funded itself almost entirely with currency. However, after Lehman, the Fed’s borrowing needs exceeded the amount of currency, so it began to borrow heavily from depository institutions: commercial banks, thrifts, and credit unions (“banks”). Banks offer funding to the Fed in the form of deposits at their Federal Reserve Bank. Those deposits are also known as “reserve balances.” At times, the Fed has needed to also borrow from money market mutual funds, GSEs, foreign official institutions and primary dealers (the broker-dealers authorized to be Fed counterparties) using overnight reverse repos collateralized by OMO securities. In 2013, the Fed began borrowing overnight reverse repos (ON RRPs) at a standing facility as a tool to support monetary policy.[5] Reverse repos had usually been zero before the GFC. In 2021–22, they were over $2 trillion but they have fallen substantially since, and as of Jan. 25, 2024, they were $558 billion.

The Federal Reserve’s book equity (using its own accounting conventions) is close to zero.[6] In 2015, Congress expropriated the Fed’s equity to help pay for the Fixing America’s Surface Transportation Act and limited the equity going forward to $10 billion.[7] In 2021, Congress limited the Fed’s equity to $6,785 million.

The Federal Reserve conducts monetary policy by setting its policy rates and adjusting its balance sheet so that the federal funds rate (the interest rate on unsecured loans between banks or between banks and GSEs) trades near the FOMC’s target or within its target range.[8] The Fed’s policy rates are the interest rate at which it lends to banks (the “primary credit rate” a/k/a the “discount rate”) and the interest rate it pays on deposits (the “interest on reserve balance” or “IORB” rate).[9] The Federal Reserve adjusts its balance sheet by engaging in repos or reverse repos or asset purchases or sales to expand and contract its assets.

Because currency, the TGA and reverse repos are unaffected by these transactions, and balance sheets must balance, by adjusting its assets the Fed also adjusts reserve balances. Setting aside the interaction between reserves and ON RRPs, the aggregate level of reserve balances—the amount the Fed needs to borrow from banks—is immutable: unchanged by transactions between banks or other financial institutions. At the same time, each bank can freely choose the reserve balance it wants. The federal funds rate, as well as other interest rates, adjusts so that banks each choose to keep deposit balances at the Fed that add up to the aggregate amount of reserve balances necessary to fund fully the Fed’s balance sheet. That’s how monetary policy works.

Banks choose to keep overnight deposits at the Fed—reserve balances—for various reasons. Until March 2020, banks were required to maintain reserves, on average over a two-week period, equal to a fixed percentage of their deposits; that is, banks were subject to a reserve requirement. In addition, a bank’s deposit at the Fed is its checking account, and banks maintain positive amounts in their accounts to support their payments activity. A deposit at the Fed is also the ideal source of contingency funding—it is perfectly safe and immediately available, and its value does not change. Banks currently keep a large amount of reserves for liquidity management purposes.

For half a century, the relationship between reserve balances and the federal funds rate has been conceptualized using a model William Poole developed in 1968 while he was an economist at the Federal Reserve Board.[10] The model is based on the behavior of a bank that must decide how much to borrow in the federal funds market at mid-day without knowing what transactions will hit its account in the afternoon. Note that the model describes the behavior of a bank over the course of one day, not over time.

Although not exactly how Poole presented it, the model as currently used is shown in Exhibit 3. The orange line is the quantity of reserves borrowed by the central bank each day, which is fixed following the morning repo operation, to adjust the size of the balance sheet. The purple line is the “demand” by banks for reserve balances, which describes what federal funds rate will correspond to any particular supply. To the left, the curve flattens out at the discount rate, because a bank should not be willing to borrow in the funds market for more than it can borrow from the Fed. To the right, the curve flattens out at the IORB rate, because a bank should not be willing to lend in the funds market for less than it can receive by simply keeping its funds on deposit at the Fed.

exhibit 3 - reserve balances and the fed funds rate

In the middle of Exhibit 3, there is a relatively steep portion of the line connecting the two flat regions. The relationship corresponded well with the behavior of the federal funds rate during the day when the Federal Reserve sought to supply the quantity of reserves demanded by banks and paid zero interest on deposits. Actually, the correspondence really only worked well on the last day of the two-week maintenance period over which banks had to meet their reserve requirements on average. For other days, the federal funds market generally cleared at the FOMC’s target rate, irrespective of modest changes in the quantity of reserve supplied or demanded, because banks knew that they could make up any excess holding or shortfall by adjusting their balances later in the maintenance period.

At the end of the maintenance period, if the Fed inadvertently created more reserves than banks wanted, the federal funds rate dropped sharply at the end of the day, because all market participants sought to shed the reserves on which they would earn zero. If the Fed inadvertently created insufficient reserves, the federal funds rate would rise until it reached the discount rate. Banks would then borrow from the discount window, expanding the Fed’s balance sheet and creating the necessary reserves. In fact, as we will discuss, banks were and are reluctant to use the discount window, so the federal funds rate would rise somewhat above the discount rate before banks would borrow.

The model is well suited for describing corridor and floor monetary policy implementation frameworks. The federal funds rate is determined within a corridor created by the discount rate, which is the ceiling, and the deposits rate, which is the floor. As shown in Exhibit 3, if the central bank is operating using a corridor system, it sets a target for the middle of the corridor and sets the supply of reserve balances in the middle of the steep section of the demand curve. If the central bank is using a floor system, it sets the target rate at the deposit rate and sets the supply of reserves well to the right of the steep slope.

It is a convenient fiction to say that the Fed operated policy using a corridor system before the GFC. From the mid-1960s to 2002, the discount rate was below-market. The Fed controlled rates using the upward supply of reserves induced by the reluctance to borrow at the discount window, changes in reserve supply through open market operations, and the downward sloping demand for reserves. From 2003 (when the Fed changed the discount rate to an above-market rate) to October 2008, the Fed had a ceiling, but the floor was zero because the Fed was not allowed to pay interest on reserves.

On Oct. 1, 2008, the Federal Reserve began paying interest on reserve balances. The Financial Services Regulatory Relief Act of 2006 authorized the Federal Reserve Banks to pay interest on bank balances, effective Oct. 1, 2011.[11] The start date was advanced to Oct. 1, 2008, by the Emergency Economic Stabilization Act of 2008.[12] The Fed had sought the authority to pay interest on required reserves, not excess ones.[13] Paying interest on required reserves eliminates the tax on banks caused by forcing them to hold unremunerated required reserves: forced zero-interest loans to the government. Moreover, doing so eliminated the incentive for banks to minimize their reserve requirements by sweeping deposits at the end of the day into money fund accounts. Sweep accounts had become so common that the level of required reserves was falling below the level necessary to implement monetary policy.

However, Congress accelerated the date for the Fed to start paying interest on reserves so that the Fed could pay interest on excess reserves as well as, incidentally, required reserves. It was necessary for the Fed to pay interest on excess reserves to conduct monetary policy once the Fed began offering more reserve balances than banks demanded as a consequence of the post-Lehman expansion of the balance sheet. By creating excessive excess reserves, the Fed switched from a corridor to a floor framework for monetary policy implementation. And as can be seen in Exhibit 4, when the Fed began to oversupply reserves, the federal funds rate fell well below the interest on excess reserves (IOER) rate. At the time, the IOER rate was 1 percent and the FOMC’s target for the fed funds rate was 2 percent, but the funds rate fell essentially to zero. The IOER rate’s failure to create a floor on the funds rate was moot by December 2008, because the Fed set a target range for the funds rate of 0 to 25 basis points. But the failure cast doubt on the effectiveness of the IOER rate to put a floor on the funds rate. This would become important five years later, when the FOMC began to contemplate raising the funds rate above zero.

exhibit 4 - graph

Although the Poole model describes the intraday behavior of the fed funds market well, it has proven a poor guide for how the demand for reserves behaves when the Fed pays interest on reserve balances and borrows massively more from banks than necessary. As Poole (1968) stated:

The model presented here concentrates on these very short-run adjustments. However, it is obvious that the bank must make further adjustments if it experiences persistent reserve drains or accretions. (p. 770)

As shown by Afonso and colleagues (2022, including John Williams, the president of the New York Fed), the minimum level of reserve balances required by banks has shifted to the right over time.[14]

Indeed, as reported in Exhibit 5, the Fed’s estimate of the quantity of reserve balances required to implement a floor system has risen dramatically over time, in line with the tremendous growth in reserve balances. In the original staff memo from April 2008 on methods to implement policy using interest on reserves, staff estimated that $35 billion would be necessary to implement policy in a floor system. In March 2016, the New York Fed raised its assumption to $100 billion.[15] In March 2018, it raised it again to $600 billion.[16] At the December 2018 FOMC meeting, at the cusp of the Fed deciding to adopt a floor system permanently, staff judged that $1 trillion would be needed, $800 to supply the reserves required by banks and a $200 billion buffer to absorb volatility in the supply caused by changes in other balance sheet items, particularly the TGA.[17] By September 2019, it had more than doubled this to $1.3 trillion.[18] In the New York Fed’s forecast of the Fed’s balance sheet published in May 2022, the estimate essentially doubled again to $2.3 trillion. Responses to the Fed’s May 2023 Senior Financial Officer Survey found that a notable fraction of banks had raised their target level of reserves substantially since late 2022.[19]

Nelson (2018a, 2022) argues that the growth is the result of bank management and bank examiners becoming accustomed to addressing bank liquidity needs with subsidized reserve balances. For example, in 2010, concern about the negative consequences of holding excess reserves led Norges Bank (the central bank of Norway) to switch from a system with abundant reserves to one with more scarce ones. When seeking comment on their decision, they noted:

When Norges Bank keeps reserves relatively high for a period, it appears that banks gradually adjust to this level. . . . With ever increasing reserves in the banking system, there is a risk that Norges Bank assumes functions that should be left to the market. It is not Norges Bank’s role to provide funding for banks. . . . If a bank has a deficit of reserves towards the end of the day, banks must be able to deal with this by trading in the interbank market.[20]

Alternatively, Acharya, Chauhan, Rajan, and Steffen (2023) point to banks’ changing their balance sheets in ways that require the added liquidity from reserve balances, such as offering more lines of credit. Similarly, Lopez-Salido and Vissing-Jorgensen (2023) argue that reserve demand increases with the level of bank deposits. Consequently, the current conceptualization of reserve demand and supply, and the behavior of money market rates, is not the Poole model with its stable demand curve and fixed steep section. Instead, similar to the relationship between short-run and long-run production functions familiar to microeconomists, the steep part of the demand curve moves gradually up and down, in reaction to the available supply of reserve balances and configuration of interest rates.

exhibit 5 fed est banks struc demand for reserve bal

If it still seems helpful to use the Poole model as an intuitive source, the relationship between the level of reserve balances and the federal funds rate appears to look like Exhibit 6. As the Fed offers more reserves, the steep part of the demand curve shifts right over time. Moreover, rather than being flat at the IORB rate, the demand curve slopes gradually down—the more reserves the Fed borrows from banks, the more it has to pay for them, relative to market rates.

exhibit 6 - floor system with ratchet

To read the full paper, please click here, or click on the download button below.


[1] Federal Open Market Committee (2008).

[2] Friedman (1969).

[3] Federal Reserve System (2008).

[4] Standard practice is to refer to a bank’s willingness to lend reserve balances to the Fed as its “demand” for reserves, just as a household’s willingness to lend deposits to a commercial bank is referred to as its demand for deposits.  Conversely, the amount that the Fed borrows from banks in the form of deposits is referred to as the Fed’s “supply” of reserves.

[5] See Nelson (2021) for more information.

[6] The Fed does not follow GAAP accounting rules; it follows its own.  See Board of Governors of the Federal Reserve System (2023a).

[7] If Congress had taken the equity but allowed the Fed to build it back up by retaining earnings rather than remitting them to Treasury, the expropriation would not have scored as a source of revenue in the budget projection.

[8] The Fed switched from targeting a level for the fed funds rate to targeting a range on Dec. 16, 2008.

[9] The Fed set reserve requirements to zero in March 2020.  Originally, the relevant policy rate was the interest on excess reserves (IOER) rate, but we will mostly refer to the rate as the IORB rate. 

[10] Poole (1968).

[11] https://www.congress.gov/109/plaws/publ351/PLAW-109publ351.pdf

[12] https://www.congress.gov/110/plaws/publ343/PLAW-110publ343.pdf

[13] Kohn (2005). 

[14]Afonso and colleagues (2022)

[15] Federal Reserve Bank of New York (2015).

[16] Federal Reserve Bank of New York (2018).

[17] Board of Governors of the Federal Reserve System (2018) p. 23.

[18] For further discussion of the Fed’s unfortunate decision to hold back its forecast of its balance sheet and income in March 2019, see Nelson (2019a).

[19] Board of Governors of the Federal Reserve System (2023b).

[20] Norges Bank (2010).