I. What are Bank Reserves?
The funds that banks use to issue loans primarily come from deposits from other customers. If banks use all deposits for lending to earn interest, although it improves the utilization of funds, they will face difficulties in meeting customer withdrawal demands whenever customers withdraw deposits. Bank reserves (Bank Reserves, also known as bank preparation funds) are cash held by financial institutions to meet central bank statutory requirements and customer withdrawal demands; these funds must be kept in the vaults of regional Federal Reserves or central bank accounts. According to the Basel III Accord, reserves are classified as high-quality liquid assets and are the safest and most direct liquidity assets for major banks. Compared to other assets, reserves are cash that can be used directly to settle debts.
The reserve system originated in medieval Europe. At that time, wealthy individuals stored gold with goldsmiths, who discovered that people would not simultaneously withdraw their gold, so they lent out a portion of the gold to earn interest, keeping only a fraction to meet occasional withdrawal demands. This was the prototype of the modern "fractional reserve system."
The Fed mandates that each bank must maintain a certain percentage of cash relative to bank deposits, known as the required reserve ratio. Historically, the reserve ratio has ranged between 0% and 10%. Before March 26, 2020, the Fed generally required a reserve ratio of 10%, but for small banks that met the criteria for "Exemption amount" or "Low reserve tranche amount," the reserve ratio could be reduced to 0% or 3%, respectively, as detailed in the figure below.
The most direct function of cash reserves is to prevent banks from experiencing runs when facing large or unexpected withdrawal demands; adequate reserves also enhance customer trust in banks during daily operations. However, this system is not perfect; the collapse of Lehman Brothers is a case in point. Unless banks can guarantee that 100% of customer funds are retained, once a severe trust crisis and a run occur, someone will inevitably be unable to withdraw their funds, potentially leading to the collapse of the entire banking system.
Many Chinese people from the mainland often hear terms like "lowering the reserve requirement" and "raising the reserve requirement," referring to the central bank adjusting the required reserve ratio to regulate interest rates and money supply. However, in the U.S., in response to the economic recession triggered by the pandemic, the Fed has lowered the required reserve requirement to zero. In fact, the U.S. is among the last developed countries to eliminate reserve requirements; Canada abolished this requirement as early as 1992, the UK has always implemented a voluntary reserve policy, and regions such as Hong Kong and New Zealand also have no reserve requirements. This raises the question: should we reevaluate traditional concepts of reserves?
II. Can the Reserve Ratio Determine Money Supply?
Let’s discuss a related concept - the money multiplier, with the core formula "Total Broad Money = Base Money (Reserves) * Money Multiplier." In simple terms, the currency issued by the central bank (reserves) is magnified by a certain factor through the bank's credit operations, and this factor is the money multiplier, calculated as 1/reserve ratio.
For a simplified example: Suppose a newly established commercial bank initially has no deposits or loans, and the reserve ratio is 10%. Customer A deposits 100 yuan into the bank, and A's account shows a deposit of 100 yuan, thus the bank's deposit level reaches 100 yuan. The bank retains 10 yuan as reserves in the central bank and lends out the remaining 90 yuan to customer B. B then deposits this 90 yuan back into the bank, and at this point, B's account shows a deposit of 90 yuan, increasing the bank's total deposits by 90 yuan. The bank will again set aside 10% (i.e., 9 yuan) of this deposit as reserves and lend the remaining 81 yuan to customer C. Customer C then also deposits this money back into the bank... This cycle continues, ultimately forming several geometric series, resulting in a remarkable outcome: the initial deposit of 100 yuan made by customer A, through credit expansion, increases the total scale of bank deposits and the deposits in customers' accounts to 1000 yuan, referred to as "deposit money," which is "1/reserve ratio" times the original deposit. Note that throughout this process, only the 100 yuan initially deposited by customer A is cash, which is the base money issued by the central bank, while the expanded portion is merely an increment in electronic accounts, not involving the actual issuance of new money, which is the charm of credit!
Of course, the calculation of the money multiplier in reality is far more complex than the model described above, as different types of deposits correspond to different reserve ratios. Nevertheless, theoretically, central banks can indeed influence the creation of "deposit money" by adjusting the reserve ratio, thereby controlling the total money supply. However, in practice, the assumptions of the money multiplier theory are overly stringent: it requires banks to lend out all funds exceeding reserves and that there is sufficient credit demand in society, and borrowers must also redeposit the funds obtained back into banks. These conditions mean that the reserve ratio can only exert influence on the upper limit of the actual money multiplier, lacking significant correlation with the actual values.
In Figure 3, the red line represents M2 (including cash, checking deposits, and other short-term deposits), the green line represents M0/base money (including cash and bank reserves), and the ratio of M2 to M0 (blue line) is the actual money multiplier; Figure 4 shows the reserve ratio in the United States. It can be seen that the reserve ratio was only reduced twice, in the 1990s and in 2020, but the money multiplier fluctuated significantly and had almost no correlation with the reserve ratio, which statistically negates the impact of the reserve ratio on the money multiplier. Especially after the pandemic, when the reserve ratio dropped to zero, mathematically zero cannot serve as a denominator, making any reciprocal relationship between the two moot.
III. From the Scarce Reserve System (SRS) to the Ample Reserve System (ARS)
1. Before 2008, reserves were scarce.
Bank reserves are divided into required reserves and excess reserves. Required reserves are the minimum amount of cash that the central bank mandates banks to retain, while any amount exceeding this requirement is called excess reserves. Generally, banks lack the incentive to hold excess reserves, as cash yields no return and may depreciate due to inflation over time. In this scenario, apart from required reserves, the fluctuation of reserves is mainly influenced by the economic environment: during periods of economic prosperity, businesses and consumers are more inclined to borrow and spend, prompting banks to utilize excess reserves as much as possible for lending; conversely, during economic downturns, credit demand decreases and banks may tighten lending standards due to concerns over defaults.
Banks typically avoid unnecessary excess reserves to reduce the opportunity cost of not investing funds into income-generating assets, thereby preferring to hold as little reserve as possible. In periods when quantitative easing was not widely used, such as just before the 2008 financial crisis, the Fed's balance sheet was about $900 billion in size, and the total amount of bank reserves was around $100 billion, almost all of which were required reserves. As illustrated below, the scale of reserves in banks' accounts at the Fed was extremely low, a structure referred to as a Scarce Reserve System (SRS).
However, banks inevitably encounter temporary funding gaps, at which point they can choose to withdraw funds from their excess reserves or borrow overnight from other commercial banks (the borrowing rate is the interest rate at which U.S. commercial banks lend to each other). Due to the general scarcity of excess reserves, overnight borrowing often becomes the more common practice. The banking system operates under conditions of total reserve scarcity and heavily relies on an active interbank market for the redistribution of reserves. The interbank borrowing rate is the interest rate level at which uncollateralized overnight borrowing occurs between financial institutions, determined by daily market transactions. The primary lever for the Fed to manipulate the economy is adjusting short-term interest rates to ensure that the effective federal funds rate (EFFR, the weighted average interest rate for overnight borrowing among banks) closely aligns with the announced target range for the federal funds rate.
In a scarce reserve system, the Fed can fine-tune the supply of reserves to alter the supply-demand relationship in the interbank lending market to adjust short-term interest rates. The Fed communicates its monetary policy stance by announcing a target for the federal funds rate (usually a range) as a target for short-term rates. Afterward, the Fed actively manages the quantity of reserves in the financial system by buying and selling short-term U.S. government bonds through open market operations to bring the EFFR closer to the target set by the Fed. The principle is that when the Fed buys U.S. government bonds or other securities, it pays the sellers using checks drawn on itself. When sellers deposit these checks into their banks, their claims on the Fed become new reserves in the banking system. In short, when the Fed purchases securities, it creates bank reserves and increases the reserve balance; when it sells bonds, bank reserves decline. For banks, withdrawing reserves and interbank borrowing are both options during times of insufficient funds. The "funds" borrowed by banks in the lending market are the balances other banks have deposited at the Federal Reserve (these balances are sometimes referred to as federal funds). If reserves are abundant, the willingness to borrow from other banks decreases, leading to a drop in interbank borrowing rates until the level of reserve sufficiency and borrowing rates reach a balance; the opposite is also true.
Before the 2008 financial crisis, the New York Fed closely monitored fluctuations in banks' reserve demand, adjusting the Fed's daily securities purchases in a timely manner to keep the EFFR close to target levels. It is important to note that while this operational approach appears similar to later QE, the difference lies in the types and quantities of securities involved. Due to the scarcity of overall reserves, interbank borrowing rates are extremely sensitive to changes in reserve balances, so the Fed typically achieves effective rate control through small purchases of short-term government bonds.
2. After 2008, Reserves Became Ample.
After 2008, the U.S. fell into a deep recession, and the Fed attempted to respond to the economic distress by expanding the volume of securities purchases through open market operations to drive down short-term interest rates to nearly zero. However, this measure was not sufficient to support the sluggish economy. Typically, long-term interest rates tend to move in sync with short-term rates, but during the crisis, long-term rates remained elevated, impacting mortgage and corporate borrowing rates. Consequently, the Fed began purchasing large quantities of long-term government bonds and mortgage-backed securities (QE) to reduce long-term rates, which had remained above zero. The U.S. shifted from a scarce reserve system to an ample reserve system.
After experiencing two recessions in 2008 and 2020, the Fed introduced liquidity tools and engaged in massive asset purchases to improve financial market conditions and stimulate the economy. Although these actions mainly converted bad assets held by banks or provided loans, the result was still a significant increase in liquidity reserves within the banking system. Additionally, prior to the 2008 financial crisis, cash reserves held by banks earned no interest. On October 1, 2008, as part of the Emergency Economic Stabilization Act, the Fed began paying interest on excess reserves (IOER, Interest on Excess Reserves) to encourage banks to hold more reserves to face financial risks. After the pandemic outbreak in 2020, the Fed eliminated the portion of required reserves, allowing all reserves to earn the same interest (IORB, Interest on Reserve Balances). These two reforms changed the traditional notion that banks would rather lend out funds than keep them in accounts with the Fed; banks are now more willing to earn a smaller but risk-free interest rate rather than lend out for slightly higher but riskier returns.
In this way, on the one hand, QE brought abundant liquidity (banks had money), and on the other hand, depositing money in the Federal Reserve could yield risk-free returns (banks were willing to deposit money). As a result, bank reserves expanded at an astonishing rate: In August 2008, the reserve balance was $15 billion; by early 2009, it had increased to $800 billion; and by the end of 2021, it reached $4 trillion. Even two years after the balance sheet reduction, bank reserves still stand at $3.2 trillion.
The shift of reserves from scarcity to abundance has made it very difficult for the Fed to control interest rates through open market operations as it previously did. If the Fed wishes to maintain the short-term interest rate target at zero, then abundant reserves are indeed not a problem, because banks are very well-off, and no one is inclined to borrow from the overnight lending market, allowing borrowing rates to remain extremely low. However, in 2015, the Fed wanted to begin raising interest rates. In such a context of abundant reserves, engaging in small-scale buying and selling of reserves (traditional open market operations) hardly has any impact, since minor changes in reserve balances do not alter the banks' liquidity situation. The Fed urgently needed a new and effective means to regulate interbank borrowing rates, which later came to be known as the "interest rate corridor." In simple terms, this corridor sets an upper and lower limit for market interest rates. This guidance is not a hard requirement, but basic economic principles ensure that interbank borrowing rates remain within this range.
The upper and lower limits of the corridor that the Fed initially envisioned are the discount rate and the reserve interest rate (IORB). The discount rate is the rate at which the Fed lends to creditworthy banks, providing them with a channel to borrow funds outside the usual market. Interbank borrowing rates generally do not exceed this upper limit, as the Fed is the safest lending counterparty; if the cost of borrowing from other sources is higher, banks will simply choose to borrow from the Fed. IORB is the yield banks earn when they deposit money with the Fed; banks have no incentive to lend at lower rates if the returns from depositing elsewhere are lower, so they will choose to lend to the Fed (in other words, to lend to the Fed).
Currently, the discount rate is 5.0%, and IORB is 4.9%. According to the logic above, the effective federal funds rate (EFFR) should theoretically always be between the blue and red lines. However, the current EFFR is 4.83%, very close to IORB but actually below it; the imagined interest rate floor has effectively become a ceiling. Something must be amiss.
"Ceiling" and "floor" may be misleading terms, as the discount rate and IORB do not strictly define the upper and lower limits of market interest rates. For instance, banks may choose to borrow in the market at rates above the discount rate out of fear that borrowing from the Fed might be interpreted by their clients as a sign of financial issues. However, compared to the "leaky" "ceiling" of the discount rate, the IORB, which was initially envisioned as a "floor," may present more significant and complex issues.
As mentioned, the Fed's IORB sets a lower limit for short-term rates; that is, no bank will lend to anyone at rates below what they earn on their deposits at the Fed. However, only banks can deposit reserves at the Fed, while other financial institutions, such as money market funds and some government-sponsored enterprises, do not have access to this interest. In a context of ample liquidity, these institutions may be willing to lend overnight at rates below IORB, leading to the IORB's role as a rate floor becoming ineffective. Further, when the market presents low-interest loans below IORB, banks will seize the arbitrage opportunity, actively borrowing these loans, depositing them back into their reserves, and profiting from the interest spread. For example, if IORB is currently 4.9%, banks will not lend at rates lower than 4.9%, because doing so would yield less interest than depositing the money at the Fed for a risk-free return. However, many money market funds cannot directly deposit money with the Fed to earn interest, so they might be willing to provide loans to banks at a rate of 4.5%. After banks borrow these loans and deposit them into their reserve accounts, they can earn a 0.4% spread. Banks will continue this arbitrage until market rates approach 4.9%, at which point it becomes unprofitable.
To address the issue of the ineffectiveness of the interest rate floor, the Fed created a separate tool to absorb cash from non-bank financial institutions - our old friend the overnight reverse repo (ON RRP, detailed in the first article of this series). Since a broader range of financial institutions (not just banks) can use the RRP tool to deposit and earn interest from the Fed, it sets a true hard lower limit for market interest rates, ensuring that no financial institution will lend at rates below RRP, thereby granting the Fed control over the bottom of short-term rates. Prior to the emergence of RRP, most cash provided by the Fed to financial institutions through securities purchases was deposited in banks, turning into reserves. With RRP, non-bank institutions have a better option for liquidity after obtaining it from the Fed.
After supplementing the above content, we now have a whole new perspective to understand the previously mentioned statement "RRP as a reservoir during the era of quantitative easing, and as a blood transfusion bag during the era of balance sheet reduction." In other words, the Fed's QT is mainly reflected on the liability side of its balance sheet, namely the decrease in RRP balances, rather than a reduction in bank reserves. When the RRP balance becomes very low, further QT will lead to a decrease in bank reserve balances, harming true liquidity. Therefore, a significant drop in RRP balances may signal the impending end of QT.
At this point, a new practical interest rate corridor has been established, with IORB and RRP rates serving as the upper and lower limits for interest rates, respectively. The effective federal funds rate (EFFR) remains close to the IORB rate, indicating that the Fed can effectively control the federal funds rate by adjusting the IORB and RRP tool rates, even in a context of abundant reserves. The FOMC announced that it will continue to implement monetary policy under the framework of an ample reserve system for the long term.
3. The Slope of the Reserve Demand Curve
To summarize the meaning of "ample reserves": when the balance of a bank's reserves is sufficiently large such that the federal funds rate (the price at which banks are willing to trade reserves) does not exhibit significant sensitivity to daily changes in total reserves, reserves are considered ample. In other words, the elasticity of the federal funds rate to reserve shocks must be very small, allowing the Fed to focus on using IORB and RRP rates to manage market interest rates without actively managing reserve supply.
This elasticity can be represented by the slope of the blue reserve demand curve in the diagram below, which depends on the total reserve amount in the banking system. It indicates how much the federal funds rate will change with even slight changes in total reserve levels. The key point here is that as reserves decline, the slope of the reserve demand curve becomes steeper. During the "moderate" or "scarce" reserve phases, reserves are limited assets for banks; thus, interbank borrowing rates are very sensitive to changes in reserve levels. In other words, even a slight shift in reserve balances (on the horizontal axis) will lead to very noticeable changes in rates. At this time, the Fed can either buy securities to increase reserves or sell securities to reduce reserves, with small changes in the supply of reserves sufficient to move the market to different points on the demand curve, resulting in corresponding changes in market interest rates.
Above a certain level of reserve balance, we are currently in the "ample" reserve range, where the slope of the reserve demand curve is close to zero. In this area, banks' reserves are more than sufficient to meet their reserve requirements and all payment needs. It can be understood that at this point, reserves have almost no value; they are abundant, and the marginal utility of increasing or decreasing reserves approaches zero. If one must assign some value to it, it likely comes from the fact that these balances represent a fully liquid, interest-earning asset, which banks can hold as part of their general liquidity pool or investment portfolio. In the U.S., this value comes from IORB, allowing the Fed to frame the upper and lower limits of market interest rates through interest rate tools like IORB.
IV. How to Define Adequate Reserves? When Will Balance Sheet Reduction End?
The Fed implements monetary policy under an ample reserve system, primarily by managing IORB and RRP rates to control short-term interest rates. For this to work, the number of reserves in the banking system needs to be sufficiently large so that market interest rates become desensitized to daily changes in bank reserves. One way to ensure ample reserves is for the Fed to provide liquidity near the transition point where the demand curve moves from a flat section to a slope (in other words, where the slope changes from zero to negative).
One major purpose of monitoring whether reserve balances are adequate is to study the timing for the Fed to stop balance sheet reduction, which is also one of the few guidelines the Fed provides regarding its QT plan. In January 2022, the FOMC outlined several principles for reducing the balance sheet. These included: "Over time, the committee intends to maintain the amount of securities held necessary for the effective implementation of monetary policy under an ample reserve system." Fed Chair Powell stated at a press conference in December 2023 that the plan is to "slow down and stop the decline in the size of the balance sheet when reserve balances are slightly above what is considered consistent with ample reserves."
As RRP funds are depleted, bank reserves are fluctuating downward, and it seems inevitable that reserve balances will fall out of the "ample range," intensifying discussions about halting QT. However, accurately determining the transition point from ample to moderate reserves is challenging. The spike in overnight borrowing rates in 2019 indicated that the Fed could not entirely accurately assess the sufficiency of reserves. Between 2014 and 2019, the Fed reduced its balance sheet, and bank reserves dropped to about $1.5 trillion. The Fed mistakenly believed this would provide the banking system with ample reserves. It turned out to be a misjudgment, as the short-term rates in the money market surged in September 2019, with banks reluctant to lend out their reserves. In response, the Fed began expanding its balance sheet and resumed reverse repo operations at that time.
The Reserve Demand Elasticity (RDE) published by the New York Fed is a referenceable indicator. As shown in the graph, the slope was significantly negative during 2010-2011, but as the Fed injected large amounts of reserves into the banking system to cope with the global financial crisis, the slope approached zero. Between 2012-2017 and after mid-2020, reserves exceeded 13% of total bank assets, and the slope neared zero again, indicating ample reserves. When the slope tends towards negative, it is usually a time for the Fed to intervene in the market to provide liquidity, especially in September 2019. Currently, this indicator has not clearly turned negative, which is not good news for those expecting a rapid end to balance sheet reduction or even a resumption of balance sheet expansion.
Another popular viewpoint in the market is that a reserve level equivalent to 10%-12% of nominal GDP is barely considered "ample," corresponding to a value between $2.7 trillion and $3.4 trillion, with current bank reserves sitting right in this range (about $3.2 trillion). In this light, the timing for ending balance sheet reduction seems to have arrived, at least no later than the first quarter of 2025; this perspective is subjective. We can pay attention to Powell's further statements on this in the next FOMC meeting.
In summary, with the implementation of quantitative easing policies, Western central banks have developed a new method for influencing short-term market interest rates, transitioning from a scarce reserve system to an ample reserve system. This shift fundamentally changed the nature of bank reserves, rendering many common theories about reserves ineffective. In a scarce reserve system, reserves serve solely as a settlement tool, and banks primarily hold reserves to complete transactions and settlements rather than for investment or yield. In an ample reserve system, reserves not only serve for settlement but can also act as a store of value, leading banks to compare the return on holding reserves with the returns from other investments, making demand more sensitive to various interest rate tools.
V. The Mother of Liquidity - Bank Reserves
Having discussed all this, some may wonder why, despite bank reserves being so important, they do not seem to reflect in dollar liquidity (the size of the Fed's balance sheet - TGA - RRP). In fact, bank reserves are known as the mother of liquidity, and dollar liquidity is essentially two sides of the same coin; fluctuations in dollar liquidity accompany changes in reserves. This can be understood by breaking down the Fed's balance sheet: since assets and liabilities are equal, after subtracting TGA and RRP from the size of the balance sheet, only bank reserves and cash in circulation remain. These two are collectively referred to as base money (corresponding to the deposit money produced by credit mentioned earlier). Cash in circulation has changed little over the past decade; therefore, the fluctuation of bank reserves is essentially equivalent to fluctuations in dollar liquidity (Figure 12), and dollar liquidity is essentially just a more precise measurement formula.
Before 2021, the Fed's liabilities mainly consisted of three components: bank reserves, TGA, and currency in circulation. Americans prefer to use paper currency; in fact, until 2010 when bank reserves surpassed it, currency in circulation was the largest liability of the Fed. Since then, physical cash has only followed GDP's slow growth, allowing us to ignore currency in circulation and primarily focus on bank reserves and TGA. Usually, the Treasury maintains TGA close to its target level (currently $750 billion) by adjusting the number and duration of government bonds issued. However, when the debt level approaches the ceiling, the Treasury often reduces TGA to meet expenditures; when the debt ceiling is lifted, the Treasury issues additional debt to fill the TGA balance. This represents the most typical fluctuation pattern of the TGA account, directly impacting bank reserves as it is the remaining part of the Fed's liabilities. During this phase, the TGA account is also the only reservoir of bank reserves (liquidity). Keeping the TGA balance high means tightening the liquidity faucet; conversely, if the Treasury implements a stimulative budget plan, liquidity can flow back into the banking system through government spending (whether through helicopter money or targeted industry subsidies).
Entering 2021, a significant new item was added to the Fed's liabilities: RRP. In 2020, due to massive fiscal stimulus and quantitative easing, the private sector experienced significant savings. Much of this entered banks in the form of deposits, and a substantial portion also flowed into money market funds (MMFs). After the 2016 money market reform, MMFs could only invest in government-issued securities. With the Treasury issuing a large amount of government bonds to cover the 2020 deficit, MMFs flowed into the TGA by purchasing government bonds. As the pandemic pressures eased in 2021, the Treasury no longer urgently needed short-term funds and began to reduce government bond issuances. However, this quickly created problems for MMFs, which had to scramble to buy the limited newly issued government bonds, leading to overcrowding in the bond market. To address this issue, the Fed opened RRP to money market funds, reducing their reliance on government bonds.
In other words, aside from the TGA, there is now another place to "store" liquidity for bank reserves - RRP. By the end of 2021, the consistent inflow of funds into RRP from money market funds squeezed bank reserves in the Fed's balance sheet. Fortunately, due to the recent massive liquidity injection, the overall reserve level remained quite sufficient, and the repo market did not experience a crisis like in 2019. However, this process did significantly impact the prices of risk assets like U.S. stocks, with the phenomenon being more pronounced in higher-risk areas. This situation persisted until mid-2022 when the Treasury completed its plan to reduce the stock of government bonds, leading to a gradual slowdown in the inflow of funds into RRP, stabilizing bank reserves and providing breathing space for the market.
In 2023, despite the economy recovering, the Biden administration is determined to push forward on economic issues that were postponed during the pandemic; massive government plans and large deficits need to be financed by resuming large-scale bond issuances. Fortunately, money market funds still have significant "savings" in RRP, making it easy to find buyers for short-term bonds. As money market funds shift funds from RRP to the Treasury and then from the Treasury to enterprises and the private sector, space on the Fed's balance sheet is released, pushing up bank reserves. Despite rising interest rates and the Fed reducing its balance sheet at a pace of $95 billion per month, the speed of RRP depletion is even faster. The increase in bank reserves makes it easier for financial markets to rebound, providing additional dollars to purchase assets. U.S. stocks and Bitcoin have reached new highs, although this strong momentum can largely be attributed to the hype around AI and Bitcoin ETFs, the underlying premise is that excess liquidity in the U.S. banking sector has been relatively well preserved in the post-pandemic era, ensuring that there are always additional dollars waiting to be invested in the latest speculations.
Since April of this year, as RRP balances fell below 300 billion, the Treasury could not issue government bonds at the same pace as last year. Since funds are still flowing into money market funds, these funds may have no choice but to continue leaving their money in RRP, and the downward trend in bank reserves has become inevitable, with the timing of a slowdown gradually emerging in this downward trend.
