(AGENPARL) - Roma, 5 Marzo 2024 - (AGENPARL) – mar 05 marzo 2024 Mercati, infrastrutture, sistemi di pagamento
(Markets, Infrastructures, Payment Systems)
Money market rate stabilization systems over the last 20 years:
the role of the minimum reserve requirement
Number
March 2024
by Patrizia Ceccacci, Barbara Mazzetta, Stefano Nobili, Filippo Perazzoli
and Mattia Persico
Mercati, infrastrutture, sistemi di pagamento
(Markets, Infrastructures, Payment Systems)
Money market rate stabilization systems over the last 20 years:
the role of the minimum reserve requirement
by Patrizia Ceccacci, Barbara Mazzetta, Stefano Nobili, Filippo Perazzoli
and Mattia Persico
Number 46 – March 2024
The papers published in the ‘Markets, Infrastructures, Payment Systems’ series provide
information and analysis on aspects regarding the institutional duties of the Bank of
Italy in relation to the monitoring of financial markets and payment systems and the
development and management of the corresponding infrastructures in order to foster
a better understanding of these issues and stimulate discussion among institutions,
economic actors and citizens.
The views expressed in the papers are those of the authors and do not necessarily reflect
those of the Bank of Italy.
The series is available online at www.bancaditalia.it.
Printed copies can be requested from the Paolo Baffi Library:
Editorial Board: Stefano Siviero, Livio Tornetta, Giuseppe Zingrillo, Guerino Ardizzi,
Paolo Libri, Giuseppe Maresca, Onofrio Panzarino, Tiziana Pietraforte,
Antonio Sparacino.
Secretariat: Alessandra Rollo.
ISSN 2724-6418 (online)
ISSN 2724-640X (print)
Banca d’Italia
Via Nazionale, 91 – 00184 Rome – Italy
Designed and printing by the Printing and Publishing Division of the Bank of Italy
Money market rate stabilization systems
over the last 20 years:
the role of the minimum reserve requirement
by Patrizia Ceccacci*, Barbara Mazzetta*, Stefano Nobili*, Filippo Perazzoli* and Mattia Persico*
Abstract
The paper analyses the Eurosystem’s minimum reserve requirement system with averaging
provision and compares it with the mechanisms for stabilizing money market rates adopted by
other central banks (such as minimum reserves with carry-over provision, voluntary reserves with
compliance bands and the quota system). The effectiveness of different approaches to stabilizing
money market interest rates is then empirically investigated using data from nine central banks
over the time period 2002-22. The results show that in corridor systems with scarce or no liquidity
surplus, all the mechanisms examined contributed to stabilizing money market rates, although
not all mechanisms proved equally effective. In floor systems with ample excess liquidity, these
mechanisms would not be needed, since central banks can control the stance by adjusting the
policy rate at which reserves are remunerated. However, the minimum reserve mechanism could
also be used in floor systems to rapidly adjust the structural amount of liquidity in the system and
as a benchmark when tiered reserve remuneration schemes are adopted.
JEL Classification: E41, E43, E51, E52, E58.
Keywords: central bank reserves, minimum reserve requirement, money markets rates, monetary
policy.
Sintesi
Il lavoro analizza il sistema di riserva obbligatoria con mobilizzazione dell’Eurosistema e lo
confronta con i meccanismi di stabilizzazione del tasso del mercato monetario adottati da altre
banche centrali (come la riserva obbligatoria con riporto, il sistema delle riserve volontarie con
bande di adempimento e il sistema a quote). Viene poi indagata empiricamente l’efficacia dei
diversi approcci nello stabilizzare i tassi di interesse del mercato monetario, impiegando dati
relativi a nove banche centrali sull’orizzonte temporale 2002-2022. I risultati mostrano che
in sistemi a corridoio con nullo o limitato eccesso di liquidità tutti i meccanismi considerati
hanno contribuito a stabilizzare i tassi del mercato monetario, sebbene con un diverso grado
di efficacia. In sistemi floor con ampio eccesso di liquidità questi meccanismi non sarebbero
necessari, in quanto la banca centrale può controllare l’orientamento di politica monetaria
mediante la fissazione del tasso di remunerazione delle riserve. Tuttavia, anche in un sistema
floor il meccanismo di riserva obbligatoria potrebbe essere utilizzato per aggiustare rapidamente
l’ammontare strutturale di liquidità nel sistema e come parametro di riferimento in caso di
adozione di un metodo di remunerazione a più livelli.
Bank of Italy, Directorate General for Markets and Payment Systems.
CONTENTS
1. Introduction
2. The Eurosystem’s minimum reserve requirement
2.1 Features
2.2 Shortcomings
2.3 Minimum reserve with tolerance band
3. Alternative reserve systems adopted by other central banks
3.1 Carry-over of reserve balances
3.2 Voluntary reserve
3.3 Quota system
3.4 Operational frameworks that do not include stabilization mechanisms
based on reserves/quotas
4. Empirical evidence
4.1 Robustness check
5. Comparison of various stabilization systems
6. Conclusions
References
Appendix
Appendix I Voluntary reserve and quota systems
Appendix II Review of systems adopted by central banks
Appendix III Analysis of relationship between central bank reserves
and money market rates
1 Introduction1
The monetary policy frameworks of central banks generally include mechanisms aimed at facilitating
the stabilization of money market rates (usually overnight). The volatility of the overnight money
market rate affects the achievement of monetary policy targets for at least two reasons. First, an
extreme variation of this rate with respect to the monetary policy reference rate can hinder the correct
functioning of the transmission mechanism, reducing the central bank’s ability to influence financial
conditions. Second, high volatility increases uncertainty, which generates higher costs for banks in
managing liquidity, and, in extreme cases, can dissuade them from participating in the market.
These stabilization mechanisms vary according to the characteristics of the overall operational
framework, such as, for example, the frequency of monetary policy operations carried out by the
central bank, the type of eligible counterparties, the monetary policy reference rate, the use of a
corridor or floor system.
In the Eurosystem, one of the tools that contributes to stabilizing money market rates is the minimum
reserve requirement, which requires each credit institution resident in the euro area to hold a
compulsory deposit with its national central bank (NCB) as a percentage of its funding. Such
obligation can be fulfilled on average over the maintenance period (reserve mobilization). Under a
corridor system, the minimum reserve requirement creates (or increases) a structural liquidity
shortage, allowing the central bank to control money market rates more easily through periodic
refinancing operations. However, the unconventional monetary policy measures2 implemented in the
last fifteen years have led to a structural liquidity surplus which diminishes the relevance of the aims
pursued through the minimum reserve requirement.
This paper: i) analyzes the Eurosystem’s minimum reserve system, the evolution of its role over time
and in the current context (paragraph 2); ii) describes the reserve systems adopted by other central
banks, contextualizing them in the overall operational framework of monetary policy and assessing
their main advantages and disadvantages (paragraph 3); iii) carries out an empirical analysis of the
effectiveness of the various mechanisms for stabilizing money market rates (paragraph 4); iv)
compares alternative stabilization mechanisms, also taking into account different liquidity scenarios
that could take place in the coming years (paragraph 5).3 Paragraph 6 shows the main conclusions.
The study is completed by three appendices, which provide empirical evidence and theoretical
elements on the various stabilization mechanisms adopted by central banks in the last 20 years.
Any views expressed in this paper are the authors’ and do not necessarily represent those of the Bank of Italy. We are
grateful to Luigi Cannari, Gioia Cellai, Alberto Locarno, Salvatore Nasti, Franco Panfili, Stefano Siviero, Livio
Tornetta and one anonymous referee for valuable comments and suggestions.
In recent years, the ECB has enriched its operational framework with many non-standard operations implemented in
order to address the 2008 financial crisis (monetary policy operations with “fixed rate full allotment”, LTRO), the
deflationary pressures starting from 2015 (TLTRO, APP, PSPP) and the consequences of the Covid-19 pandemic from
2020 (TLTRO III, PEPP).
Looking at the ECB, see Monetary policy decisions, December 2022: “By the end of 2023, the Governing Council
will also review its operational framework for steering short-term interest rates, which will provide information
regarding
endpoint
balance
sheet
normalization
process”.
https://www.ecb.europa.eu/press/pr/date/2022/html/ecb.mp221215~f3461d7b6e.en.html. See also the speech by
Christine Lagarde at the Hearing of the Committee on Economic and Monetary Affairs of the European Parliament,
September 2023: “We aim to conclude this review [of the operational framework] by spring 2024”.
https://www.ecb.europa.eu/press/key/date/2023/html/ecb.sp230925~036083efca.en.html
2 The Eurosystem’s minimum reserve requirement
2.1 Features
In general, minimum reserve systems differ according to the following main elements: a) the
liabilities that are subject to the reserve requirement; b) the reserve ratio; c) the reserve maintenance
period, (i.e. the number of days with respect to which the compliance with the obligation is assessed);
d) the time lag between the maintenance period and the period over which the amount of the reserve
is calculated; e) the methods for fulfilling the obligation (punctual daily compliance, on average over
the maintenance period or within a tolerance band); f) the remuneration of the reserve; g) the
application of penalties in the event of non-compliance.
The Eurosystem’s minimum reserve system was introduced in the third stage of European Monetary
Union (EMU,1 January 1999). Its legal framework is based on Article 19 of the Statute of the ESCB
and on various Council and ECB Regulations. The system was designed taking into account that not
all NCBs had used a similar mechanism before the start of EMU.
The value of the minimum reserve is determined by applying specific coefficients to the so-called
“reserve base”, defined in relation to some bank balance sheet items, reported at the end of the second
month preceding the start of the maintenance period to which the reserve relates.
In more detail, the reserve base includes bank liabilities in the form of deposits and debt securities
issued; liabilities vis-à-vis the ECB, the Eurosystem NCBs and the banks subject to reserve
requirement are excluded. The applicable coefficients are currently 0 for liabilities with a maturity
above 2 years and 1 per cent for the remaining items of the aggregate. All banks can deduct a lumpsum allowance of 100,000 euros from the reserve due, introduced in order to reduce the operational
burden of managing the mechanism for small banks. To this end, the legislation provides for the
possibility of fulfilling the obligation indirectly, through an intermediary bank, also subject to the
minimum reserve.
The reserve requirement is deemed to be fulfilled when the average of the daily balances of the reserve
accounts held by each bank during the maintenance period is equal to or greater than the value of the
reserve requirement.
In the design of the mechanism, the objective of stabilizing money market rates is pursued through
three elements:
1) The possibility to fully mobilize the reserve, associated with the average compliance
mechanism. This allows the banking system to manage its liquidity in a flexible manner,
absorbing the daily fluctuations in liquidity needs and taking advantage of short-term arbitrage
opportunity in the money market.
2) The setting of a sufficiently high reserve ratio in order to force banks to have a liquidity buffer
that can be used to absorb any liquidity shocks in the market, once settlement needs are met.
3) The setting of a sufficiently long maintenance period (currently, on average, 6-7 weeks), to
allow banks to perform the balancing function of rates, and with a starting date coinciding
with the settlement day of the main refinancing operation (MRO) following the Governing
Council meeting dedicated to monetary policy decisions. The latter arrangement ensures that
the change in the ECB’s policy rates generally takes effect only starting from the next
maintenance period, thus avoiding expectations of changes in official interest rates during the
maintenance period.
In particular, the possibility to mobilize reserves gives banks some flexibility in determining how to
meet their liquidity needs. In general, banks will try to hold more reserves on days when they expect
the market interest rate to be lower than the expected future interest rates and fewer reserves on days
when they expect the rate to be higher. In other words, the reserve mobilization within a maintenance
period tends to flatten the demand curve at least in the days preceding the last day of the maintenance
period and thus makes it easier for the central bank to control the variability of the money market rate
around the policy rate.
With regard to the other characteristics of the Eurosystem’s reserve regime, both a remuneration
system and the application of penalties in the event of non-compliance are featured. The remuneration
of minimum reserve was set at 0% starting from September 2023 to improve the efficiency of
monetary policy.4 Such decision has, indeed, reduced the overall amount of interest that needed to be
paid on reserves in order to implement the appropriate monetary policy stance. At the same time, it
has preserved the effectiveness of monetary policy by maintaining the current degree of control over
the stance and ensuring the full pass-through of interest rate decisions to money markets.
In the event of non-compliance, a financial penalty is imposed, calculated by applying a penalty rate
2.5 percentage points higher than the average interest rate of the marginal lending facility to the
amount of the non-compliance. There is also a higher penalty in the event of repeated noncompliance.
In the operational framework of the euro area, the minimum required reserve was combined with a
symmetrical corridor system around the policy rate, with upper and lower limits being the rates of the
two standing facilities: the marginal lending facility and the deposit with the Eurosystem. An analysis
of the trend in money market rates from 2002 to 2022 (see Appendix II5) shows that, in a period
characterized by a liquidity deficit, the mechanism proved to be quite effective in stabilizing money
market rates around the policy rate, with the exception of the last days of the maintenance period.
Since November 2005, the Eurosystem has regularly used fine-tuning operations to offset unexpected
liquidity changes on the last day of the maintenance period. Subsequently, with the structural excess
of liquidity created by non-standard operations, the system has de facto turned into a floor system,
where the overnight money market rate is aligned to the deposit facility rate, and the minimum reserve
requirement lost its original function.6
ECB Governing Council decision on 27 July 2023. Before that date, the rate was equal to the weighted average rate
of the main refinancing operations carried out during the maintenance period, up to 20 December 2022, and it was
equal to the weighted average of the deposit facility rate starting from 21 December 2022.
Appendix II presents an overview of what has happened in the main central banks over the last 20 years, highlighting
how the various operational frameworks have changed in a context characterized by a structural excess of liquidity.
Moreover, it analyzes the effectiveness of monetary rate stabilization mechanisms through specific indicators.
In the period from October 2019 to September 2022, minimum reserve requirement worked as a benchmark for the
remuneration of excess reserves. With the negative rates and the growth of excess liquidity, ECB established a twotier system (TTS) for remuneration of excess reserves, which exempted part of the banks’ excess liquidity holdings,
equal to a multiple of the required reserve, from the application of the negative interest rate of the deposit facility.
Following the increase above zero of the rate on deposits with the ECB, the system became unnecessary and was
revoked by the Governing Council at its meeting on 8 September 2022.
2.2 Shortcomings
Over the years, starting even earlier than the 2008 global financial crisis, a number of possible
improvements to the Eurosystem’s minimum reserves have been identified.
1) As empirical evidence shows, in a structural liquidity deficit regime the mechanism requires
the use of additional instruments to prevent volatility at the end of the maintenance period.
This issue, in theory, could be eliminated through mechanisms used by other central banks,
such as compliance with the obligation within a tolerance band or the possibility of carrying
over a liquidity balance to the next maintenance period. These features are analyzed in
paragraphs 2.3 and 3.1. On the other hand, in a situation of structural excess of liquidity, such
as the present one, the mechanism becomes not (strictly) needed, against the persistence of
operational costs for central banks and intermediaries.
2) The aggregate subject to minimum reserves (or reserve base, i.e. the amount to which the
ratios for calculating the obligation apply) is made of balance sheet items (liabilities with a
maturity of up to two years), which do not necessarily reflect the liquidity needs of banks for
the purpose of fulfilling their payments. Therefore, wide divergences can arise between the
reserve base and the banks’ real liquidity needs, which would require a re-composition of the
items or recalibration of the coefficients. We considered three variables as proxies of the
liquidity needs of each bank: the transactions on the TARGET2 payment system, the volatility
of deposits and the width of swings in excess liquidity deposited with the central bank.7 We
analyzed the relationship between the reserve base and these proxies of the liquidity needs.
The results show non-significant coefficients for all the considered variables for the sample
between the fourth maintenance period 2021 and the fourth maintenance period 2022 (see
Table 1).
Table 1 – Relationship between reserve base and TARGET2 transactions, excess liquidity swings,
bank’s deposits volatility
3) The mechanism can be seen as an inefficient imposition in times of crisis, when the banking
system may have to resort to the central bank to obtain funds by pledging collateral and a
shortage of guarantees may arise. In such cases, banks may wish to use the buffer constituted
These variables are calculated for each bank as the average, per maintenance period, of the sum of daily inflows and
outflows, the standard deviation of the monthly variation rate of bank’s deposits (households and non-financial
corporations), the difference between liquidity deposited with the central bank (in excess of the minimum reserve
requirement) at the end of the month and its average of the last 12 months, respectively.
by the minimum reserve, but this would result in non-compliance and would be subject to a
penalty. This inefficiency is accentuated by the fact that in the euro area the minimum reserve
cannot be included in the LCR numerator, and so it does not contribute to compliance with
the aforementioned liquidity requirement, which becomes more difficult in times of market
tension. During a crisis, the problem could be solved by setting to zero the ratio (in the euro
area, this is always possible through a decision of the ECB Governing Council), which would
also facilitate the fulfilment of the prudential liquidity requirements imposed by the
Supervisory Authority.8
4) The mechanism is operationally burdensome for banks. In fact, the fulfillment on average
compels banks to keep track of their fulfillment path; in addition, the shock absorbing capacity
decreases as the end of the maintenance period approaches. Various studies (Bindseil, 2016
and 2018; Perez – Quiros and Rodriguez, 2016) show that the mechanism based on the optimal
reserve fulfillment path in a maintenance period is far from obvious, even under simplified
model assumptions. In fact, each day of the maintenance period is different, and the path of
the interest rate during the remainder of the maintenance period depends on past events within
the maintenance period.
Before comparing the Eurosystem’s minimum reserve requirement with other tools with the same
aim adopted by other central banks, it is useful to consider that, in the past, minimum reserve
requirements were also imposed for reasons other than stabilizing monetary rates, such as prudential
and monetary control (S. Gray, 2011).9 With reference to the prudential objective, the minimum
reserves provided some protection against the liquidity and solvency risks of the banks. However, the
prudential benefits of such mechanism are now obsolete, as the same objectives are more effectively
achieved through the combination of supervision and regulation (with related capital adequacy and
liquidity requirements), deposit insurance and central bank financing. Furthermore, the prudential
role is substantially weakened where the fulfillment of the reserve requirement on average is allowed.
Similarly, the use of the reserve for monetary control purposes (exercised through two channels: the
money multiplier and the impact on the spread between deposits and lending rates) is no longer
widespread among central banks of world’s advanced economies. However, if there are excess
reserve balances in the economy, central banks may consider increasing minimum reserve
requirements as a cheap way of sterilizing the impact of the surplus. Recently, De Grauwe and Ji
(2023a, 2023b, 2023c) advocated the raising of minimum reserve requirements in the euro area as a
tool to avoid the excessive subsidies to banks following the increases in monetary policy rates.
According to the authors, the introduction of a two-tier system whereby part of the bank reserves
In April 2020, following the outbreak of the Covid 19 pandemic crisis, in order to support the flow of credit to the
real economy, a supervisory measure was introduced, which allowed banks to temporarily operate below the liquidity
level defined by the LCR. Since the establishment of the Eurosystem, the minimum reserve ratio has been changed
only once in 2011, when it was lowered from 2 to 1 percent.
This was also the case in some of the countries that joined the Eurosystem. Della Valle et al. (2022) recall that
minimum reserve requirement could be used as a macroprudential tool to contain systemic risks from both broadbased credit growth and systemic liquidity shocks. However, the authors highlight that now the central banks and the
macroprudential authorities have other options to manage these risks. Specifically, the countercyclical capital buffer
offers an effective alternative to the minimum reserve requirement to smooth credit cycles and central banks can buy
or sell marketable securities, manage the issuance of their own securities, and deal in more complex instruments such
as cross-currency swaps markets (e.g., New Zealand) to manage the systemic liquidity shocks. These market-based
instruments are less likely to be distortive than the minimum reserve requirement.
would be converted into unremunerated minimum reserves can reduce the banks’ equity channel
effect, contributing to enhance the effectiveness of monetary policies.10
2.3 Minimum reserve with tolerance band
Besides the mobilization of reserves within a maintenance period, another instrument for the central
bank to control the volatility of money market rates around the policy rate is the introduction of
tolerance (or compliance) bands. In this case the penalizing rates do not apply when the balance (daily
or calculated as an average of the maintenance period)11 falls below or exceeds the level of the
required reserves, but only where the balance deviates by a percentage greater than the tolerance band,
for example +/- 10 per cent around the exact value of the reserve requirement. The tolerance band
may be associated with a minimum reserve or with a voluntary reserve (see paragraph 3.2): for
example, it was used by the Bank of England in conjunction with the voluntary reserve from 2006 to
2009. The wider the fulfillment band, the lower the sensitivity of money market interest rates to
liquidity shocks is. This implies a reduction in interest rate volatility (Armenter, 2016). However,
very large compliance bands reduce the incentive for banks to exchange funds (Armenter, 2016; Lee,
2016; Ennis and Keister, 2008).
3. Alternative reserve systems adopted by other central banks
This paper analyzes nine central banks (in addition to the ECB, the central banks of Australia, Canada,
Japan, Norway, New Zealand, the United Kingdom, the United States and Sweden). Among these,
only the Eurosystem and the Bank of Japan currently use a minimum reserve system with
mobilization. Over the past twenty years, some central banks have adopted reserve systems with
different characteristics from those of the Eurosystem, as shown in the following sub-paragraphs.
In the first article, De Grauwe and Ji (2023a) suggested transforming all the existing bank reserves into unremunerated
minimum reserves. As a consequence, banks would have a larger proportion of their balance sheet in the form of
assets that have no return. In order to restore their overall interest spread (the difference between the interest earned
on their assets and the interest paid on their liabilities), they would have to increase the interest rate they apply on their
loan portfolio: this would lead to a generalized increase in interest rates, contributing to the central banks’ strategy to
fight inflation. At the same time the increase in the minimum reserve requirement would facilitate the return to a
reserve scarcity regime, such as the one before the financial crisis. With this proposal, however the central bank would
lose the opportunity to use the interest rate on bank reserves as policy rate. To avoid this effect, in following articles
(De Grauwe and Ji, 2023b and 2023c) the authors propose an increase of non-interest-bearing minimum reserve
requirements on part of bank reserves. This would reduce the excessive subsidies to banks while maintaining the
current operating procedure used by central banks. According to the authors, by imposing unremunerated minimum
reserves the central bank could reduce the positive equity effect of a rate hike. A such use of the minimum reserve
requirement, however, would have the negative consequence of penalizing more banks with higher amount of
deposits, with possibly unintended distributional consequences on the banking sector (Deuber and Zobl, 2023; Kwapil,
2023). Furthermore, it can create incentives for banks to erode the reserve base in order to circumvent the measure
(McCauley and Pinter, 2024).
Tolerance bands may be applied to the balance calculated both daily and as an average of the maintenance period. In
the first case, the band would be an alternative to mobilization. In the second case, the two systems can coexist: the
average level of the balance in the maintenance period may deviate from the exact value of the requirement up to the
amount of the band.
3.1 Carry-over of reserve balances
Carry-over was a feature related to reserve requirements employed in the United States by the Federal
Reserve (Fed) from 1981 to 2012 (see Y. C. Obrien 2007). The carry-over provision allows banks to
avoid non-compliance with the reserve requirement if they register a shortfall in a maintenance
period, since they may carry-over the reserve balances – in excess or in deficit – to the following
maintenance period, where they are to be used or recovered. The carry-over facilitates banks’ liquidity
management by providing a buffer to partially compensate for unexpected payment flows, with the
likely effect of reducing overnight interest rates fluctuations, especially at the end of the maintenance
period.
However, the carry-over provision can also be critical: firstly, it complicates central bank monitoring
of reserve requirements; secondly, it could lead to speculative behaviors by banks on the occasion of
policy rate decisions. For example, if banks expect policy rates to rise in the following maintenance
period, they will tend to carry forward a higher balance than required to the future; similarly, if they
expect rates to fall, they tend to carry forward a balance lower than the obligation to the future.
3.2 Voluntary reserve
In a voluntary reserve (VR) system, each bank commits in advance to a target for reserves at the
central bank, taking into account a number of factors such as preference for liquidity, the
characteristics of the collateral framework, the regulatory requirements and the reserve remuneration.
The central bank still maintains the control of the aggregate reserves by setting a range for individual
targets, within which each bank must place its own target. Within a VR system, the central bank may
use tolerance bands (or compliance bands, see paragraph 2.3) to better control the variability of rates
around the policy rate.
To date, the VR system has only been applied by the Bank of England (BoE), which adopted it in
May 2006 and then suspended it in March 2009, due to the excess liquidity provided through the
activation of the quantitative easing (QE) programme.12 The BoE’s reserve system also included
tolerance bands (Bank of England, 2006; Clews, 2005)13, as well as a narrowing of the corridor on
A system that had some features of voluntary reserves was the Fed’s Contractual Clearing Balance (CCB), an
instrument which supplemented the minimum reserve system with carry-over provision. The CCB was an additional
period-average amount that a depository institution (DI) could voluntarily contract to maintain with a Federal Reserve
Bank in addition to its required reserve balance held in its account with the central bank. It was remunerated in the
form of credits, which could be used to offset charges of Federal Reserve services, such as check clearing, and Fedwire
transfer of funds and securities. The amount of the CCB had to be determined before the start of a maintenance period.
A DI’s daily average of the clearing balances in a given maintenance period had to be within a tolerance band. The
CCB system reduced the likelihood that a bank would have to recur to the Fed’s overnight overdrafts. Indeed, in order
to ensure the proper functioning of the payment system, the Fed allows banks to use daily overdrafts which, if not
reset to zero at the end day, become overnight overdrafts. This exposure can be risky for the Fed, as it is not mandatory
to collateralize this credit. Therefore, with the aim of discouraging banks from using the overnight overdraft, the Fed
applies a very high penalty rate, equal to the primary credit rate + 4% for the first three days of use during the year;
from the fourth day onwards, 100 basis points are added for each additional day of use. Sometimes, the additional
requirement could be imposed by the central bank (required clearing balance, RCB), rather than being voluntary, if a
DI had frequent overdrafts (during the day or overnight). For more details, see M. Naber, R. Sambasivam, and M-F.
Styczynski (2017).
Prior to September 2007, the BoE’s tolerance band was limited to only ±1 percent around the target. During periods
of market tension, the BoE widened the band to ±60 percent and offered additional liquidity above banks’ aggregate
the last day of the maintenance period (from ±100 bps to ±25 bps). In the BoE system, if the effective
reserves, as an average of the maintenance period, were positioned within the tolerance bands, they
were remunerated at the BoE’s official interest rate (known as the bank rate).
In the case of reserves outside the tolerance band (above the upper limit or below the lower limit),
remuneration was subject to a penalty: excess reserves were remunerated at the bank rate minus a
spread, while banks in deficit would have to borrow the missing part from the BoE at the bank rate
plus the same spread. The remuneration scheme acted similarly to a corridor system, with the
remuneration rate for excess reserves acting as a floor and the penalty rate for deficiencies acting as
a ceiling (Armenter, 2016; Berentsen and Monnet, 2008).
Banks had to specify their (non-negative) reserve targets at least two business days before the start of
each maintenance period, within individual ceilings. The latter were determined by the BoE twice a
year and were equal to the higher between £1 billion and two per cent of average eligible liabilities
in sterling in the six months preceding May and November.14
In a series of papers, Baughman and Carapella (2019a, 2019b) proposed the adoption of this tool for
the Fed’s operational framework, as an alternative to a system based exclusively on the floor system.15
From the bank’s point of view, this system is more flexible than minimum reserve requirements. If
market stress leads to an increase in banks’ demand for reserves, banks may adjust their reserve target
for the following period. Awareness of being able to meet changing liquidity needs from one period
to the next can help reduce liquidity premiums and make banks more inclined to trade in the interbank
market. Furthermore, compared to a system of required reserves, the VR system allows banks with
different business models to “correctly size” their reserve requirements, which would otherwise be
fixed according to the same formulas for all banks.
From the point of view of the central bank, too, the system appears advantageous in theory: the central
bank, in fact, can still target an aggregate level of reserves by setting maximum and minimum limits
for the individual banks commitments and can foresee more easily the liquidity demand as banks
reveal their expectations of liquidity needs (Winters, 2012). As long as banks have a fairly good
understanding of the factors affecting their reserve holdings, VRs eliminate the need for the central
bank to continuously adjust supply in order to meet the interest rate target. Banks would adjust their
targets to respond to the expected change in supply, leaving market rates unchanged and neutralizing
the shock, thus making the central bank intervention necessary only in exceptional cases. This reduces
administrative and operational costs associated with frequent fine-tuning operations. Furthermore,
reserve targets through its refinancing operations. These measures provided banks with additional flexibility in their
liquidity managing, which contributed at least in part to reduce the monetary rate volatility (Lee, 2016).
Eligible liabilities included: deposits with an original maturity of up to 2 years (sight deposits, time deposits and
certificates of deposit); commercial paper and other issues with an original maturity of up to 5 years; reverse
repurchase agreements (repo); other residual and pending items.
For a similar proposal for the Fed see Filardo A. (2020). This proposal is based on “strengthening a Corridor-Type
Framework with a Sequestered Reserves Rule (SRR)”. Under the SRR, monetary and supervisory authorities should
regularly negotiate with financial institutions the appropriate level of reserves needed to meet liquidity regulations.
Being solely centered on financial stability purposes, the negotiations would focus on the objectives of avoiding
excessive accumulation of reserves and incentivizing the use of reserves as part of a well-diversified basket of high
quality liquid assets (HQLA). According to the author, the negotiations would also aim to clarify the intentions of the
authorities and encourage a better mutual understanding of each other’s expectations.
the VR system offers the central bank flexibility in choosing its counterparties: the system can also
be applied to intermediaries with a different liability structure than banks.
VRs allow central banks flexibility in determining the size of their balance sheets, regardless of their
interest rate targets, and encourage money market activity, regardless of the amount of reserves held
by each individual bank. From a theoretical point of view, Baughman and Carapella (2019b) show
that banks will be induced to set VR targets that bring the expected money market rate to the target
reserve rate, determined by the central bank, provided that the remuneration function of reserves is
additively separable into a portion depending on the reserve target chosen by each bank and into a
portion depending on the difference between the reserves held and the target (for further details, see
Appendix I16). Under this condition, the authors show that the VRs deliver the condition of separation
between interest rate control and liquidity provision. Thus, by simply setting the rate on target
reserves, a central bank moves the money market rate, without the need for frequent refinancing
operations, having a simple and flexible system (such as a floor system), but still providing an
incentive for money market activity (a feature not present in a floor system).17
If banks expect a shortage of reserves, i.e. the money market rate will be higher than the VR rate,
they will reduce their targets and have an incentive to lend on the market, thus eliminating the scarcity
phenomenon. On the other hand, if they expect reserves to be abundant, i.e. the money market rate
will be lower than the VR rate, they will raise their targets and borrow from the market. The only
assumption consistent with an equilibrium is that expected money market rate is equal to the VR rate
(Armenter, 2016; Baughman and Carapella, 2019b).
In the event of unexpected changes in the autonomous factors leading to a deviation between the
aggregate reserves in the system and the aggregate targets, with the money market rate drifting away
from the VR rate, the central bank wishing to control rate volatility could inject or drain reserves. In
this regard, Baughman and Carapella (2019a) argue that, in order to inject or drain liquidity into the
system without affecting money market volatility, it is important that the central bank credibly
announces the details of the intervention before banks choose their targets or, alternatively, when
more frequent operations are planned, allows banks to adjust their targets more often. The key
condition is that the injection or drainage of liquidity is credible and fully anticipated by the banks.
Lee (2016) points out that, although the BoE’s VR system has largely succeeded in controlling rate
volatility, the average money market rate has consistently been above the BoE’s target. Baughman
and Carapella (2019b) show that this persistent deviation of market rates from the voluntary reserve
rate is mostly attributable to the endogeneity of the targets with the tolerance bands, envisaged by the
BoE system. They suggest to solve the problem by providing tolerance bands that are not linked to
the targets (for example, setting them on the basis of historical average values of total assets or other
balance sheet items). In fact, a remuneration function that violates the condition of being additively
separable can provide perverse incentives, driving money market rates away from the VR rate. In
Appendix I provides analytical details on reserve systems other than minimum reserve requirements.
By way of comparison, the separation condition would not be satisfied by reserve requirements, since under this
system the central bank should necessarily increase the supply of reserves to reduce money market rates, autonomous
factors being equal. The separation condition is instead guaranteed by a floor system without reserve requirements,
where the central bank controls market rates by varying the rate of the deposited reserves. However, this is possible
only if the supply of liquidity exceeds demand (see Keister, Martin and McAndrews, 2008).
particular, with a remuneration function where tolerance bands are based on the VRs, the expected
market rate exceeds the target rate due to the banks’ optimal targeting strategy.18
The experience of the BoE shows that even this system, which in theory appears very flexible and
effective, has some critical issues that are listed below.
1) In event of a crisis, it does not seem to improve the liquidity forecasts made by the central
bank, even though it can rely on the preferences revealed by the banks. Indeed, crisis are
typically the result of aggregate shocks and therefore individual intermediaries do not have an
information advantage over the central bank. The BoE’s experience at the onset of the global
financial crisis suggests that banks’ predetermined reserve targets turned out to be lower than
their actual needs, resulting in increased money market volatility, which compelled the BoE
to conduct frequent liquidity providing operations and to increase the tolerance bands around
the targets.
2) It implies the operational burden for the central bank of setting the range of accepted voluntary
reserve targets, which requires careful maintenance, having to be fine-tuned also according to
the banks’ financial structure (e.g. taking account that some institutions, such as settlement
banks, have higher liquidity needs).
3) It requires adjustments with two types of unconventional measures, such as QE and negative
rates. Voluntary reserves could be very difficult to manage in a situation of large excess
liquidity, as an imbalance could be created between the large supply of reserves and the
demand for liquidity by banks, which would find it increasingly difficult to calibrate their
reserve targets. In order to prevent this from happening, the communication by the central
bank is essential, as well as the possibility of allowing, in exceptional circumstances, banks
to change their targets even during maintenance periods. The VR system would also require
arrangements for managing negative interest rates. With negative rates and a large supply of
liquidity from the central bank, it would be necessary to introduce, also in this system, a
concept of excess reserves and possibly a new (negative) deposit rate at which they are
remunerated.
3.3 Quota systems
An alternative to reserves (required or voluntary) is represented by quota-based systems. In these
systems, the central bank sets for each counterparty the amount of reserves remunerated at the official
rate that can be deposited and/or withdrawn from it (Target rate limited access, TARALAC).
Individual deposit or funding quotas are allocated on the basis of certain characteristics of banks, such
as balance sheet size or volume of transactions in the payment system. In a system with individual
deposit quotas, any amount of reserves placed on deposit above the allocated quota each day by banks
is remunerated at the deposit rate, which is lower than the official rate. Similarly, in a system with
individual funding quotas, the use of central bank borrowing in excess of the allocated quota is paid
If the width of the tolerance band was proportional to the targets that banks set themselves, they would have an
incentive to set higher targets than they would otherwise have done, with the aim of increasing the width of their
tolerance band to reduce the risk of being outside it.
at the marginal lending rate, which is higher than the official rate. By limiting recourse to the central
bank, banks’ activity on the money market is encouraged: the official rate influences market
conditions and the overnight money market rate should be close to the official rate (for more
theoretical details, see Appendix I). Quotas can be one-sided (limited access to only one of the two
facilities) or two-sided (limited access to both facilities).
Systems based on “one-sided deposit quotas” were used by the Reserve Bank of New Zealand
(RBNZ) from 2007 to 2020 and by Norges Bank from 2011 to the present.
However, defining the optimal TARALAC quota is rather complex and not obvious because there is
a trade-off between the volatility of the target operating rate and money market activity. The higher
the quota, the more likely it is to be sufficient to absorb shocks and the less likely it is that banks will
have to resort to one of the two standing facilities at penalizing rates. The overall effect is a more
stable overnight rate. If the stability of the overnight rate is the central bank’s sole criterion of choice,
it will be more easily pursued with a high quota; however, in this case, there will be less incentive for
banks to trade on the money market. If, on the other hand, the priority is not to dry up the money
market excessively, the quotas must be low or unilateral. Bindseil (2016) leads simulations showing
that, under specific parameter choices, when open market operations are less frequent (each operation
has to cover more days), a larger quota stabilizes the overnight rate without further penalizing money
market turnover. In addition, the determination of the optimal quota level must take into account all
the liquidity requirements in force in that jurisdiction, especially in the short term.
Compared to the minimum reserve requirements with mobilization, this system appears flexible for
banks, as it eliminates the complexity of having to consider their fulfillment path during the
maintenance period.
The RBNZ was the first central bank to introduce a quota-system and it did so in a period of positive
rates (Bowman et al., 2010). It was introduced to counter the sharp decline in interbank market trading
and as a response to the previous excessive hoarding of bank reserves, experienced under the RBNZ’s
floor system.19
New Zealand maintained the quota system until March 2020, when it was suspended following the
implementation of a QE programme involving a very large reserve supply, which made it increasingly
difficult to keep short-term rates anchored to OCR. As a result, the RBNZ decided to remunerate all
deposited balances at the OCR rate and remove all tiering limits beyond which a lower remuneration
rate applied (Reserve Bank of New Zealand, 2020), thereby formally introducing a floor system.
Similarly, Norges Bank has implemented a one-sided (limited access to deposit) quota-based system
since October 2011, where each bank’s quota is determined to be sufficient to meet daily settlement
Under the quota system, banks’ overnight deposits were remunerated at the reference rate (the Official Cash Rate,
OCR) up to the limit set for each individual bank by the central bank (“tier”), mainly on the basis of the likely
settlement needs in the case of a tail event. Balances held in excess of the limit accrued interest at a rate equal to the
OCR minus 100 basis points. The relatively low rate of excess balances discouraged holding balances in excess of the
limit, thus incentivizing banks with excess balances to lend them in the interbank market and to invest in liquid
instruments other than central bank reserves. For more details on the evolution of the RBNZ’s operational framework,
see Appendix II.
needs and it is remunerated at the reference rate.20 Norges Bank normally reviews the level of the
quota allocated to each bank twice a year (Norges Bank, 2021).21
The Norwegian central bank maintained the quota system even after the outbreak of the pandemic
crisis, unlike New Zealand, but increased its open market operations. Following the crisis, the central
bank conducted several extraordinary financing operations with maturities of up to 12 months; the
injected excess liquidity was drained from the banking system using daily deposits with a one-day
maturity, with rates close to the policy rate.
The aforementioned central banks did not appear to face unexpected fluctuations in autonomous
factors that led to large liquidity surpluses and overnight rate decreases or liquidity shortages that
gave recourse to the credit facility. In any case, when these occurred, they intervened by draining or
supplying reserves, on a daily basis if necessary. For the Eurosystem, where autonomous factors have
shown higher volatility, a larger quota may be needed, which would then affect interbank market
activity.
The main disadvantages of the quota system are:
1) the inherent complexity of quota setting, as there is a trade-off between the precision of
steering short-term market rates toward the key policy rate and money market activity;
2) the need for the central bank to be ready to intervene on a daily basis in the event of an
imbalance due to unexpected autonomous factors, even if the calibration works well;
3) the inadequate functioning of the system in the presence of autonomous factors subject to
frequent and unexpected fluctuations. Central banks facing high variability in autonomous
factors are likely to have to calibrate very large quotas for the system to function well,
resulting in negative impacts on money market trading.
3.4 Operational frameworks that do not include stabilization mechanisms based on
reserves/quotas
Some central banks (Australia, Canada, New Zealand, Sweden) have adopted operational frameworks
that do not provide for either reserve mechanisms or quotas, having equally succeeded in effectively
controlling short-term interest rates (Sellin and Sommar, 2014).
The reserves of each individual bank above the pre-determined quota are remunerated at a penalizing rate (reserve
rate, 100 bps below the policy rate), in order to incentivize banks to lend liquidity in the interbank market. At the same
time, Norges Bank aims to keep total reserves in the banking system (around NOK 35 billion, with flexibility of ±
NOK 5 billion) below the sum of the quota that can be remunerated at the policy rate (NOK 45 billion), so that banks
can always deposit outstanding reserves in the banking system at the policy rate. Outstanding liquidity is controlled
through open market operations (drainage or supply of reserves), where the interest rate is determined by auction. See
Appendix II for more details.
The quotas are determined on the basis of a top-down approach. First, Norges Bank determines the total quota of bank
reserves remunerated at the official rate. Second, banks are divided into three groups based on data from the Norges
Bank’s Settlement System (NBO). The aggregate quota of each group is determined by the share of the respective
groups in the total assets. All banks in a group are assigned the same quota. Banks that also settle payments for other
banks are allocated an additional quota determined by their size in relation to the size of the banks for which they
settle payments.
These countries have adopted a narrow and symmetrical corridor of standing facilities around the
policy rate to achieve control of money market rates, with daily open market operations. Overnight
overdrafts are automatically accounted for as central bank loans at the marginal lending rate.
This framework has the following criticalities: 1) a significant operational burden for the central bank
due to the necessary use of daily operations; 2) in the event of substantial changes in autonomous
factors, high intraday rate volatility pending the daily central bank intervention; 3) central bank
intermediation might substitute market trading activity and reduce money market volume noticeably.
A reserve/quota mechanism becomes superfluous in an operational framework based on a floor
system. In this case, theoretically, the central bank limits interest rate volatility by setting the deposit
facility (DF) rate as the reference rate for monetary policy and injecting an amount of reserves into
the system such that the liquidity supply curve intersects the demand curve at the DF rate when the
demand curve is inelastic (i.e., at its flat right-hand portion). In practice, the DF rate often represents
a “leaky floor”, both because non-bank financial intermediaries do not have access to the DF and
because leverage requirements prevent banks from arbitraging the difference between DF rate and
overnight money market rate. Therefore, this system does not completely eliminate rate volatility and
some central banks have been forced to introduce new instruments (for example, in September 2014
the Fed implemented the Overnight Reverse Repo Purchase, ON RRP) to provide a lower bound to
money market rates (see Appendix II for more details).
The floor system is considered useful for simplifying the implementation of monetary policy and has
recently been applied by many central banks, as a response to the excess liquidity injected into the
system to deal with the global financial crisis first and the spread of the pandemic in more recent
times. In fact, since it is not based on a reserve maintenance period to generate the flat region of the
demand curve, nor on reserve requirements (either required or voluntary) to induce banks to hold
liquidity, it eliminates the significant administrative burdens for both banks and central banks
associated with such mechanisms (see previous paragraphs).
Moreover, even though the system has many operational advantages in stabilizing money market
rates (there is no need to accurately estimate the demand for reserves or to use frequent fine-tuning
operations) it also has some criticalities. First, it requires a large central bank balance sheet, with
inevitable consequences for the financial risks and economic burdens associated with the
remuneration of large outstanding reserves. Second, in case of high levels of excess liquidity, it leads
to money market rates falling below the deposit facility rate. This may require additional instruments
to control the level of rates and the need to admit generally unauthorized dealers as monetary policy
counterparties (as, for example, in the United States with the ON RRP facility).
Table 1A in the Appendix summarizes the pros and cons of reserve/quota-based stabilization
mechanisms comparing them with those of operational framework without such mechanisms.
4. Empirical evidence
This paragraph compares, firstly, the effectiveness of operational frameworks involving reserve/quota
systems (minimum reserve requirements, voluntary reserves, quotas); subsequently, the results are
compared with those reported by the operational frameworks that do not envisage such mechanisms.
The analysis uses data from the nine central banks mentioned in previous paragraphs (in addition to
the ECB, the central banks of Australia, Canada, Japan, Norway, New Zealand, the United Kingdom,
the United States and Sweden). The time horizon covers a 20-year period, from January 2002 to June
2022.
For each central bank, several phases have been identified which coincide with significant changes
in the operational framework of monetary policy. In some cases, these changes are due to the external
context, such as the outbreak of the great financial crisis or the spread of the Covid-19 pandemic.
The analysis was performed by comparing the volatilities of money market rates estimated through
the GARCH model. This model takes the significant clustering in the volatility of money market rates
(for example, Hamilton, 1996 and Klee et al., 2019) into account. This allows to evaluate both the
persistence of volatility phenomena (i.e. how much the current estimate of volatility is influenced by
its past values) and the response to innovations (shocks).22
Table 2 shows a summary of the evolution of operational frameworks and rate stabilization
mechanisms applied by different central banks in the period 2002-2022, highlighting different levels
of reserves with different colors. Specifically, reserves are classified on the basis of an econometric
analysis that takes into account of: i) the probability that a central bank implements a floor system in
correspondence with the different values of the reserve-to-GDP ratio; ii) the relationship between the
reserve level (relative to GDP) and the spread between the money market rate and the deposit rate
with the central bank. This analysis defines the lower bound for reserves under which money market
rate volatility could occur.23
GARCH (Generalized Auto-Regressive Conditional Heteroskedastic) is a statistical model used to model the variance
of time series in the presence of heteroschedasticity. In a GARCH model of order (p, q), variance is a function of the
values assumed by squared residuals (? process) in previous p periods and by the variance of previous q periods. More
details are presented in Appendix II.
In particular, reserves are classified as: 1) abundant (reserves/GDP ratio > 9 %), situation in which the demand curve
is flat and the money market rate is very close to the deposit facility rate; 2) ample (ratio between 7 and 9%), where
the demand curve is slightly downward sloping; 3) moderate (ratio between 4 and 7%) where the negative slope of
the curve increases and reserve changes start to induce non-negligible variations in money market rates; 4) scarce
(ratio under 4%), where reserves are very close to the aggregate requirement and the demand curve has a high negative
slope, leading to a strong rate reaction against small changes in reserves. For more details, see Appendix III. The
proposed classification is slightly different from the one that can be found in literature (e.g. Afonso et al., 2022), in
which only three categories are considered for reserves: abundant, ample and scarce.
Note: On the thresholds used for reserves’ classification see note 23 and Appendix III for details.
Table 2 – Evolution of rate stabilization mechanisms in Central Banks
The performance of the various operational frameworks was evaluated by creating 3 different groups:
1) operational frameworks with minimum reserve; 2) operational frameworks with voluntary
reserves; 3) operational frameworks with quotas. The results must be interpreted with due caution
associated with the fact that the analysis includes countries that have money markets with different
characteristics and participants, so that greater volatility could depend on structural factors.24 The
results were compared with those of central banks with frameworks without such mechanisms (central
banks with “narrow corridor and daily operations”, central banks with “floor”). Each group includes
data from central banks that have adopted the system to which the group refers, for the period in
which this occurred. Therefore, for different periods, some central banks are included in different
groups. The analysis excludes periods when central banks had in place reserve requirements within
de facto floor systems or, in any case, with at least ample reserve levels. Some systems have been
adopted only by one country and for a limited period (such as voluntary reserves in the United
Kingdom and reserve requirement with carry-over provision in the United States), so that any
generalization of results should be treated with caution.
The following tables (3.1 and 3.2) show the results of the analysis of the daily spread between the
money market rate and the monetary policy reference rate and its volatility.
In summary, with reference to the systems based on reserves/quotas, the analysis highlights the
following results:
the examined central banks were effective in pursuing the objective of keeping the overnight
money market rate close to the monetary policy reference rate, albeit with differences. In a
context of low reserves, the average absolute spread over the considered periods is between 2
basis points (bps) (Japan and Norway) and 8 bps (euro area and New Zealand), while in a
context of moderate reserves, the euro area has the highest value of 47 bps. With regard to the
volatility, it ranges from 0.02 (Norway) to 0.13 (euro area) under regimes with scarce reserves
while in case of moderate reserves the highest value is again for the euro area (0.22).25
the two central banks having a minimum reserve system with mobilization (ECB and Bank of
Japan) present very different results. While for Japan, both the average spread and the
volatility were very low regardless of the amount of outstanding reserves, for the euro area
these values are much higher, particularly in the period of moderate reserves (where the spread
in absolute value and volatility were respectively 47 bps and 0.22);
for the reserve requirements system with the carry-over provision implemented in the United
States, the volatility of the differential is in line with that of the euro area in the period of
scarce reserves (0.07);
in terms of spread volatility, moderate values are reported by the BoE and Norges Bank, which
have adopted systems based on voluntary reserves and quota-based systems, respectively.
For example, the structure of the banking system, with prevalence of small banks that are not very active in the money
market, the weight of non-bank intermediaries on the money market, autonomous factors, the share of over-the-counter
transactions.
To deal with the 2008 financial crisis, the ECB introduced fixed rate full allotment auctions and longer-term loan
programs (LTRO), allowing the overnight money market rate (EONIA) to move towards the lower limit of the
corridor. To evaluate the effects of these changes on the metrics of Table 3.1 see the next sub-paragraph 4.1
‘Robustness Check’.
Table 3.1 – Comparison of the performance of different reserve mechanisms26
Macro
SYSTEMS WITH RESERVE/QUOTA
group
Group
Central
Minimum
reserve
systems Averaging
Eurozone
2002-2008
Eurozone
2008-2012
Eurozone
2013-2015
Japan
2002-2006
Japan
2006-2008
SYSTEMS WITHOUT RESERVE/QUOTA
Median
GARCH
Volatility
0.053
0.060
0.079
0.083
-0.453
-0.553
0.219
0.469
14.8%
-0.155
-0.108
0.128
0.170
52.5%
0.001
0.001
0.002
0.001
calculable27
0.008
0.005
0.027
0.016
Absolute Volatility/
Value
Average
policy rate
Minimum
Reserve
System Averaging +
Carry-over
United States28
2002-2008
0.002
0.000
0.073
0.046
Voluntary
reserve
systems
United
Kingdom
2006-2008
0.053
0.045
0.055
0.059
-0.004
-0.010
0.024
0.016
-0.119
-0.145
0.131
0.156
-0.068
-0.050
0.089
0.085
0.000
0.000
0.001
0.000
-0.004
-0.003
0.008
0.007
0.022
0.000
0.037
0.022
0.116
0.100
0.048
0.119
-0.049
-0.050
0.027
0.049
-0.074
-0.070
0.020
0.074
11.7%
-0.037
-0.030
0.045
0.044
-0.041
-0.042
0.026
0.046
-0.041
-0.050
0.026
0.043
12.1%
-0.065
-0.058
0.084
0.071
213.8%
Quota system
(TARALAC)
Reserve
Level
Average
Width
between
Standing
Facilities rates
Systems with
narrow
corridor and
daily
operations
Floor systems
Norway
2011-present
New Zealand
2007-2009
New Zealand
2009-2020
Australia
2002-2020
Canada
2002-2020
New Zealand
2002-2007
Sweden29
2002-2008
Canada
2020-present
Australia
2020-present
New Zealand
2020-present
United
Kingdom
2009-present
United States
2020- present
Sweden
2020- present
The outlier values (0.5%) of the entire sample were removed.
It is not possible to calculate the ratio between volatility and the average policy rate as the policy rate was 0 per cent
for the whole period.
For United States, the column of the Average width between Standing Facilities rates contains the average difference
between the Primary Credit Rate and the Federal Funds Rate, as the deposits with the central banks were not
remunerated (for details, see Appendix II).
The corridor was wide (± 75 basis points) until April 2009, was then reduced until July 2010 (± 50 basis points) and
finally widened again (± 75 basis points) until 2019.
Table 3.2 – Comparison of the aggregate performance of different reserve mechanisms.
Macro-group
SYSTEMS WITH
RESERVE /
QUOTA
Group
Reserve Level
GARCH
Volatility
Absolute value
Minimum reserve systems
– Averaging
0.012
0.075
0.081
-0.208
0.102
0.217
0.002
0.073
0.046
0.053
0.055
0.059
-0.035
0.055
0.049
-0.119
0.131
0.156
Minimum reserve systems
– Avg. + Carry-over
Voluntary reserve systems
Quota systems
(TARALAC)
Systems with narrow
0.017
0.014
0.022
corridor and daily
operations
-0.047
0.032
0.051
Floor Systems
Note: Tables 3.1 and 3.2 exclude the periods in which central banks had minimum reserve requirements under, de facto, floor
systems, or, in any case, with at least ample reserve levels. In the Reserve Level column, S indicates Scarce reserves (reserves
to GDP ratio value 9%). The Average Width between Standing Facilities rates column contains the
average difference between the two standing facilities (marginal lending and deposit) rates in the considered period. The Mean
and Median columns show the mean and median value, calculated on daily data, of the spread between the money market rate
and the policy rate. The GARCH Volatility column shows the average volatility of the spread, measured through a GARCH, in
the considered period. The Absolute value mean column indicates the mean of the spread calculated on the absolute values.
Finally, the Volatility / Average Policy Rate column compares the average volatility of GARCH to the average level of the
policy rate, both calculated over the considered period. The daily money market rates considered are: EONIA (Euro OverNight
Index Average) until October 2019 and the €STR (Euro Short-Term Rate) thereafter for the Eurosystem, Cash Rate for Australia,
OMMFR (Overnight Money Market Financing Rate) for Canada, Overnight Call Rate for Japan, NIBOR (Norwegian InterBank Offer Rate) until October 2011 and NOWA (Norwegian Overnight Weighted Average) thereafter for Norway, OICR
(Overnight Interbank Cash Rate) for New Zealand, SONIA (Sterling Over Night Index Average) for the United Kingdom, EFFR
(Effective Federal Funds Rate) for United States, STIBOR (Stockholm Interbank Offered Rate) until September 2021 and
SWESTR (Swedish krona Short Term Rate) thereafter for Sweden. For more details, see Appendix II.
SYSTEMS
WITHOUT
RESERVE/QUOTA
If we compare the results of the systems with reserves/quotas with those without them (systems with
daily operations and floor systems), the second macro-group reports lower values both for the mean
and for the volatility of the spread. In particular, systems with narrow corridor and daily operations,
in contexts of structural liquidity deficits, record an overall difference in absolute value equal to 0.02,
while for floor systems, in contexts of excess liquidity, this value is equal to 0.05. In both cases, the
volatility of the spread is very low (equal to 0.01 and 0.03, respectively).30 This latter result shows
that under these systems the spread between money market rates and the policy rate has been stable
and predictable, indicating an effective steering of short term rates as volatility has been kept low.
In summary, systems based on narrow corridor and daily operations and floor systems, which do not
require additional stabilization mechanisms, are more effective. However, they may present
disadvantages, such as: in the first case, excessive operational burdens; in the second case,
comparatively higher costs and higher financial risk.
For Canada and Australia, the volatility of money market rates is slightly lower under the narrow corridor system than
the one observed under the floor system. The former system is very effective in stabilizing the money market rate, but
implies a significant operational burden for the central bank requiring the use of daily operations.
4.1. Robustness check
In the previous paragraph, the classification of reserves is based on an econometric model that
assumes constant thresholds across jurisdictions and time. As a robustness check, we estimated the
model without these assumptions.31
Firstly, we estimated the model adding a dummy time variable to capture time effects; secondly, we
tested the model allowing the coefficients to vary across jurisdictions. Both estimates confirm, on
average, the previous results (see Tables 4A and 5A in Appendix III).
In addition, we used econometric analysis where the probability of a central bank implementing a
floor system is related to the value of the spread between the money market rate and the rate applied
to bank deposits at the central bank, instead of the value of the reserves-to-GDP ratio as previously.
The estimates show a negative relation between the spread and the probability that the central bank
implements a floor system: the closer the money market rate to deposit rate, the greater the probability
that central banks adopt a floor system (see Tables 6A, 7A and 8A in Appendix III).
In light of these latest results, the Eurozone could be considered a de facto floor system already in the
period 2009-2011, since it registered a low spread between EONIA and Deposit Facility Rate
(DFR).32 Considering the DFR as de facto policy rate instead of the main refinancing operation
(MRO) rate modifies the metrics in Table 3.1. As expected, the mean, the median value of the spread
between the money market rate and the reference rate as well as the mean of the spread calculated on
the absolute values decreased, compared to the previous analysis, to 0.160, 0.100 and 0.160,
respectively. It is also important to note that the average volatility of the spread shows a strong
reduction (to 0.114, measured through a GARCH), but it increases relative to the average reference
rate (to 45.4%). Even with the new metrics, however, the evaluations made in the previous paragraph
remain unchanged.
5. Comparison of various stabilization systems
Each of the systems analyzed in the previous paragraphs has advantages and drawbacks that vary
across different liquidity conditions. Historically, the stabilization systems analyzed in this paper
were part of central banks’ operational framework to ensure monetary control and to provide banks
with a liquidity buffer. In particular, such systems generate a predictable minimum demand for central
bank credit by the banks. These funding needs provide the central bank with a tool to adjust its stance
by changing the rate applied to liquidity provision operations.
In a corridor system with scarce or no liquidity surplus, minimum reserve requirements, when
accompanied with other provisions (such as averaging, carry-over or the compliance band), would
maintain their role in helping to stabilize short-term rates through intertemporal smoothing of
autonomous factor shocks.
However, in today’s circumstances, it must be considered that there is less need to establish a
structural demand for central bank reserves through minimum reserve requirements. This is because
See Appendix III for details.
See Appendix II.
there is ample liquidity demand driven by autonomous factors (such as banknotes), and most banks
may wish to maintain additional reserve buffers for payment purposes.
Furthermore, the rationale for implementing minimum reserve requirements as a tool both to offer
banks a liquidity buffer and to curb excessive maturity transformation activities has become less
relevant. The regulatory reforms implemented after the global financial crisis have delegated the
former function to the liquidity coverage ratio (LCR) and the latter to the net stable funding ratio
(NSFR).
In a floor system, minimum reserve requirements would not be needed to stabilize money market
rates, as the central bank can control the stance by adjusting the policy rate at which reserves are
remunerated. However, even under such a system, the minimum reserve mechanism could be justified
by several motives.
The central bank can consider adjusting the minimum reserve requirements ratio as an option to
rapidly modify the structural amount of liquidity in the system. If, at a future date, the central bank
decides to reduce the structural liquidity position, it can increase the minimum reserve requirements
ratio. Nevertheless, any such decision should be carefully weighed against possible drawbacks.33
Furthermore, in an environment of ample excess liquidity, minimum reserve requirements could be
used as a benchmark in the case of the adoption of a tiered reserve remuneration structure.34 When
tiering is aimed at mitigating the effects of negative rates on banks’ net interest margins in the
presence of nominal rigidity of deposit rates at the zero lower bound, the use of minimum reserve
requirements (or a proxy based on bank deposits) ensures better alignment between the purpose
pursued and the allocation of allowances (Boucinha et al., 2022). Similarly, minimum reserve
requirements could serve as a benchmark in a reverse tiering system aimed at reducing the interest
paid to banks on reserves in situations of positive interest rates and large excess liquidity. However,
such a mechanism may have adverse effects, potentially leading to unintended distributional
consequences within the banking sector.
In a “parsimonious” floor system, minimum reserve requirements with averaging may help to
calibrate monetary policy operations (Della Valle et al., 2022).35 Compared with the corridor system,
in a “parsimonious” floor system, the central bank aims to satisfy slightly more than the precautionary
demand for reserves to maintain the market rate at the floor, which is remunerated at the policy rate.
The reserve supply, based on estimates of the demand for reserves and forecasts of autonomous
factors, could be calibrated without a minimum reserve requirement. However, if a minimum reserve
requirement is in place and is high enough – constituting a known, predetermined portion of the total
reserves of the system – the central bank would encounter fewer uncertainties in determining the
reserve supply needed to maintain the market rate at the floor.
See note 10.
The tiered reserve remuneration structure implies a different rate of remuneration for part of a credit institution’s
current account holdings in excess of minimum reserve requirements. In particular, the Eurosystem adopted this
system to exempt part of these holdings from the deposit facility rate (DFR) when the rate was negative (see note 6
for more details).
Della Valle et al. (2022) define “parsimonious” floor system as the minimum central bank balance sheet size to
implement a floor. This is different from a floor system achieved only through asset purchase programs.
Regardless of the prevailing liquidity conditions, Table 4 shows the performance of various
mechanisms with reference to different aspects.
Table 4 – Pros and cons of reserve/quota systems
Liquidity
forecasts
accuracy *
Frequency of
fine tuning
operations
Operational
burden for banks
Independence
between interest
rate target and
liquidity
provision**
Adaptability to
the business
model
Minimum reserve
with averaging
Minimum reserve
with tolerance
Minimum reserve
with carry-over
Voluntary reserve
with tolerance
Quota systems
(TARALAC)
Medium
Medium
Medium
Medium
Medium
Medium
Medium
* Accuracy required in central bank liquidity estimates.
** Under the condition of independence, the equilibrium interest rate does not depend on the exact quantity of reserve
balances supplied by the central bank. The interest rate target can be set according to the usual monetary policy concerns,
while the quantity of reserves can be set independently.
Note: Negative characteristics are highlighted in red, neutral in yellow, and positive in green.
The minimum reserve with averaging on one hand implies: i) a high complexity to obtain an accurate
liquidity forecast by the central bank, in order to correctly estimate the liquidity needed by the banking
system. This drawback would be less relevant in the context of a floor system with ample excess
liquidity; ii) a high operational burden for banks, because the fulfillment on average compels banks
to keep track of their fulfillment path; iii) a low adaptability to the business model of the banks: the
requirement is fixed according to the same formula for all banks, which takes nearly no individual
peculiarity into account. On the other hand, it supports money market activity, although this would
also depend on the regulatory costs of accessing the interbank market.
Compared to the mechanism with averaging, minimum reserves with tolerance band and minimum
reserve with carry-over have a lower operational burden for banks and require less frequent finetuning operations, because the banks’ requirements are more flexible and therefore they need a lower
accuracy of liquidity forecasts by the central bank. As the mechanism with averaging, these two
systems encourage money market activity.
Voluntary reserve with tolerance band is more flexible than the previous systems, since it allows
banks to set their own requirement, while the tolerance bands lower the sensitivity of money market
interest rate to liquidity shocks. However, voluntary reserves could be very difficult to manage in a
situation of large excess liquidity; under such circumstances, banks could find it burdensome and
increasingly difficult to calibrate their reserve targets.
Quota systems relax the need for very accurate liquidity forecasts, although quotas would need to be
defined, reviewed and closely monitored by the central bank. The central bank should set the quotas
in such a way to support money market activity while the volatility of short-term rates remains low.
However, under this system the central bank must be ready to intervene on a daily basis in the event
of an imbalance due to unexpected autonomous factors, even if the calibration of the quotas works
well.
Both the system with voluntary reserves and with quotas have the advantage of being able to adapt
requirements to the business model of counterparties; however, both can become burdensome for the
central bank when applied to a large monetary area characterized by wide fluctuations in the
autonomous factors.
6. Conclusions
This paper analyzes the Eurosystem’s minimum reserve system with averaging, assessing its main
advantages and disadvantages and comparing it with alternative reserve systems adopted by other
central banks (such as minimum reserves with carry-over provision, voluntary reserves with
compliance bands and the quota system – TARALAC).
Minimum reserve requirements generate a predictable minimum demand for central bank credit by
the banks, regardless of the operational framework in which they are established. This mechanism,
when accompanied with an averaging provision, is effective in a corridor system, helping to stabilize
short-term rates through intertemporal smoothing of autonomous factor shocks. In a floor system,
minimum reserve requirements would not be needed for this purpose, as the central bank can control
the stance by adjusting the policy rate at which reserves are remunerated. However, even under a
floor system, the minimum reserve mechanisms could be used to rapidly adjust the structural amount
of the liquidity in the system as well as a benchmark when a tiered reserve remuneration scheme is
adopted.
References
Åberg, P., Corsi, M., Grossmann-Wirth, V., Hudepohl, T., Mudde, Y., Rosolin, T. and Schobert, F.
(2021). Demand for central bank reserves and monetary policy implementation frameworks: the case
of the Eurosystem. ECB Occasional Papers No. 282 / September 2021.
Afonso, G., Giannone, D., La Spada, G. and Williams, John C. (2022). Scarce, Abundant, or Ample?
A Time-Varying Model of the Reserve Demand Curve. FRB of New York Staff Report No. 1019,
2022.
Afonso, G., Cipriani, M., Copeland, A., Kovner, A., La Spada, G. and Martin, A. (2020a). The market
events of mid-September 2019. Federal Reserve Bank of New York Staff Reports No.918.
Afonso, G., Kim, K., Martin, A., Nosal, E., Potter, S. and Schulhofer-Wohl S. (2020b). Monetary
policy implementation with an ample supply of reserves. Federal Reserve Bank of Chicago Working
Paper, (WP 2020-02).
Armenter, R. (2016). A tractable model of the demand for reserves under nonlinear remuneration
schemes. Federal Reserve Bank of Philadelphia Working Paper, (16-35).
Bank of Canada. Framework for market operations and liquidity provision.
https://www.bankofcanada.ca/markets/market-operations-liquidity-provision/framework-marketoperations-liquidity-provision/.
Bank of England (2006). The framework for the Bank of England’s operations in the sterling money
markets (Red book).
Bank of England (2018a). Monetary policy summary and minutes of the Monetary Policy Committee
meeting ending on 20 June 2018.
Bank of England (2018b). The Bank of England’s future balance sheet and framework for controlling
interest rates, August 2008. https://www.bankofengland.co.uk/-/media/boe/files/paper/2018/boefuture-balance-sheet-and-framework-for-controlling-interest-rates.pdf.
Bank of England (2022). Monetary policy summary and minutes of the Monetary Policy Committee
meeting ending on 4 August 2022.
Bank of Japan (2001). New Procedures for Money Market Operations and Monetary Easing, 19
March 2001. https://www.boj.or.jp/en/mopo/mpmdeci/mpr_2001/k010319a.htm.
Baughman, G. and Carapella, F. (2019a). Voluntary reserve targets. Journal of Money, Credit, and
Banking.
Baughman, G. and Carapella, F. (2019b). A simple model of voluntary reserve targets with tolerance
bands. Finance and Economics Discussion Series 2019-060. Washington: Board of Governors of the
Federal Reserve System, https://doi.org/10.17016/FEDS.2019.060.
Berentsen, A. and Monnet, C. (2008). Monetary policy in a channel system. Journal of Monetary
Economics, 55 (6), 1067-1080.
Bindseil, U. (2004). The operational target of monetary policy and the rise and fall of Reserve position
doctrine. ECB WP No. 372.
Bindseil, U. (2016). Evaluating monetary policy operational frameworks. Discussion paper at the
Federal Reserve Bank of Kansas City, Economic Policy Symposium in Jackson Hole, Aug. 25-27,
2016.
Bindseil, U. (2018). What Monetary Policy Operational Framework after the Crisis? Revue française
d’économie, 2018/3 Vol. XXXIII.
Bindseil, U. and Würtz, F. (2008). Efficient and universal frameworks (EUF) for monetary policy
implementation. Working Paper, August (presented at the Ausschuss für Geldtheorie und Geldpolitik
des Vereins für Socialpolitik, Frankfurt am Main, Feb. 20 2009).
Boucinha, M., Burlon, L., Corsi, M., Della Valle, G., Eisenschmidt, J., Marmara, I., Pool, S.,
Schumacher, J. and Vergote, O. (2022). Two-tier system for remunerating excess reserve holdings.
ECB Occasional papers, Number 302.
Bowman, D., Gagnon, E. and Leahy, O. (2010). Interest on excess reserves as a monetary policy
instrument: the experience of foreign Central Banks. Board of Governors of the Federal Reserve
System, International Finance Discussion Papers, Number 996.
Clews, R. (2005). Implementing monetary policy: reforms to the Bank of England’s operations in the
money market. Bank of England Quarterly Bulletin: Summer 2005.
De Grauwe, P. and Ji, Y. (2023a). Monetary policies that do not subsidise banks. VoxEU.org, 9
January.
De Grauwe, P. and Ji, Y. (2023b). Monetary policies with fewer subsidies for banks: A two-tier
system of minimum reserve requirements. VoxEU.org, 23 March.
De Grauwe, P. and Ji, Y. (2023c). Fighting inflation more effectively without transferring central
banks’ profits to banks. CESifo Discussion Paper.
Della Valle, G., King, D. and Veyrune, R. (2022). Monetary Operations and Domestic Market
Development: Reserve Requirements. IMF Technical Assistance Handbook.
Deuber, G. and Zobl, F. (2023). ECB Minimum Reserves – 10% or 10% less Government Bonds.
SUERF Policy Brief 741, November.
Ennis, H. M. and Keister, T. (2008). Understanding monetary policy implementation. Economic
Quarterly, 2008, 94 (3). Federal Reserve Bank of Richmond.
European Central Bank (2022). Monetary policy decisions, 15 December
https://www.ecb.europa.eu/press/pr/date/2022/html/ecb.mp221215~f3461d7b6e.en.html.
2022,
European Central Bank (2023). Monetary policy decisions, 27 July
https://www.ecb.europa.eu/press/pr/date/2023/html/ecb.mp230727~da80cfcf24.en.html.
2023,
Federal Reserve System (2008). Monetary Policy Decision,
2008.
October
Filardo, A. (2020). Monetary policy operating frameworks: are reforms heading in the right direction?
Cato Journal, Vol. 40, No. 2 (Spring/Summer 2020).
Gray, S. (2011). Central Bank Balances and Reserve Requirements. IMF working paper, 11/36.
Hamilton, J. D. and Lin, G. (1996). Stock market volatility and the business cycle. Journal of applied
econometrics, vol. 11, 573-593.
Hauser, A. (2019). Waiting for the exit: QT and the Bank of England’s long-term balance sheet.
Speech at the European Bank for Reconstruction and Development, London, and hosted by the
Association for Financial Markets in Europe (AFME), International Swaps and Derivatives
Association (ISDA) & International Capital Market Association (ICMA), July 17.
Hauser, A. (2021). Bigger, broader, faster, stronger? How much should tomorrow’s central bank
balance sheets do – and what should we leave to financial markets? Some principles for good
parenting. Speech at the International Finance and Banking Society (IFABS) 2021 Oxford
Conference on “The Financial System(s) of Tomorrow”, September 13.
Ihrig, J., Senyuz, Z. and Weinbach, G. C. (2020). The Fed’s “ample-reserves” approach to
implementing monetary policy. Finance and Economics Discussion Series 2020-022. Washington:
Board of Governors of the Federal Reserve System, https://doi.org/10.17016/FEDS.2020.022.
Keister, T., Martin, A. and McAndrews, J. (2008). Divorcing money from monetary policy. Federal
Reserve Bank of New York Economic Policy Review, (Sep), 41-56.
Klee, E., Senyuz, Z. and Yoldas, E. (2019). Effects of changing monetary and regulatory policy on
money markets. International Journal of Central Banking, October 2019.
Kwapil, C. (2023). A two-tier system of reserve requirements by De Grauwe and Ji(2023): A closer
look. SUERF Policy Brief No 702, October.
Jackson, C. and Sim, M. (2013). Recent developments in the sterling overnight money market. Bank
of England Quarterly Bulletin: Q3 2013.
Lagarde, C. (2023). Speech at the Hearing of the Committee on Economic and Monetary Affairs of
European
Parliament,
September
2023,
https://www.ecb.europa.eu/press/key/date/2023/html/ecb.sp230925~036083efca.en.html.
Lee, J. (2016). Corridor system and interest rates: volatility and asymmetry. Journal of Money, Credit
and Banking, 48 (8), 1815-1838.
Logan, L. K. (2019). Observations on Implementing Monetary Policy in an Ample-Reserves Regime,
17 April 2019. https://www.newyorkfed.org/newsevents/speeches/2019/log190417.
Logan, L. K. (2022). Speech on Monetary Policy Operations, the Fed’s New Standing Repo Facility,
and the Future of the Fed’s Balance Sheet, 10 January 2022. https://www.mercatus.org/macromusings/lorie-logan-monetary-policy-operations-feds-new-standing-repo-facility-and-future.
Logan, L. K. (2023). Ample reserves and the Friedman rule, 10 November 2023.
https://www.dallasfed.org/news/speeches/logan/2023/lkl231110#n9.
McCauley, R. and Pinter, J. (2024). Unremunerated reserves in the Eurosystem, part 2: Tax incidence
and deposit relocation risks. VoxEU.org, 16 January.
Naber, J. M., Sambasivam, R. and Styczynski, M-F. (2017). Demand for Voluntary Balance
Requirements: the U.S. Experience with Contractual Clearing Balances from 2000 to 2007, FEDS
Notes. Washington: Board of Governors of the Federal Reserve System, January 4, 2017.
Nakaso, H. (2017). Evolving Monetary Policy: The Bank of Japan’s Experience, Speech at the Central
Banking Seminar hosted by the Federal Reserve Bank of New York, 18 October 2017.
https://www.boj.or.jp/en/about/press/koen_2017/data/ko171019a1.pdf
Norges Bank. Background for the system for managing bank reserves in Norway.
https://www.norges-bank.no/en/topics/liquidity-and-markets/The-liquidity-managementsystem/The-management-of-bank-reserves-The-system-in-Norway/Background-system-managingbank-reserves/.
Obrien, Y. C. (2007). Reserve Requirement Systems in OECD Countries, Federal Reserve Board,
Washington, 2007-54.
Reserve Bank of Australia. How the Reserve Bank Implements Monetary Policy.
https://www.rba.gov.au/education/resources/explainers/how-rba-implements-monetary-policy.html.
Reserve
Australia.
Monetary
Policy
Implementation
Australia.
https://www.rba.gov.au/education/resources/in-a-nutshell/pdf/monetary-policy-implementation.pdf.
Reserve
Australia.
Unconventional
Monetary
Policy.
https://www.rba.gov.au/education/resources/explainers/unconventional-monetary-policy.html.
Reserve Bank of New Zealand (2022). New Zealand’s Monetary Policy Implementation Framework,
07 September 2022. https://www.rbnz.govt.nz/hub/publications/speech/2022/speech2022-09-07.
Schnabel, I. (2023). Speech at the Money Market Contact Group meeting, 2 March 2023.
https://www.ecb.europa.eu/press/key/date/2023/html/ecb.sp230302~41273ad467.en.html.
Sveriges Riksbank (2020). Monetary Policy Report, 28 April 2020. https://www.riksbank.se/engb/monetary-policy/monetary-policy-report/2020/monetary-policy-report-april-20202/.
Sveriges Riksbank (2020). The Riksbank’s balance sheet is growing, 28 April 2020.
https://www.riksbank.se/globalassets/media/rapporter/ppr/fordjupningar/engelska/2020/theriksbanks-balance-sheet-is-growing-article-in-monetary-policy-report-april-2020.pdf.
Sveriges
Riksbank
(2021).
Monetary
Policy
Report,
September
2021.
https://www.riksbank.se/globalassets/media/rapporter/ppr/engelska/2021/210921/monetary-policyreport-september-2021.pdf.
APPENDIX
Table 1A – Pros and cons of reserve/quota-based stabilization mechanisms and operational
framework without such mechanisms
counterparties
For central bank
For counterparties
For central bank
Systems with reserve/quota
? Easy to implement
? In case of liquidity deficit,
effective in stabilizing rates
during maintenance period
? If combined with wide
corridor, it supports interbank
trading
Minimum
reserve with
averaging
Minimum
reserve with
tolerance bands
Minimum
reserve with
carry-over
Voluntary
reserve with
tolerance bands
Quota system
(Taralac)
? Flexible and
operationally
undemanding
? In case of liquidity deficit, it
eliminates volatility at the
end of the maintenance
period
? Flexible and
operationally
undemanding
? In case of liquidity deficit, it
eliminates volatility at the
end of the maintenance
period
? Flexible
? It reflects the real
liquidity needs of
banks
? It can help to
reduce the
liquidity premium
? It can improve forecasts
ability for central bank
? Flexibility for the
counterparties choice
? It allows independence
between interest rate target
and the provision of liquidity
? It supports interbank trading
? Operationally
undemanding
? Inflexible and burdensome
from the operational point of
view due to the need to keep
track of the fulfillment path
? Based on an aggregate that
does not reflect the actual
liquidity needs of each bank.
? It can be an inefficient tool in
times of crisis
? Based on aggregate that does
not reflect the actual liquidity
needs of each bank
? Based on aggregate that does
not reflect the actual liquidity
needs of each bank
? Possible difficulties in
calculating reserve amounts
for small banks
? In situations of liquidity
deficit,
requires
additional
fine-tuning
interventions at the end of
the maintenance period
? In case of excess liquidity,
it is not strictly needed
? If the band is too wide, it
reduces interbank trading
? In case of excess liquidity,
it is not strictly needed
? It complicates central bank
monitoring of reserve
requirement compliance
? It leads to speculation about
the next maintenance
period’s key rate decisions
? In case of excess liquidity,
it is not strictly needed
? Complicated to manage in
the presence of QE and
negative rates
? Complexity
quota
setting due to the trade-off
between operational target
control
ability
interbank market activity
? Possibly unsuitable in case
of high volatility of
autonomous factors
? It does not need very accurate
liquidity forecasts
Systems without reserve/quota
Daily monetary
policy operations
and narrow
corridor
Floor
? Flexible
? Accurate control of money
market rates at the end of the
? Flexible and
operationally
undemanding
? It allows balance sheet to be
determined independently of
interest rate targets
? It does not need accurate
liquidity forecasts
? It does not need fine tuning
operations
? Operationally demanding
? Operationally demanding
? Possible high volatility of
intraday rates pending
central bank intervention
? It needs a large balance
sheet
? It may request admission to
the central bank deposit of
operators other than banks
Appendix I – Voluntary reserve and quota systems
This appendix presents some theoretical elements about voluntary reserve and quota systems.
Voluntary reserve
In a voluntary reserve system, each bank commits in advance to a target for reserves at the central
bank, while the central bank still maintains the control of the aggregate reserves by setting a range
for individual targets, within which each bank must place its own target.
Baughman and Carapella (2019b) show that if the reserve remuneration function is additively
separable, depending in part on the reserve target chosen by each bank (T) and in part on the difference
between the reserves holdings and the target (D–T), banks will be induced to set voluntary reserve
targets that bring the expected interbank market rate to equal the central bank-determined reserve
rate.
If we denote by ? the penalizing rate that banks would be required to pay if they held reserves lower
than the target, by iE the rate that is paid on reserves in excess of the target and by iT the rate on
reserves up to the target, we obtain that at the end of a maintenance period the remuneration of the
reserves held D with respect to a target T can be expressed with the following form:
?(?, ?) = {
? ? × ? + ?? × (? ? ?)
? ? × ? ? ? × (? ? ?)
?? ? ? ?
?? ? < ?
The precondition for the central bank to encourage banks to align their reserves with the targets is the
following relationship between the rates:
?? < ? ? < ?
The remuneration function can be rewritten as ?(?, ?) = ? ? × ? + ??, where:
(? + ?) × (? ? ?)
?? = { ?
?? × (? ? ?)
?? ? < ?
?? ? ? ?
This functional form is additively separable into a portion that depends on the target T and a portion
that depends on the difference D-T, which makes interbank market trading choices an affine function
of target reserves T. Baughman and Carapella (2019b) show that with this remuneration function, in
equilibrium, banks will set target levels such that expected money market rates equal the target rate:
E[iM ] = iT.
In the case of a system with tolerance bands, reserve balance values that differ from the target but fall
within the band would not imply consequences for the banks, which would in any case obtain the
voluntary reserves rate as remuneration. In this case, the remuneration function taking into account
the tolerance bands (of magnitude 2?T) becomes:
? ? × (1 + ?) × ? + ?? × (? ? (1 + ?) × ?)
(?, ?) = { ? ? × ?
? ? × ? ? ? × ((1 ? ?)? ? ?)
?? ? ? (1 + ?)?
?? (1 ? ?)? ? ? < (1 + ?)?
?? ? ?)+iT × ?(-? < ? ? (? + ?) < ?) + ?×P(M-(? + ?)?) = P(M-(? + ?)?) = P(-??)+iT × ?(0 < ? ? (? + ?) < ?) + ?×P(M-(? + ?)?) = P(M-(? + ?)?) = P( ? ? ) = P( ? ? < ? )? P(? )
In other words, for symmetric probability distributions, if the central bank adopts a TARALAC
facility with limited access to the deposit, the central bank’s open market operations must inject an
amount of reserves equal to the sum of the autonomous factors (at the beginning of the day) plus half
of the quota (Q/2).
Appendix II – Review of systems adopted by central banks
This appendix analyzes the evolution of the monetary policy framework of nine central banks, over
the period 2002-2022, focusing on their efficiency in terms of money market rates stabilization around
the policy rate.
For each central bank, we show a figure in which the upper panel illustrates trends in the monetary
policy rate, the overnight money market rate and the level of reserves deposited on accounts with the
central bank (in this case the reference axis is on the right).36 The middle panel shows the trend of the
spread between the money market rate and the monetary policy reference rate; in the event that for a
central bank the changes to the operational framework have led to a change in the reference rate, the
trend of the new spread is also shown in the graph. Finally, the bottom panel illustrates the volatility
of the spread between money market rate and monetary policy reference rate, measured by applying
a GARCH (Generalised Autoregressive Conditional Heteroskedasticity) model. When it comes to
capturing the volatility clustering of financial time series, GARCH models dominate.
In a GARCH model of order (p, q), the process ? is characterized by:
?? = ?? ???
where ?t is an independent and identically distributed random variable, with zero mean and unit
variance (white noise process), and where ht evolves according to an ARMA process:
?? = ?0 +
? ?? ????
?=1
+ ? ?? ????
?=1
ht is the conditional variance of the process ? and is dependent upon previous q own lags and p squared
error terms. In our analysis a GARCH (1,1) has been estimated: in most cases, this model is sufficient
to capture the clustering of volatility in the data. The coefficient ?1 represents the response to shocks,
i.e. how the conditional variance reacts to the lagged value of the squared error term (innovation); the
coefficient ?1 is the proportion of how much the previously estimated value of the conditional
variance transfers to the current estimate. The sum of the coefficients ?1 and ?1 represents the
persistence of the process, i.e. for how long the past values influence the variance.37
For each central bank we identified different phases which coincide with significant changes in the
operational framework of monetary policy. In some cases, these changes are due to the external
context, such as the outbreak of the great financial crisis or the spread of the Covid-19 pandemic.38
The amount of reserves deposited on accounts with the central bank, in relation to GDP, have been
defined as: 1) abundant (value of the ratio of reserves to GDP > 9 percent); 2) ample (value of the
Figure presents total reserves, regardless of the presence of a minimum reserves requirement.
As proof of robustness, the GARCH analysis was replicated by excluding both the 0.5% and 2.5% of the smallest
extreme values and the largest extreme values, respectively, focusing only on the remaining 95% and 90% of the set
of observations (trimming or truncation). As an alternative, GARCH was also estimated with winsored data, in which
observations with values greater or less than those of the considered percentiles are replaced with values exactly
corresponding to the percentiles. This allows us to align with several empirical works in which end-of-month, end-ofquarter or end-of-year data are eliminated from the dataset, since they are influenced by window dressing phenomena.
Since there was a substantial easing of monetary policy during the period under consideration, in order to analyze the
development of differentials in the different phases, in addition to comparing their mean and standard deviation, we
also took into account measures that combine them, such as the coefficient of variation, which is obtained as the ratio
between the standard deviation and the mean in absolute value.
ratio between 7 and 9 percent); 3) moderate (ratio value between 4 and 7 percent); 4) scarce (value
of the ratio of reserves to GDP < 4 percent). For more details about this classification, see Appendix
Eurozone
With reference to the operational framework of the Eurozone, a gradual transition from a corridor
system to a de facto floor system has taken place in the years under review. In particular, it is possible
to distinguish three phases. The reserve requirement mechanism with averaging is present in all three
phases.
Phase I (2002-2008): system with wide corridor
In the first period, which runs up to October 2008, the operational framework of the Eurosystem was
based on a system with a wide corridor (standing facilities at ±100 bps relative to main refinancing
operations rate). As already mentioned, the minimum reserves system with averaging contributed to
the stabilization of rates.
In this period, the differential with the policy rate (main refinancing operations rate) averaged 5 bps,
but with a standard deviation of 0.11 percent. Considering the absolute value of the difference, the
spread increases, on average, to 8 bps, due to the volatility of the rate on the last day of the
maintenance period which forced the ECB to intervene with fine-tuning operations. Overall, the
volatility of the spread, as measured by GARCH, was high.
Phase II (2008 – 2015): system with a narrower corridor and moderate excess liquidity
In the second phase, which followed the great financial crisis and included the sovereign debt crisis,
fixed rate full allotment and longer-term loan programs were introduced (first LTRO and then
TLTRO). This contributed to a situation of growing, albeit moderate, excess liquidity. The width of
the corridor was reduced (standing facilities at ±50 basis points with respect to the rate of the main
refinancing operations). The spread to the policy rate averaged -44 bps; the EONIA rate gradually
moved towards the deposit facility rate: the spread to this rate averaged 20 bps, with a standard
deviation of 0.17 percent. Volatility showed some periods characterized by sudden increases.
Phase III (2015 – 2022): ample and abundant liquidity and negative rates
The third phase coincides with the launch of the quantitative easing programs at the beginning of
2015. From an operational point of view, this period is characterized by the significant growth in
central bank reserves, the adoption of an asymmetric corridor against a zero policy rate (marginal
lending at +25 bps over the main refinancing operations rate, with the deposit facility first at -40 bps
and then at -50 bps) and negative money market rates.
The system became a de facto floor. The spread between the money market rate39 and the deposit
facility rate averaged 4.4 bps, with a very low standard deviation (0.03 percent). With the adoption
of €STR starting from October 2019, the spread between the latter rate and the deposit facility rate
Since October 2019, EONIA has been replaced by euro short-term rate (€STR) as the reference money market rate in
the Eurozone; EONIA continued to be published until December 2021, but was calculated as €STR + 8.5 bps. In the
analyses, the money market rate used is given by the EONIA until data are available (2002 – 2021), and by €STR +
8.5 bps for the subsequent period (2021 – 2022).
averaged -6 bps with a standard deviation of 0.02 percent. The negative average of the spread is due
to the fact that in this period the money market is represented by transactions which also involve
operators that do not access the deposit facility. Volatility is lower than that of the previous phase.
Figure 1A – Rates and level of central bank reserves in Eurozone
Australia
The framework adopted by the Reserve Bank of Australia (RBA) for the implementation of monetary
policy is a corridor system focused on the key policy rate (target cash rate). The operational objective
of the RBA is to maintain the overnight-unsecured interbank market rate, the cash rate,40 at a level
close to the target cash rate. To pursue this objective, the Reserve Bank sets standing facility rates
and frequently conducts open market operations, both in Australian dollars and in foreign currency.
There are no reserve requirements for the banking system. Banks and other financial institutions hold
settlement accounts with the RBA (called Exchange Settlement Accounts, ESAs) to settle payments
and manage liquidity. The RBA does not have a target for the level of settlement balances, but instead
controls the supply of liquidity to keep the cash rate close to the target.
The RBA conducts daily open market operations in order to maintain ESA account balances at an
appropriate level to guarantee the smooth functioning of the money market. The daily open market
operations (mainly repurchase agreement and currency swaps)41 are performed primarily to offset the
The cash rate is calculated as a weighted average of the interest rates at which uncollateralized overnight funds are
traded in the interbank market.
Open market operations are carried out at least once a week, every Wednesday. The RBA may carry out additional
operations for other working days. Moreover, an additional round of open market operations can be decided every
afternoon. As regards the maturity of the operations, normally, it does not exceed 28 days. In addition to the open
market operations, there are also liquidity facilities, whose function is to provide liquidity to the banks following the
provision of collateral for the correct functioning of the payment system (RITS). Details are given below: 1) Intraday
SF Repo: interest-free intraday funding; 2) Overnight SF Repo: overnight funding, generally through the extension of
the intraday maturity to the following business day. The buyback price is set at 25 bps above the key policy rate; 3)
liquidity effects of the numerous transactions between the banking system and the Australian
government. It is possible to identify two phases, which differ in the symmetry of the width of the
corridor and/or in the amount of liquidity injected into the banking system.
Phase I (2002 – 2020): system with narrow, symmetrical corridor
In the first period, the corridor was narrow and symmetrical: standing facilities rates were set at ± 25
bps relative to the reference rate. In November 2013, following the introduction of new provisions to
ensure that financial institutions with settlement accounts with the central bank had sufficient reserves
to meet their interbank payment obligations and facilitate same-day settlement (including after normal
banking hours), balances held with the RBA increased (financed by using open repo transactions with
the central bank, which do not have a maturity date).
The RBA has been very effective in keeping the overnight market rate close to the official rate.
Deviations were rare, even during the period of financial market stress. The differential averaged 0
bps, with a very low standard deviation (amounting to 0.002 percent). The volatility of the differential
between the cash rate and the target was limited and stable, during both tightening and easing periods
of monetary policy. Several elements contributed to this stability, in addition to the presence of a
narrow, symmetrical corridor and the high frequency of transactions, such as: i) the small number of
active participants in the overnight market (15 to 20 institutions), which facilitates the search for
counterparties with positions to offset; ii) market participants openly disclosed their position (short
or long) on settlement balances and, by convention, generally traded overnight funds at the target
cash rate; iii) the possibility for the RBA to adjust intraday liquidity, if necessary, or use moral suasion
to encourage market participants to bring the overnight interbank rate back to target (Bowman et al.,
2010).
Phase II (2020 – 2022): system with asymmetrical corridor and abundant excess liquidity
After the outbreak of the pandemic, the RBA adopted an asymmetric corridor system, which set the
marginal lending rate and the deposit rate, respectively, 25 bps above and 15 bps below the target
cash rate. From November 2020, the deposit facility rate was set at 10 bps below the target cash rate;
it was, therefore, 0 percent when the target cash rate was reduced to 0.10 percent.
The central bank also: 1) launched a program to purchase government securities on the secondary
market (which had as a target the interest rate on 3-year securities);42 2) launched the Term Funding
Facility, i.e. 3-year loans with an interest rate equal to the target cash rate (with a disbursement
window from April 2020 to June 2021 and final maturity no later than June 2024); 3) set up a swap
facility with the Fed for liquidity up to 60 billion US dollars in exchange for Australian dollars. These
Open SF Repo: Repos from the Reserve Bank without a maturity date. It has the function of facilitating the liquidity
buffer that banks operating in the payment system must have and assumes particular relevance for "afterhours"
payments. The possibility of using this facility reduces the use of the intraday repo; 4) Term SF Repo: funding with a
maturity greater than one working day up to a maximum of 30 days. This is supplemented by the Committed Liquidity
Facility, targeted at certain monetary policy counterparties. This is an emergency facility linked to compliance with
regulatory requirements (LCR) for the banking system. It foresees a commitment fee of 15 bps. The potential amount
to be allocated is equal to the difference between the assessment of the LCR carried out by the macro-prudential
supervisory authority (APRA) and the assessment of the Reserve Bank regarding the quantity of HQLA securities that
the banking system can hold without jeopardizing the proper functioning of the market.
The initial envelope of the program was equal to 100 billion Australian dollars, subsequently increased by another
100 billion, totaling about 10 percent of GDP.
measures implied a sharp increase in central bank reserves held by the banking system. Because of
the central bank's liquidity injections, the spread between the unsecured overnight money market rate
and the key policy rate averaged -7 bps, with a standard deviation of 0.03 percent over the last two
years. However, the volatility of the spread, as measured by the GARCH, has been sharply increasing.
Figure 2A – Rates and level of central bank reserves in Australia
Canada
The operational framework of the Bank of Canada (BoC) was based on a narrow and symmetrical
corridor (except for a brief period following the global financial crisis) in the period from 2002 to
2020. Since the outbreak of the pandemic a floor system has been adopted. The Canadian operational
framework does not provide for minimum reserves.
Phase I (2002 – 2020): system with narrow and symmetrical corridor
The operational objective of the BoC is to peg the overnight money market interest rate to the
monetary policy reference rate (target for the overnight rate). 43 In the first phase, this objective was
pursued in the context of a corridor system, whose bounds were the deposit rate (target rate -25 bps)
and the bank rate (target rate +25 bps). When necessary, the BoC also conducted overnight open
market operations at the target rate. The BoC’s implementation of monetary policy is closely linked
to the level of liquidity in Canada’s payment system (Large Value Transfer System, LVTS). Since
the inception of LVTS in 1999,44 the BoC had set an overall target for account balances slightly above
The benchmark overnight market rate is a collateralized rate. Specifically, the BoC uses two indicators as proxies for
this rate. The first is the Overnight Money Market Financing Rate (OMMFR, shown in Figure 3A), which is the result
of an end-of-day survey of major participants in the overnight secured market. Some counterparties in this market,
such as pension funds, may not have access to the Bank of Canada’s deposit facilities. The second indicator is the
Canadian Overnight Repo Rate Average (or CORRA), which is a measure of the cost of overnight general collateral
repo transactions using Government of Canada (GoC) bonds as collateral. During 2020, the calculation methodology
was changed. The BoC calculates CORRA based on data from repo transactions traded by primary dealers and other
major players in the Canadian government debt market. The difference between the two rates averaged -0.2 bps.
Replaced by Lynx system in 2021.
zero (liquidity surplus) to reduce transaction costs and end-of-business day frictions for the execution
of payments, making the recourse to central bank overnight lending less frequent for counterparties
with deficit balances.45
Even though the BoC did not use the liquidity injected into the system (which translated into targets
for overall account balances in the payments system) to signal policy changes, the central bank
changed this aggregate to reinforce the operational target if (temporary) pressures pushed the
overnight rate away from the target.
The BoC increased the liquidity surplus in mid-August 2007 in response to increased financial
uncertainty and concomitant upward pressure on the overnight rate. Then it returned to a more typical
target (CAD 25 million) in March 2008. In 2009, when the BoC lowered the policy rate to 0.25
percent, bringing it to the lower bound of the corridor (de facto, adopting a floor system), it injected
CAD 3 billion surplus of liquidity to support the transition to the lower bound of the corridor. If,
therefore, the use of the deposit facility was usually very limited, as it was advantageous and
operationally feasible to find counterparties in the market with a liquidity shortage willing to pay the
target rate, with the transition to the floor system, the amount deposited daily to the deposit facility
was close to CAD 3 billion. The BoC in setting a clear target for excess reserves, rather than letting
the excess fluctuate above a certain threshold, believed it would help limit uncertainty among banks.
As tensions decreased, the BoC returned to using the narrow, symmetrical corridor system in 2010.
Figure 3A shows that the BoC managed to keep the overnight rate close to its target, even during the
financial crisis. Over the entire period, the average spread was -0.4 bps, with a standard deviation of
0.01 percent. Volatility, as measured by GARCH, was also low. This success probably reflects the
BoC’s ability to keep the supply of overnight balances stable and close to zero (with the exception of
the 2009 – 2010 period). The small number of participants in the interbank market and the presence
of a real-time payments system also helped in generating low uncertainty about payments at the end
of the day.
Phase II (2020 – 2022): floor system with abundant reserves
Following the outbreak of the pandemic, the BoC reduced the policy rate to 0.25 percent, bringing it
in line with the deposit rate, moving back to a floor system. It also implemented a series of expansive
measures as the GBPP (Government of Canada Bond Purchase Program) and medium to long-term
liquidity operations (Standing Term Liquidity Facility), in order to safeguard the smooth functioning
of markets.
The overnight interest rate fluctuated below the policy rate, with an average difference of -5 bps and
a standard deviation of 0.03 percent. The excess liquidity held at the end of the day by financial
institutions not directly participating in the payment system generated downward pressure for the
For example, an increase in the target on balances, implying an increase in surplus liquidity, would have determined
an incentive for banks to lend funds in the money market with consequent downward pressure on the overnight rate.
Indeed, as the level of surplus liquidity (current account balances target) increases, the incentive for banks to operate
in the money market is reinforced, since the liquidity surplus must otherwise be deposited overnight with the central
bank at a remuneration 25 bps lower than the policy rate and thus probably lower than the overnight money market
rate.
money market rate, pushing it below the deposit facility rate.46 It increased the volatility of the spread,
which was more reactive to shocks affecting the system.
Figure 3A – Rates and level of central bank reserves in Canada
Japan
The Bank of Japan's (BoJ) monetary policy operating framework over the past 20 years has been
characterized by an asymmetric corridor, with the level of reserves varying over time:
moderate/scarce between 2002 and 2008; scarce/moderate/ample between 2008 and 2013; abundant
after 2013. The BoJ also uses the minimum reserve requirement as a monetary policy tool. Following
the various monetary easing programs adopted by the BoJ, the stabilizing role of the reserve
requirement has substantially lost relevance. The reserve requirement ratio has remained unchanged
since 1991.
Phase I (2002 – 2008): Quantitative Easing, zero target rate and subsequent normalization
The operational target of the BoJ since 2001, with the introduction of the Quantitative Easing Policy,
was to control the overall level of reserves in the banking system, in order to counterbalance negative
inflation rates.47 The operational framework was based on an asymmetric and narrow corridor system,
in which the lower bound was the policy rate (monetary policy reference rate, corresponding to the
remuneration of current accounts held by financial institutions with the BoJ), while the upper bound
was the basic loan rate. The system was a de facto floor system. During the period of reserves increase
corridor was widened to 25 bps (from the initial 10 bps).
To correct this situation, the BoC would generally have to drain liquidity from the system, but given the emergency
due to Covid, it did not choose this option.
The Bank of Japan has undertaken to continue this policy "until the annual rate of change in the consumer price index
(CPI) is stable at or above zero per cent".
At this stage, the spread between the overnight money market rate48 and the policy rate averaged 0.4
bps, with a standard deviation of 0.02 percent49. Based on GARCH, the volatility of the differential
was low (particularly until early 2006), persistent, and with a relatively small response to shocks.
Phase II (2008 – 2013): asymmetric corridor and moderate liquidity
Starting from 2008, following the Global Financial Crisis, the BoJ reduced the policy rate50 and
narrowed the corridor with the basic loan rate (to +10 bps). At the same time, a new rate was
introduced to remunerate reserves in excess of the minimum reserve requirement (complementary
deposit facility rate), initially set at -20 bps relative to the policy rate (November 2008), later set equal
to the policy rate (from December 2008). In addition, the long-term interest rate was set as a new
operational target and a program to purchase long-term securities was initiated to pursue it (Asset
Purchase Program, APP), leading to an increase of liquidity.
During this period, the spread between the overnight money market rate and the policy rate was on
average negative and narrow (-1 bps), with a standard deviation of 0.01 percent.51 Based on GARCH,
the volatility of the differential was low, not particularly persistent and with a greater response to
shocks than in the previous period.
Phase III (2013 – 2016): Quantitative Qualitative Monetary Easing and abundant liquidity
Starting from 2013, in order to bring inflation closer to target levels, the BoJ redefined the monetary
policy framework by introducing the "price stability target," that is the achievement of the 2 percent
inflation target as quickly as possible.52 In addition, in April 2013 BoJ launched the Quantitative and
Qualitative Monetary Easing (QQE), a massive purchase program for long-term securities, which
generated a large excess of liquidity in the system.
In this period, there was a marked reduction in the coefficient of variation of both the money market
rate (0.10), with an average value of the rate equal to 7 bps and a standard deviation of 0.07 percent,
and the spread between the money market rate and the policy rate (0.25), with an average of -3 bps
and a standard deviation of 0.01 percent. Based on the GARCH, spread volatility was low but
persistent.
Phase IV (2016 – 2022): abundant liquidity, negative target rate and three-tier system
Since 2016, the BoJ, with the constant expansion of the balance sheet, which has led to abundant
excess liquidity in the system, has set the policy rate at negative values (-10 bps) and the target for
the Japanese 10-year government bond rate at 0 bps. In addition, BoJ implemented a tier system, in
which the balance of each financial institution with the BoJ is divided into 3 brackets, remunerated
respectively at a positive rate of +0.1 percent ("basic balance"), at a zero rate (“macro add-on
balance”) and a negative rate of -0.1 percent (“policy rate balance”). The latter rate coincides with
The overnight call rate is an unsecured rate and is calculated as the daily weighted average of the rates on transactions
in the unsecured overnight market.
The average value of the money market rate was equal to 15 bps; the standard deviation was low (0.22 percent) while
the coefficient of variation was moderately high (1.45).
First to 30 bps (November 2008), then to 10 bps (since December 2008).
The average value of the money market rate was equal to 10 bps; the standard deviation was very small (0.04 percent)
as the coefficient of variation (0.38).
This target replaces the previous "price stability goal" with a medium- to long-term horizon, which placed the
quantitative inflation target in a positive range of 2 percent or less. The BoJ had set this target at 1 percent.
the policy rate.53 This system aims, on the one hand, to support banks’ profitability by ensuring a
positive remuneration for a fixed amount, and on the other hand, to discourage excessive deposit of
liquidity on the BoJ’s accounts by encouraging trading in the interbank market and keeping its rate
in the band between the macro add-on rate and the basic rate.
During this period, the differential between the overnight money market rate and the policy rate was,
on average, +6 bps, with a standard deviation of 0.02 percent.54 The volatility of the differential was
increasing, showing a relevant reaction to shocks.
Figure 4A – Rates and level of central bank reserves in Japan
New Zealand
The Reserve Bank of New Zealand (RBNZ) was the first central bank to implement a quota-based
system (called "tier"). It is possible to distinguish three phases of the operational framework: the first,
prior to the adoption of the tier system, the second characterized by this system, and the third, postpandemic, in which this system was abandoned for the floor system. During the period under review,
the RBNZ’s operational framework did not provide for minimum reserve requirement.
Phase I (2002 – 2007): system with a narrow corridor
In the first period, the RBNZ had a symmetrical and narrow corridor system, with the Official Cash
Rate (OCR) as the reference rate.55 The central bank remunerated all overnight positions on the
The quotas are determined as follows: the "basic balance" quota is the average level of reserves that each institution
held in the period between January 2015 and December 2015 (fixed quota); the "macro add-on balance" is calculated
as the difference between the "basic balance" and the sum of the minimum reserve requirement, the amount of the
special operations in response to Covid-19 and the amount of the support program to loans, multiplied by the
Benchmark Ratio (determined by the BoJ on a quarterly basis); the "policy rate balance" quota is calculated on a
residual basis, by subtracting the other two portions from the volume of reserves held by banks on BoJ accounts.
The average value of the monetary rate was -4 bps and the standard deviation was very low (0.02 percent); the
coefficient of variation was 0.54.
In 1999, the RBNZ abandoned a system in which monetary policy was implemented through a target for reserves on
settlement accounts in favor of a corridor system centered on the Official Cash Rate (OCR).
settlement accounts (overnight reverse repurchase facility, ORRF) at 25 bps below the OCR, while
overnight-secured loans were offered on demand at 25 bps above the OCR. The central bank
controlled short-term money market rate close to the reference rate through daily repo and reverse
repo operations. Under this system, banks relied heavily on the intra-day liquidity provided by the
RBNZ to meet their payment and settlement needs (so-called automated intra-day repurchase facility,
Auto-repo). Given the important role of nonresident banks in the New Zealand banking system, the
central bank also used foreign exchange swap transactions to inject liquidity into the system.
During this period, the average overnight interbank cash rate (OICR) 56 was high (6.3 percent), with
a standard deviation of 0.94 percent; the coefficient of variation was small (0.15). The differential
from OCR was very small: 2 bps on average, with a standard deviation of 0.06 percent. The volatility
of the differential (at least until mid-2006) was low.
Despite its good ability to control the domestic money market rate (thanks to daily operations), the
central bank observed several signs of stress in both the foreign exchange swap market and open
market operations. These symptoms suggested that sometimes there was insufficient liquidity in the
banking system or that the liquidity was inefficiently redistributed (Nield, 2006). Some examples
were (i) relatively high incidence of delayed settlements; (ii) frequent use of standing facilities; (iii)
underbidding in RBNZ open market operations; (iv) increased demand for government bonds to be
used as collateral that led, in the face of essentially stable supply, to a surge in their prices; and (v)
significant changes between the domestic overnight interbank rate and the overnight rate available in
the foreign exchange swap market.
Consequently, the RBNZ decided to change its operational framework to a fully cashed-up system in
2006, which resulted in a substantial increase in the reserve supply and the progressive removal of
central bank intraday liquidity. The transition to the new system was implemented from July to
October 2006 and, to encourage banks to hold cash on accounts, the RBNZ switched to remunerating
all overnight balances at the OCR. As a result, between October 2006 and August 2007, the RBNZ
operated an asymmetric corridor, with the OCR standing at the corridor lower bound.57 This system
determined a liquidity hoarding phenomena by some banks, to which the RBNZ responded at the end
of August 2007 by implementing the quota system.
Phase II (2007 – 2020): quota system with a wide and asymmetric corridor
Under the quota system, banks’ overnight deposits were remunerated at the OCR up to the limit set
for each individual bank by the central bank ("tier"), mainly based on likely settlement needs in a tail
event.58 Balances held in excess of the limit accrued interest at a rate equal to the OCR minus 100
bps.59 The relatively low return on excess balances discouraged holding balances in excess of the
quota, incentivizing banks with excess balances to lend in the interbank market and invest in liquid
instruments other than central bank liquidity.
The OICR is the average interest rate of overnight transactions (secured and unsecured) reported by domestic market
price makers in a daily survey.
The spread between the central bank overnight lending rate and the OCR was reduced to 50 bps.
The determination of quotas was based on a liquidity algorithm that analyzed banks' balance sheets (including
prudential requirements), payment flow data, and intraday liquidity use. Once the limits were assigned, each bank, if
it believed they were not sufficient, could open a business case with the central bank.
The spread between the central bank overnight lending rate and the OCR was reduced to 25 bps.
At this stage, the spread between the OICR and the OCR averaged -7.5 bps, with a standard deviation
of 0.1 percent.60 The volatility of the differential was high and persistent.
Phase III (2020 – 2022): floor system with ample/abundant liquidity
The tier-based system was adopted until March 2020, when it was suspended following the launch of
a QE program (Large Scale Asset Purchases, LSAP) and targeted long-term lending operations
(Funding for Lending Programme, FLP), to counter the effects of the pandemic. This resulted in a
very large supply of reserves, which made it difficult to keep short-term rates anchored to the OCR.
As a result, the RBNZ decided to remunerate all balances at the OCR and remove all tiering limits
beyond which a lower remuneration rate was applied (Reserve Bank of New Zealand, 2020), thereby
introducing a floor system.
At this stage, the money market rate anchored to the OCR61 and the spread has narrowed to -3.7 bps,
with a very low standard deviation (0.05 percent).
Figure 5A – Rates and level of central bank reserves in New Zealand
Norway
Norges Bank is one of two central banks examined (along with the RBNZ) that has implemented a
one-sided (limited access to deposit) quota-based system. In the analyzed period it is possible to
distinguish two phases: one before the adoption of the quota system (in October 2011) and a
subsequent one. Indeed, the Norwegian central bank has maintained the quota system even after the
The OICR mean was 2.94 percent with a standard deviation of 1.86 percent, implying a coefficient of variation of
0.63. The rate was persistent and unresponsive to shocks.
The OICR mean was 0.49 percent with a standard deviation of 0.45 percent, implying a coefficient of variation of
0.92.
outbreak of the pandemic crisis (unlike New Zealand). There is no reserve requirement in the
Norwegian system.
Phase I (2002 – 2011): system with asymmetric corridor
Prior to 2011, the operational framework of Norges Bank was based on an asymmetric corridor
system. Banks received interest on all deposits with the central bank at the sight deposit rate, which
is also the central bank’s policy rate. The sight deposit rate was the floor for shorter-term money
market rates, as banks had no incentive to lend out reserves at an interest rate lower than the rate they
would receive from the central bank. Similarly, the overnight lending rate (the so-called D loan rate)
represented the ceiling for short-term money market rates. In this system, the central bank ensured
that there was an adequate surplus of reserves in the banking system to push overnight market rates
towards the sight deposit rate. Norges Bank provided liquidity through F loans when it believed that
more reserves were necessary. The interest rate on the F loans was usually a few basis points higher
than the sight deposit rate.
During this period, the spread between the overnight money market rate and the sight deposit rate
averaged 47 bps;62 the standard deviation was small (0.24 percent), so that the coefficient of variation
was equal to 0.51. Based on GARCH, the volatility of the differential was high, persistent and with a
relatively low response to shocks.
Phase II (2011 – 2022): quota system
With the use of the asymmetric corridor system, over time Norges Bank observed that the demand at
the F-loan auctions was higher than what would have been necessary for the banks to settle payments
in the domestic settlement system. In particular, after the great financial crisis, when the central bank
temporarily injected large amounts of liquidity into the financial system, the level of reserves rose
strongly and continued to be above pre-crisis level. The central bank also noted a reduction in reserve
redistribution activity in the interbank market. There were signs that the short-term money-market
rate had to rise considerably before a lender would enter it.
Therefore, as from October 2011, Norges Bank implemented the quota system, where each bank’s
"quota" is determined to be sufficient to meet daily settlement requirements and is remunerated at the
reference rate. The system also foresaw a penalizing rate (reserve rate, 100 bps below the sight deposit
rate) at which reserves deposited by banks above the pre-determined quota are remunerated in order
to incentivize banks to lend liquidity in the interbank market.
Following the pandemic crisis, Norges Bank carried out several extraordinary refinancing operations
with maturities of up to 12 months. However, it was able to maintain the quota system because the
excess liquidity injected was drained from the banking system using daily deposits with a one-day
maturity, with rates close to the policy rate.
During the floor system period, only the Norwegian Inter-Bank Offer Rate (NIBOR) with maturity tomorrow next
(data provided on the central bank’s website) was available as a reference of the short-term money market rate. The
central bank provides two data series: 1) nominal rates; 2) effective interest rate. This second series was used in this
work. Because of the tomorrow next deadline, the series has been lagged by one day. NIBOR was based on indicative
rates quoted by a group of banks active in the Norwegian market. Since the participating banks quoted, in general,
wide bid-ask spreads, the actual rates at which market transactions took place were often closer to the sight deposit
rate than what was apparent based on the NIBOR.
Under the quota system, the spread between the overnight money market rate (NOWA rate)63 and the
sight deposit rate averaged 0 bps; the standard deviation equal to 0.10 percent. If we exclude the
observations at the end of the quarter, the differential averaged around -2 bps, with a standard
deviation of 0.04 percent. Based on the GARCH, spread volatility was very low, persistent and with
limited response to shocks.
Figure 6A – Rates and level of central bank reserves in Norway
Sweden
During the period under review, Riksbank has operated under a corridor system, in which there are
no minimum reserves. The Riksbank uses standing facilities and open market operations to steer the
short-term interbank interest rate64 towards the target set by the central bank, the reference rate (repo
rate, which from 2019 has been renamed policy rate) to indicate where the market rate should be
positioned.
Before the Riksbank payment system (RIX) daily closure, banks' accounts must be balanced. If
necessary, banks can borrow or deposit funds with the Riksbank overnight through the central bank's
marginal lending and deposit facilities. The corridor system was supposed to ensure the liquidity
needed for the proper functioning of RIX, where banks, clearing organizations, and the Swedish
With the introduction of the quota-based system, Norges Bank, in cooperation with Finance Norway, began collecting
rates on reserve trades by banks in the overnight market, which have since been published as the Norwegian Overnight
Weighted Average (NOWA) rate. The NOWA is a weighted average of interest rates on overnight-unsecured
interbank loans traded in NOK and is the first money market rate in Norway based on actual trades. The monetary
rate series used in the analysis is obtained by sequentially merging NIBOR and NOWA (for October and November
2011, the transition period between the two rates, an average has been calculated between the two).
The benchmark money market rate was the STIBOR until September 2021, when it was replaced by the SWESTR.
While the former was a rate based on the intentions of market participants, the latter is based on rates of actual
transactions.
Treasury have their accounts. Banks can have intraday credit up to the maximum amount of collateral
pledged as collateral.
We have identified three phases of the operational framework, which differ in the width of the
corridor, in the use of fine-tuning operations or in the liquidity injected into the system.
Phase I (2002 – 2008): system with a wide symmetrical corridor and liquidity deficit
In the first period, the corridor was wide (the standing facilities rate was set at ± 75 bps relative to the
policy rate). The system had a structural liquidity deficit and the central bank used weekly one-week
repo operations to provide the necessary liquidity to the banking system. During this period, the
spread between the money market rate and the policy rate averaged 12 bps, with a low standard
deviation (equal to 0.06 percent); the volatility of the spread was also low, at least until the end of
2007.
Phase II (2008 – 2019): system with symmetrical corridor and use of fine-tuning operations
With the outbreak of the financial crisis in October 2008, Riksbank launched a longer-term Swedish
krona loan facility to ease pressure on the term funding markets. For a brief period in early October
2008, these funding pressures pushed STIBOR above the Riksbank's marginal lending rate. With the
new operations, overnight balances held at Riksbank increased substantially. Subsequently, Riksbank
relied frequently on daily fine-tuning operations to anchor the money market rate in the middle of the
corridor. The fine-tuning rate was 10 bps below the policy rate. To drain the liquidity surplus,
Riksbank also started issuing certificates, which have become a regular instrument for conducting
monetary policy, with weekly issuance.
From 2015 (and up to the beginning of 2020), the policy rate was negative (with a minimum of -0.5
percent) and a securities purchase program was launched.
Over the whole period, the spread between the money market rate and the policy rate averaged 5 bps,
with a fairly high standard deviation (equal to 0.13 percent). After the outbreak of the global financial
crisis, the volatility of the spread has been significant; since 2013 it has decreased, except for a few
spikes at the end of the year.
Phase III (2019 – 2022): system with narrow corridor and ample/abundant liquidity
From the end of 2019, Riksbank stopped conducting daily fine-tuning operations and reduced the
width of the corridor, narrowing it to ?/+10 bps below/above the policy rate, with the deposit facility
rate effectively replacing the fine-tuning operations rate. The objective was to reduce the sensitivity
of the overnight rate to the liquidity distribution in the banking system at the end of the day, to ensure
that it remained pegged to the policy rate, and to automate the deposit and lending process at the end
of the day, without the operational risk characterizing fine-tuning operations.
With the outbreak of the pandemic, Riksbank conducted 12-month financing operations and expanded
its securities purchasing activity. As a result, there was a considerable increase in the liquidity
deposited at the central bank at the end of the day. The liquidity surplus was partially drained by the
increase in certificate issuances. Riksbank certificates with one-week maturity are currently offered
at the policy rate.
The spread between the money market rate and the policy rate averaged -6 bps, with a standard
deviation of 0.12 percent. Therefore, the money market rate moved towards the lower limit of the
corridor; at irregular intervals it was a few basis points (maximum 2 bps) below the deposit facility
rate. Banks seemed to have a preference for overnight reserve holdings, but this interfered with the
steering of rates towards the policy rate at the middle of the corridor. The volatility of the spread was
low (with the exception of one anomalous phenomenon on the last day of 2021).
Figure 7A – Rates and level of central bank reserves in Sweden
United Kingdom
During the period under review, the Bank of England (BoE) used three distinct operational
frameworks for conducting monetary policy. Initially, the BoE operated with a symmetric corridor
system in which banks had an incentive to hold zero reserves on their overnight balances and the
central bank provided the necessary liquidity through open market operations at the reference rate. In
2006 the central bank adopted the system of voluntary reserves (VR), and then suspended it from
March 2009, due to the excess liquidity provided through the launch of the quantitative easing (QE)
programme. Since that date, the operating framework has been based on a floor system, in which all
reserves are remunerated at the reference rate.
Phase I (2002 – 2006): system with a wide symmetrical corridor without minimum reserves
In the first period, the liquidity required by the banks was provided by very frequent open market
operations (between 2 and 4 operations every day), conducted with a limited number of counterparties
(less than 20 “discount houses”) at the reference rate (Bank rate). Reserve balances were not
remunerated and there was no minimum reserve requirement.65 The system had had a wide
Banks were required to deposit a minimum amount of cash on accounts at the BoE. This measure was known as Cash
Ratio Deposits (CRDs). The liquidity was invested in government bonds with the aim of financing the BoE's
institutional functions. Whitesell (2006) identifies these amounts as required reserves. However, their nature appears
symmetrical corridor (±100 bps relative to the bank rate) until 2005. It had subsequently been
narrowed (±25 bps relative to the Bank rate).
The system was not particularly efficient. The volatility of the spread between the overnight money
market rate (SONIA) and the Bank rate was very high and persistent: against an average of -4 bps,
the standard deviation was 0.37 percent.66
The implementation of a corridor system usually generates a reduction in the volatility of the spread
if there is unrestricted access to the standing facilities (Bindseil, 2004). In the UK, however, the
availability of central bank lending at the end of the day was limited to what the BoE had foreseen
would be necessary. Thus, the high volatility of the market rate relative to the policy rate could be
due to BoE’s forecasting errors or high market concentration, which allowed some counterparties to
obtain a large share of the limited liquidity made available and therefore behave as marginal suppliers
for the rest of the intermediaries (Osborne, 2016).
Phase II (2006 – 2009): voluntary reserve system
To deal with this excessive volatility, in May 2006 BoE implemented a system of voluntary reserves.
If the effective reserves were positioned, as an average of the maintenance period, within the tolerance
band of the voluntary reserves, they were remunerated at the Bank rate. If reserves were outside the
tolerance band (above the upper limit or below the lower limit), remuneration suffered a penalty:
excess reserves were remunerated at the official rate minus a spread, while banks in deficit would
have to borrow from the BoE at the official rate plus the same spread, generating a sort of symmetrical
corridor.
The BoE’s tolerance band was limited to only ±1% around the target until September 2007. In times
of market tensions, the BoE widened the band to ±60% and offered additional liquidity above banks'
aggregate reserve targets through its open market operations.
These measures contributed to reduce, at least in part, the volatility of the interbank rate. Over this
period, the average spread was -0.5 bps and the standard deviation was reduced to 0.20 percent. The
GARCH analysis shows, especially in the initial phase of the outbreak of the financial crisis (August
2007), episodes of extreme volatility, which are also reflected in the estimate of the reaction
coefficient to innovations.
Phase III (2009 – 2022): floor system
Following the bankruptcy of Lehman Brothers in September 2008, volatility in the interbank market
had picked up again. Several factors had contributed to this phenomenon. Banks' demand for reserves
had become broader and less predictable and banks were reluctant to resort to the Overnight Lending
Facility for fear of the stigma effect. In addition to the widening of the tolerance band, the BoE
increased the frequency of its liquidity operations. Between August 2008 and March 2009, the Bank
rate was reduced by 4.5 percentage points (to 0.5 percent).
Following the launch of a large-scale asset purchase (QE) programme, the voluntary reserve system
was replaced by a floor system, in which all available reserves on bank balances were (and are)
different from classical required reserves, given their selected purpose and the impossibility of banks to use them in
the payments system.
The b1 GARCH coefficient was high, while the squared innovation coefficient (a 1) was relatively low.
remunerated at the Bank rate, without minimum reserve requirement, except for central counterparties
and international central depositories. For monetary policy counterparties there are no limits
(minimum and maximum) on the balance of accounts with the BoE, which are remunerated at the
Bank rate. However, only in the event of a negative interest rate, the BoE envisages the possibility of
maintaining a minimum balance and adopting a tiering remuneration system.
The change from a voluntary reserves system to a floor system was motivated by the fact that, in a
normal situation with a stable liquidity stock, the former ensured effective control of the interest rate
on the money market, but in times of stress, with strong increases in liquidity demand, it was not very
efficient. The BoE system was exposed to the interaction (endogeneity) of the targets with the
tolerance bands, which violated the condition of separability of the remuneration function.
The floor system, which satisfies the demand for central bank reserves at all times, made it possible
to keep the interest rate on the money market anchored to the bank rate. The volatility of the difference
between the two rates decreased considerably. The standard deviation was only 0.03 percent,
compared to an average of -4 bps. Volatility, measured by the GARCH, was very low and
significantly explained by the response to shocks. The negative average of the spread is mainly due
to the fact that in this period the money market also includes entities that are not eligible for monetary
policy operations (such as, for example, money market funds).
The BoE has stated that it will continue to use a variant of the current floor system to control shortterm interest rates,67 based on the following three considerations (Bank of England, 2022 and 2018a;
Hauser, 2021 and 2019). First, the current framework has proven effective in keeping market rates in
line with the monetary policy reference rate. Second, maintaining key elements of the current system
guarantees business continuity and simplicity for both counterparties and the central bank. Third, the
system would remain effective in the event of increases in the supply of reserves in the future, either
due to future asset purchases or the liquidity facilities that the central bank would need to activate to
deal with any shocks. The variant consists in the implementation of a "preferred minimum range of
reserves (PMRR)", defined as the optimal range for the amount of reserves in the system calculated
by the central bank, at the lower limit of which reserves start to become scarce.68 This range will take
into account all the buffers banks hold above their actual minimum needs to address uncertainties
about potential stressed outflows, access to wholesale markets and market risks.
See https://www.bankofengland.co.uk/-/media/boe/files/monetary-policy-summary-and-minutes/2018/june-2018.pdf
and https://www.bankofengland.co.uk/monetary-policy-summary-and-minutes/2022/august-2022 .
Since central bank reserves ensure the correct and efficient functioning of the payment system and constitute the first
safeguard for unexpected liquidity shocks, the BoE, in determining the desired amount of central bank reserves,
identifies two extreme situations characterized by shortage and excess reserves, thus delineating the PMRR. Within
this interval, it is assumed that as reserves vary, the overnight rate will have minimal deviations from the reference
rate. The BoE’s framework relies on a new credit operation called the "Short-term Repo Facility", which is a 7-day
open market operation, priced at the BoE's Bank rate on deposits + 0 bps (with the rate indexed to the bank rate). The
framework may be considered as a zero corridor system, where reserves are remunerated daily but provided weekly
with the same rate against HQLA collateral. This facility will ensure that excess liquidity remains within the fixed
liquidity range, while the BoE, in parallel, pursues the reduction of its government bond portfolio. Furthermore, the
BoE still has in place two standing facilities for urgent liquidity needs between the weekly operations, priced with a
25 bps spread around Bank rate.
Figure 8A – Rates and level of central bank reserves in United Kingdom
United States
The Federal Reserve (Fed)’s monetary policy operating framework has radically changed over the
past 20 years. Prior to the 2008 global financial crisis, the Fed monetary policy was based on a wide
asymmetric corridor system with scarce reserves, and the liquidity supply was carefully calibrated to
maintain control of short-term interest rates. Subsequently, Fed implemented a floor system, where
active reserve management no longer plays a role in rate control, but rather money market rates are
affected by the Fed's setting of "administered" rates. Initially, reserves were ample but after the
pandemic crisis they became abundant.
Phase I (2002 – 2008): system with a wide asymmetric corridor and scarce reserves
In the first period, the Federal Open Market Committee (FOMC) of the Fed communicated the
monetary policy stance by announcing a target for the federal funds rate, which represented the
reference rate. The Fed used open market operations (which could be daily) to make the necessary
adjustments to the supply of reserves so that the money market rate (effective federal funds rate,
EFFR)69 was as close as possible to the target set by the FOMC. The monetary policy implementation
system was based on an asymmetric corridor system, in which the discount rate (primary credit rate)
represented the upper part of the corridor (+100 bps relative to the reference rate up to August 2017
and then +50 bps) and a zero fixed interest rate was the bottom. This is a key feature that distinguished
this corridor from those implemented by many other central banks. Since the minimum rate was fixed,
the width of the corridor varied with the official rate.
The Fed had reserve requirements among its instruments. A reserve ratio of 10 percent was applied
to deposits (net transaction deposits). If a financial institution had deposits below $48.3 million or
$7.8 million (2007 data; these thresholds varied over time) the required reserve ratio was reduced to
The EFFR is calculated as the weighted median of the rates on overnight transaction between financial institutions
licensed to operate in the money market.
3 percent or 0 percent, respectively. The maintenance period was generally 14 consecutive days and
it was possible to carry over balances from one maintenance period to another.70
In this framework, banks had an incentive to hold as little reserves as possible on account balances
with the Fed, as these did not accrue interest, since the Fed did not have the authority to pay interest
on reserves (even minimum reserves).71 Thus, the banking system operated with a shortage of
aggregate reserves and relied on the redistribution of these through an active interbank market.
During this period, the average spread between the EFFR and the federal funds rate was 0 bps, with
a standard deviation of 0.11 percent. The volatility of the spread, as measured by the GARCH, was
relatively low, but increasing from August 2007.
Phase II (2008 – 2020): ample/abundant reserves and floor system
With the onset of the global financial crisis, the Fed implemented new liquidity instruments and
conducted Large Scale Asset Purchases to improve conditions in financial markets and stimulate the
economy. These measures generated a significant increase of reserves in the banking system. As
Figure 9A shows, banking system reserves exceeded $800 billion at the start of 2009, up from around
$10 billion in the pre-crisis period. Over the next several years, the FOMC continued to conduct asset
purchases to promote a stronger economic recovery, and reserves continued to increase until the end
of 2014.
With the significant increase of reserves in the banking system, it was difficult for the Fed to maintain
control of rates through the previous corridor system. Small changes in the supply of reserves no
longer affected the Fed funds rate. With the Fed's authorization to pay interest on reserves to banks,
introduced in October 2008, the Interest Rate on Reserve Balance (IORB) became the policy rate, as
it set a level below which the banks have no interest in lending. In the new system, which represented
a floor system, the operational target of the Fed was to maintain the money market rate within a predetermined range (Fed funds target range).
During this period, the spread between the EFFR and the IORB fluctuated moderately: the average
was -10 bps with a standard deviation of 0.08 percent. A negative relationship can be found between
Up to a maximum given by the greater between $50,000 and 4 percent of the total requirement. The total requirement
is the sum between the required reserves and the Contractual Clearing Balance (CCB). The latter supplemented the
minimum reserve system with carry-over provision. The CCB was an additional amount, with respect to the mandatory
requirement, that a bank (depository institution, DI) could voluntarily hold in its account at the central bank; it was
remunerated in the form of credits, which could be used to pay for services that the central bank offered to banks, such
as, for example, the cost of Fedwire transactions. The amount of the CCB had to be determined before the start of a
maintenance period. A DI's daily average clearing balances in a given maintenance period had to be within the
expected tolerance band around its requirement. The CCB system reduced the likelihood that a bank would have to
resort to the Fed's overnight overdrafts. Indeed, in order to ensure the proper functioning of the payment system, the
Fed allows banks to use daily overdrafts which, if not reset to zero at the end day, become overnight overdrafts. This
exposure to the Fed can be risky, as it is not mandatory to collateralize this credit. Therefore, in order to discourage
banks from using the overnight overdraft, the Fed applies a very high penalty rate equal to the primary credit rate +
4% for the first three days of use during the year; from the fourth onwards, 100 bps are added for each additional day
of use. Sometimes, the additional requirement could be imposed by the central bank (required clearing balance, RCB),
rather than being voluntary, if a DI had a history of frequent overdrafts (during the day or overnight). For more details,
see M. Naber, R. Sambasivam, and M-F. Styczynski (2017).
The authorization to pay interest on reserves was granted to the Fed with the Financial Services Regulatory Relief Act
of 2006. This act provided that the start date of the innovation would be October 2011, then anticipated to October
2008.
the spread and the level of reserves. Volatility, as measured by GARCH, was low, at least up to early
2018. Since 2015, spikes in volatility were recorded at month-end dates.
As a consequence of the large availability of reserves, only a small fraction of money market
transactions was conducted between banks (depository institutions), since none of them needed to
borrow on the market to fulfill reserve requirements. The money market was dominated by other
investors (such as government sponsored enterprises, GSEs, and money market funds) to get a return
on their overnight cash balances. This explains the negative differential between the EFFR and the
IORB, which forced the Fed to implement the Overnight Reverse Repo Purchase (ON RRP) in
September 2014, a facility that became available also to non-depository institutions, with the goal of
providing a lower bound on money market rates.72 The IORB for depository institutions and the ON
RRP rate for other financial institutions represent, respectively, a floor below which it is not
convenient to lend funds. The spread between the two rates was on average 20 bps between September
2014 and March 2020.
Between 2017 and 2019, as the supply of reserves decreased, the spread between the EFFR and the
IORB narrowed; in 2019, the EFFR exceeded the IORB for short periods.
To steer money market trading without changing the monetary policy stance (ensuring that the EFFR
remains in the target range), one option available to the Fed is to implement “technical adjustments”.
A technical adjustment is a change in the IORB and/or the ON RRP rate that aims at improving the
effectiveness of monetary policy implementation by promoting money market trading at rates within
the target range, rather than changing the stance of monetary policy itself. It can also coincide with a
change in the FOMC’s target range.73
Phase III (2020 – 2022): abundant reserves and floor system
In recent years, two stress situations can be isolated: one in September 2019 with an unexpected
shortage of liquidity on the repo market74 and another in March 2020 with the outbreak of the
pandemic. In both cases the tensions generated an increase in the EFFR-OIRB spread; in the first case
over 20 bps with the money market rate outside the federal funds target range. The exceptional
measures in response to these events, with injections of liquidity, lowered the EFFR-OIRB spread;
the latter returned to being slightly negative in the period following the outbreak of the pandemic,
with a liquidity level that reached its maximum peak of over $4 trillion.
Under this facility, the Fed conducts reverse repo transactions at a pre-announced bid rate set by the FOMC, offering
US Treasuries held in the System Open Market Account (SOMA) portfolio. A wide range of counterparties – including
money market funds and GSEs – can participate in this facility. Each counterparty may invest funds up to an assigned
limit. This facility is used exclusively to drain the liquidity available on the accounts of these counterparties overnight.
A technical adjustment can occur when there is no change in the target range and the Fed adjusts the administered
rates (IORB and/or ON RRP rate) to the existing target range. It can also coincide with a change in the FOMC's target
range. In this case, one or both of the administered rates are changed by a different amount than the target range shift.
A technical adjustment does not imply a change in the monetary policy stance. In early 2018, the IORB was set at the
upper limit of the target range, as it had been since the FOMC moved to a target range in late 2008. As the supply of
reserves declined in the first half of the 2018, the EFFR increased, very close to the top of the target range. With the
EFFR floating just five bps from the upper limit of the target range, the Fed implemented its first technical adjustment
in June 2018. The IORB was lowered to five bps below the top of the range, when the FOMC increased its target
range and ON RRP rate by 25 bps, while the IORB was only increased by 20 bps. The EFFR responded quickly to
IORB adjustments.
For more details on the market events of September 2019, see Afonso et al. (2020a).
Following the outbreak of the pandemic, the minimum reserve requirement ratio was set to zero. To
keep the money market rate in the target range, the Fed also intervened with two technical adjustments
(in 2020 and 2021). As cash was traded in the market at rates close to the lower limit of the range,
the IORB and ON RRP rates were raised to support trading in the Fed funds market within the range.75
In the most recent period, the volatility of the spread has been very low. In January 2019, the Federal
Reserve announced that the implementation of long-term monetary policy would continue to be
accomplished under a "large reserve" system, in which "administered" interest rates (such as the rate
paid on reserves) are the main instruments and where, instead, “active” adjustments in the supply of
reserves are no longer necessary to implement it.76 With such a system, the Fed does not face the
trade-off between controlling interest rates and providing liquidity to the market during times of stress
(Afonso et al., 2020b; Ihrig et al., 2020).
Figure 9A – Rates and level of central bank reserves in United States
In January 2020 and mid-June 2021, the IORB and ON RRP rates were increased by 5bps after the EFFR had moved
within 5bps above the lower bound of the range. The 2020 and 2021 technical adjustments were effective in supporting
trading in the Fed funds market at rates within the target range.
See the statement “Statement Regarding Monetary Policy Implementation and Balance Sheet Normalization.”
Appendix III – Analysis of the relationship between central bank reserves and money market
rates
The analysis aims to investigate the relationship between:
i) central banks' decision to adopt a floor system and the reserves-to-GDP ratio (whose trend for
various countries is shown in Figures 10A and 11A).77 Additionally, it explores the link with the
spread between the money market rate and the rate applied to bank deposits with the central bank;
ii) the reserves-to-GDP ratio and the spread between the money market rate and the rate applied to
bank deposits with the central bank.
Figure 10A – Reserves-to-GDP ratio
Figure 11A – Reserves-to-GDP ratio (Japan excluded)
A logistic model was estimated to identify how the ratio between reserves and GDP influences the
probability that a central bank adopts a floor system. We considered a dummy variable which was
equal to 1 in periods in which a central bank adopted a floor system as dependent variable.78
?(????? ??????,? = 1) = ? (? + ? ×
?????????,?
????,??1?
The GDP at the denominator of the ratio is annual and the value at the end of the previous year was considered. Figure
11A excludes Japan to better highlight the performance of other countries.
To identify floor system, we used official statements of the central banks, if any; in absence of official statements, we
considered speeches/communications of the central banks’ members or other central banks’ documents (see table 10A
for details).
where ?(????? ??????,? = 1) is the probability that the central bank of country i adopts the floor
system at time t and F is the logistic distribution function. The estimate has a positive and significant
coefficient ?, which assumes a value of 0.67 (Table 2A).
Table 2A: Estimate of the logistic model with reserves on GDP ratio
Dependent Variable
dummy_floor_system
Reserves / GDP
0.672***
(0.008)
Number of observations
Log Likelihood
Pseudo R2
Akaike Inf. Crit.
40,428
-11,996
0.485
23,995
Standard errors in parentheses: ***p