Alexander Kroeger, John McGowan, and Asani Sarkar
T P-C M
P I
F
F R B N Y E P R , . , O
T
he Federal Reserve’s (the Fed’s) operating framework for
monetary policy changed during the nancial crisis of
2007-09. is change occurred because the Fed implemented
an accommodative monetary policy to facilitate economic
recovery from the crisis by substantially increasing the
amount of reserves in the banking system and by reducing
interest rates to close to zero (Bech and Klee 2011). By
comparison, in the pre-crisis period the supply of reserves
was relatively scarce. e aim of this article is to assess the
Fed’s monetary policy framework prior to the crisis in order
to better understand the changes in the implementation of
monetary policy since the crisis.
A monetary policy framework is a means of implement-
ing a central bank’s monetary policy (Bindseil 2004). Such a
framework consists of an operational target and an operating
framework for achieving the target. e Fed’s statutory
mandate in conducting monetary policy is to promote price
stability consistent with full employment.
1
To implement this
mandate, the Federal Open Market Committee (FOMC) sets
a target for the overnight rate in the federal funds market,
where banks trade reserve balances (“reserves”), which are
deposits held by banks at the Fed.
2
Changes in the federal
• Before the 2007-09 nancial
crisis, the Fed’s operating
framework for monetary policy
reected a banking system in
which the scarcity of reserves
meant that small changes
in reserves would affect fed
funds rates.
• The authors assess the
framework and nd that it
met the Fed’s monetary policy
objectives by keeping rates
close to target but had certain
negative effects on nancial
market functioning and
employed operating procedures
that were rather opaque and
inefficient.
• During the crisis, the Fed
boosted reserves in a bid to
foster economic recovery, and
this increase necessitated
changes in how the Fed
conducts monetary policy.
The new approach has also
controlled rates well since the
crisis, suggesting that alternate
frameworks can be effective.
Alexander Kroeger is a former research assistant, John McGowan an assistant vice president,
and Asani Sarkar an assistant vice president at the Federal Reserve Bank of New York .
Email: alexrk[email protected]m; johnp.mcgowan@ny.frb.org; asani.sarkar@ny.frb.org.
e views expressed in this article are those of the authors and do not necessarily reect
the position of the Federal Reserve Bank of Ne w York or the Federal Reserve System.
To view the authors’ disclosure statements, visit https://www.newyorkfed.org/research/
epr/2018/epr_2018_pre-crisis-framework_sarkar.
OVERVIEW
F R B N Y E P R , . , O
T P-C M P I F
funds rate are, in turn, expected to be transmitted to other interest rates and, ultimately, to the
real economy. e pre-crisis operational framework consisted of monetary policy instruments
(mainly the conduct of open market operations, or OMOs) and procedures for using these
instruments to encourage banks to trade fed funds near the stated target rate. e Ne w Yor k
Fed’s Open Market Trading Desk (“the Desk”) carried out OMOs on a daily basis to keep the
overnight fed funds rate close to its target.
e fed funds market represents the market for bank reserves. Fluctuations in the fed funds
rate reect changes in the demand for and supply of reserves. Prior to the crisis, the demand
for reserves arose mainly from banks’ need to meet uncertain intraday payment ows, aer
satisfying minimum reserve requirements. Because no interest was paid on reserves, banks
wished to minimize their reserve holdings. e aggregate demand for reserves was sensitive
to interest rates since reserves were scarce—the Fed supplied only a small amount of reserves
in excess of what banks were required to hold in the aggregate. e daily variation in the
supply of reserves was mainly determined by so-called autonomous factors (such as currency
in circulation) outside the direct control of the Fed. erefore, the Desk’s job was to forecast
the evolution of the autonomous factors and the demand for reserves and, on an ex ante basis,
supply enough reserves to keep the market for reserve balances in equilibrium. e aggregate
amount of reserves was distributed to individual banks through the fed funds market.
In this article, we show that during the pre-crisis period the Desk was generally successful
in achieving its primary objective of meeting the fed funds target. Overnight rates were
generally close to the target fed funds rate, even during periods of relatively high liquidity
demand. Further, when the fed funds rate, on occasion, deviated from its target (such as at
the end of quarters), it reverted to the target within a day or two. Finally, changes in the fed
funds rate were quickly transmitted to other overnight money market rates.
In addition to the primary objective of controlling its operational target, a central bank
might consider other criteria to evaluate the eectiveness of its operational framework:
eciency (that is, meeting objectives with as few resources as possible), transparency (that
is, operating in a manner well understood by market participants), universality (that is,
being able to implement monetary policy under a range of economic conditions), and the
promotion of nancial stability (that is, ensuring that the operational framework does not
impair market functioning).
3
While the pre-crisis monetary policy framework was successful in meeting its monetary
objectives, the associated operational procedures were complex and opaque. e operational
framework relied on a discretionary and interventionist approach (Logan 2017) based on
daily management of the supply of reserves that required detailed market intelligence and
expert judgment (Bernanke 2005). e Desk had to provide daily forecasts of reserve demand
and supply over multiple days, and to conduct repurchase (“repo”) or reverse repo operations
almost daily. Reserve demand was dicult to forecast daily, and even predictable changes
required OMOs on most days (Logan 2017). Forecasting the autonomous factors that caused
daily variations in the supply of reserves was also challenging. Liquidity management in such
a framework appears more complex than in a “symmetric corridor system” with standing
deposit and lending facilities, as operated by some other central banks.
4
e system also lacked
transparency, since the Fed, unlike other central banks, did not publish its forecast of auton-
omous factors (Hilton 2008). Regarding universality, the pre-crisis operational framework
faced diculties in the post-crisis environment when demand for reserves became highly
F R B N Y E P R , . , O
T P-C M P I F
volatile (Hilton 2008). Also, supplying reserves to meet forecasted demand became impractical
post-crisis, when the amount of reserves in the banking system exceeded demand for reserves
by a wide margin.
Turning to the goal of not impairing nancial market functioning, we rst focus on the
functioning of payment systems. In particular, in the pre-crisis period the Fed routinely
extended large amounts of (sometimes unsecured) intraday credit to banks to meet payment
system demands, for a fee. Since banks needed these funds for only a few hours a day, they did
not nd it cost eective to borrow overnight in the fed funds market. While these “daylight
overdras” were necessary to facilitate payments, they also exposed the Fed to the potential
for loss. For the banks, the need to avoid overdras and meet reserve requirements, combined
with the lack of interest payments on reserves, implied that their cash management system was
rather costly (Logan 2017).
e Fed aected nancial markets in two other areas prior to the crisis: asset eligibility criteria
for OMOs and money market functioning. Assets that were eligible for purchase by the Desk,
including repo operations, might benet from enhanced liquidity and the ability to obtain central
bank credit, compared with ineligible assets. Since the Fed accepts only highly liquid assets in its
OMOs, including Treasury and agency securities, any distortionary eects on asset prices were
likely minimized. Regarding money market functioning, the scarcity of reserve balances prior
to the crisis (relative to the required and precautionary demand for reserves) resulted in large
trading volumes in the fed funds market because, toward the end of the trading day, banks with
surplus reserves had an incentive to trade with banks with too few reserves. While it is unclear
whether an active fed funds market should be a goal of a monetary policy framework, the market
activity likely facilitated both rate discovery (that is, the determination of an equilibrium rate
through trading) and the quick transmission of the target rate to related money markets.
e article is organized as follows. Section 1 discusses the basic economic premise
underlying the pre-crisis framework, and then describes how rate determination in actuality
deviated substantively from the textbook example. Section 2 details how the framework was
implemented in practice, and includes a description of the role of the reserve maintenance
period. We discuss the eectiveness of the framework in meeting the primary monetary policy
objectives in Section 3 and evaluate how well the framework met the objectives of operational
eectiveness and nancial stability in Section 4. Section 5 concludes with a brief summary and
remarks on some aspects of the pre-crisis framework that have changed since the crisis.
1. T E P-C M P
O F
In this section, we discuss the economic foundation of the monetary policy operating frame-
work in terms of the demand for and supply of reserves. We show that the pre-crisis monetary
regime can be viewed as managing the supply of reserves so that equilibrium is maintained on
the steeper, relatively inelastic portion of the demand curve for reserves. However, we further
note how the actual framework deviated signicantly from this idealized model.
e aggregate demand for reserves is plotted in Chart 1, with the horizontal axis represent-
ing the total reserve balances of banks and the vertical axis representing the eective federal
F R B N Y E P R , . , O
T P-C M P I F
funds rate (EFFR), calculated as a volume-weighted average rate of each business day’s fed
funds transactions.
5
A target EFFR was the Fed’s primary monetary policy tool prior to the
crisis. In the fed funds market, banks traded reserves with each other on an unsecured basis,
typically with an overnight tenor. e supply of fed funds was determined exogenously (from
the point of view of market participants) by the Federal Reserve, which, through OMOs, tar-
geted a specic amount of reserves, R*, on a daily basis in order to meet the Desk’s forecast of
reserve demand.
e demand for reserves is downward sloping, reecting the opportunity cost of holding
reserves, except at the ceiling, R
U
, and the oor, R
L
, of the EFFR, where it is at (Keister,
Martin, and McAndrews 2008). Reserve requirements necessitated that banks hold minimum
reserve balances on average (as a percentage of their net transaction accounts) in their
accounts with Federal Reserve Banks. However, because of the uncertainty of payment ows,
banks could not meet their requirements exactly. In deciding how much additional reserves to
hold, banks had to balance the income forgone from holding excess reserves against the cost of
borrowing fed funds at the EFFR. Higher levels of the EFFR increase the opportunity costs of
holding reserves and reduce the demand for reserves.
e demand for reserves is at at the lower and upper bounds of the EFFR. e lower
bound for the EFFR was zero because banks had no incentive to lend reserves at a negative
rate, since they could earn zero interest by simply keeping reserves in their Fed account. At R
L
,
where the demand curve intersects the horizontal axis, banks hold sucient reserves to meet
all possible payment needs. us, banks are indierent to holding any reserves to the right
of R
L
because the opportunity cost of holding reserves is zero. Since the discount window’s
primary credit facility is an alternative to the fed funds market as a source of reserves for nan-
cially sound banks with adequate collateral,
6
the primary credit rate (which exceeds the target
rate) acts as the upper bound above which banks would not borrow in the fed funds market.
7
When the EFFR equals the primary credit rate, banks are indierent between holding reserves
and borrowing at the discount window, and so the demand curve is at to the left of R
U
.
Before the crisis, the Federal Reserve carried out monetary policy by operating in the
downward sloping region of the demand curve for reserves. is implies that the Fed raised
rates by draining reserves (decreasing supply) and lowered rates by adding reserves (increasing
supply) to the system. Empirically, in a simple plot of the eective federal funds rate against excess
C
The Market for Reserves
0
0
Effective federal funds rate
Primary
credit
rate
Supply
of reserves
Demand for
reserves
Reserve balances
R*0
Target
rate
R
U
R
L
F R B N Y E P R , . , O
T P-C M P I F
reserves (with both averaged over maintenance periods, a two-week time period over which
reserve requirements are applied), the tted relationship is negative and statistically signicant
(see Chart 2).
However, as Chart 2 notes, excess reserve balances explain only 10 percent (R
2
= 0.10) of
the variation in the fed funds rate. e high level of noise in the relationship between rates and
reserves in the data indicates that, in practice, the relationship between reserve balances and
the fed funds rate is more complicated than the stylized theory illustrated in Chart 1 (as also
noted by Judson and Klee [2010]). One complication is that the distribution of reserves across
banks matters. Since larger institutions traded excess reserve balances more actively than
smaller institutions, a temporary concentration of reserves in large institutions could have
entailed lower rates. erefore, the aggregate amount of reserves was not the only variable that
mattered. Nevertheless, Ennis and Keister (2008) show how the basic conclusions from the
simple analysis do not change, even aer accounting for bank heterogeneity.
An additional complication is that the demand for reserves likely shis over time because
of both long-term changes in the need for liquidity (for example, as a result of technological
and regulatory changes) and short-term uctuations in liquidity needs and expectations of
rate changes throughout the maintenance period. For instance, Carpenter and Demiralp
(2006a) present evidence of increases in demand for bank reserves in expectation of an FOMC
rate increase, illustrated as the shi from D
1
to D
2
in Chart 3. ese demand movements
Sources: Federal Reserve Bank of St. Louis; authors’ calculations.
Notes: The chart shows maintenance period averages from January 1, 2000, through July 1, 2007. Periods
with exceptionally high reserve balances, such as around September 11, 2001, have been excluded. PCE is
personal consumption expenditures.
C
The Empirical Relationship between Excess Reserve Balances and the EFFR: 2000-07
R² = 0.103
0
2
4
6
8
0.5 1.0 1.5 2.0 2.5 3.0
Excess reserve balances
(billions of PCE-weighted 2009 dollars)
Effective federal funds rate
Fitted values
(percent)
EFFR
T P-C M P I F
F R B N Y E P R , . , O
complicate the relationship between the Desk’s actions and changes in the fed funds rate, since
the EFFR can move in the absence of any intervention by the Desk. Several researchers have
identied the demand curve more precisely by estimating unexpected shocks to the supply
of reserves (see Hamilton [1997], Carpenter and Demiralp [2006b], Judson and Klee [2010],
and Ihrig, Meade, and Weinbach [2015]).
2. C M O P-C E
Just as actual shis in the demand for reserves occurred for reasons absent in the stylized
model, the day-to-day implementation of monetary policy also involved complications beyond
those discussed earlier. For example, in managing daily liquidity, the Desk had to account
for variations within a reserve maintenance period.
8
Depository institutions only needed to
maintain the required reserve balance on average over the reserve maintenance period. e
task of the Desk was to accurately forecast the supply and demand for reserve balances for
each day of the two-week maintenance period, adjusting it daily based on market conditions
and the distribution of reserves among banks. In the remainder of this section, we describe the
maintenance period structure and the Desk’s forecasting exercise.
2.1 e Reserve Maintenance Period
Reserve maintenance periods begin on a ursday and end on the second Wednesday
thereaer. Some smaller depository institutions have a weekly maintenance period. Reserve
requirements and the portion that is satised with cash holdings (vault cash) are calculated
before the start of each reserve maintenance period (known as “lagged reserve accounting”).
In order to allow depository institutions greater exibility in maintaining account balances,
the Desk averaged banks’ holdings of reserve balances over these two weeks when determining
whether banks’ reserve holdings met requirements.
9
Averaging allowed banks to eectively
manage unexpected payment shocks that would cause them to hold too few or too many reserves
C
Shifts in the Demand for Reserves
0
Effective federal funds rate
Primary credit rate
Supply
of reserves
D
1
D
2
Reserve balances
R*
Initial target rate
New target rate
F R B N Y E P R , . , O
T P-C M P I F
relative to requirements on any given day in a maintenance period. Since the exibility oered by
averaging diminishes as the number of days remaining in a maintenance period declines (until
they have no exibility on the maintenance period settlement day), banks generally tended to
hold relatively few balances early in a maintenance period in order to maximize their exibility
in absorbing payment shocks later in the period. Another feature of the reserve maintenance
period that helped smooth the volatility of the EFFR toward the end of the period was depository
institutions’ ability to carry over (subject to restrictions) excess balances from one maintenance
period to the next. is ability reduced distortions that could have resulted from the incentive to
ooad excess reserves in the last few hours of the maintenance period.
2.2 e Desk’s Forecasts and Operations
In order to ensure that rates remained responsive to changes in reserves, the Desk typically le
a “structural decit” in the banking system. In other words, the Desk le the total amount of
reserves backed by outright Treasury purchases (that is, purchases of Treasury securities in the
secondary trading markets) just below the level of aggregate reserves required by the banking
system. Maintaining a structural decit helped the Desk interact eciently with its primary
dealer counterparties. Since primary dealers are the Desk’s traditional counterparties, and
dealers are natural seekers of funding and providers of collateral, maintaining a repo book of
variable size with the dealers was more eective than maintaining a reverse repo book, because
dealers typically had limited capacity to invest funds or receive collateral.
An implication of the “structural decit” was that the Desk eectively faced a downward-
sloping demand for reserves (Chart 1). Further, the practice of not paying interest on reserves
meant that banks were highly sensitive to the opportunity cost of holding reserves—in other
words, the slope of the demand curve was relatively steep. Given its forecasts of the demand
for reserves and of changes in the supply of reserves, the Desk would fine-tune the level
of reserves by conducting daily repo operations, thereby adding reserves to or subtracting
reserves from the system. is procedure, when successfully carried out, ensured that the
EFFR remained close to the target rate on a daily basis. e aggregate reserves were then redis-
tributed within the banking system as reserve-deficit banks traded with reserve-surplus banks
in the fed funds market.
e demand for reserves had three components: required reserves, contractual clearing
balances, and “excess” reserves to meet intraday payment ows (Exhibit 1). For example,
in 2004, required reserves averaged $11 billion, contractual clearing balances averaged
$10.4 billion, and excess reserves averaged $1.6 billion (Board of Governors 2005). Banks
are required to hold reserves against transaction deposits, which are checking accounts and
other interest-bearing accounts oering unlimited checking privileges. In practice, changes in
required reserves reected changes in transaction deposits, since the Federal Reserve rarely
changed the required reserves ratio (Board of Governors 2005).
Some banks voluntarily held signicant levels of contractual clearing balances at their
Reserve Banks, in addition to their required reserve balances. Clearing balances provided
banks with increased exibility in holding reserves across the maintenance period. Banks were
compensated on their clearing balances based on three-month Treasury bill rates. However,
the income credits could only be used to defray the cost of Federal Reserve services, such as
F R B N Y E P R , . , O
T P-C M P I F
check clearing and Fedwire services, thus limiting their value (Hilton 2008). Penalties applied
if a bank had not accumulated sucient balances over a two-week maintenance period to meet
its reserve requirements and clearing balance obligations, or if it ended any day overdrawn
on its Fed account (Hilton 2008). erefore, the sum of reserve requirements and contractual
balances created a predictable level of demand for reserves.
“Excess reserves” were the amount of reserves that a bank held in excess of required
reserves and contractual balances to meet unexpected intraday payment needs that might
otherwise have created an intraday or overnight overdra on its account. e daily demand for
excess reserves was the least predictable element of the demand for reserves, since it depended
on the volume and volatility of daily payment ows (Board of Governors 2005). Average
reserve balances in 2006 were about $17.5 billion, of which excess reserves were $2.0 billion.
e total level of reserves was much smaller than daily payment ows, a disparity that had
signicant implications for banks’ ability to meet payment needs during the day, as we describe
in Section 4 and in Box 6.
Using reserve requirements along with the anticipated demand for liquidity, the Desk
forecasted the average excess reserves over a maintenance period based on the expectation
that dierent types of banks typically hold dierent levels of reserve balances. In particular,
small banks with limited access to funding markets demanded some level of excess reserves
each day—typically between $1.5 billion and $2 billion—as a cushion against liquidity shocks
(Hilton 2008). e Desk had to take this component of reserve demand into account in its daily
calculations of the reserve supply needed to maintain equilibrium in the fed funds market.
e Desk estimated total demand for reserves for the entire fourteen-day maintenance period.
For example, if the Desk observed that a bank already held more reserves than it needed to meet
its requirement for the entire maintenance period—a situation known as a “lock-in”— then the
Desk would increase its estimate for excess reserve demand for that specic maintenance period
(since the “locked-in” reserves are not available to be lent to banks with a reserve decit). In
addition, for each day of the maintenance period, the Desk estimated the demand for reserves
based, in part, on the maintenance-period-to-date distribution of reserve holdings.
E 1
The Market for Balances at the Federal Reserve before 2007
Required Reserve Balances SOMA Securities Portfolio
• Held to satisfy reserve requirements
• Do not earn interest
• Holdings of U.S. Treasury and agency mortgage-
backed securities (MBS) and repurchase agreements
Contractual Clearing Balances Discount Window Loans
• Held based on contractually agreed-upon amounts
• Generate earnings credits that defray the cost of
Federal Reserve priced services
• Credit extended to depository institutions through
the discount window
Excess Reserves Autonomous Factors
• Held to provide additional protection against over-
night overdras and reserve or clearing balance
deciencies
• Other items on the Federal Reserve’s balance sheet
such as Federal Reserve notes, Treasury’s balance at
the Federal Reserve, and Federal Reserve oat
F R B N Y E P R , . , O
T P-C M P I F
e supply of Federal Reserve balances to banks comes from three sources: the Fed’s
portfolio of securities and repurchase agreements; loans through the Fed’s discount window
facility; and liabilities on the Fed’s balance sheet that are outside the Desk’s control, known
as autonomous factors (Exhibit 1). e securities portfolio, which consisted of outright pur-
chases of securities and repurchase agreement operations, was the most important source
of reserve supply. Discount window lending was the least important, since banks rarely
borrowed from the facility. For example, no discount window loan was outstanding on the
Fed’s balance sheet on August 8, 2007, the start of the nancial crisis (Hilton 2008). Auton-
omous factors caused large daily variations in the supply of reserves. e major categories
of autonomous factors are currency in circulation, the Treasury’s balance at the Fed, foreign
central bank investments in a “repo pool,” and Federal Reserve “oat” (see Box 1 for further
discussion of the foreign repo pool and other autonomous factors).
e Desk had to forecast changes in autonomous factors extending several weeks into
the future (Board of Governors 2005) as well as the resulting impact of these changes on
reserves so that these eects could be factored into the desired size of daily OMOs. For
example, if changes in autonomous factors were forecasted to increase (reduce) reserves
by, say, $1.0 billion, then the Desk might reduce (increase) the size of its outstanding repo
operations by the same amount, all else equal. Federal Reserve notes represent the largest
autonomous factor. When the Fed issues currency to a bank, it debits the bank’s account
at the Fed, causing reserves to fall. e Treasury’s account at the Fed is the next largest
contributor to uctuations in autonomous factors. Since the Treasury is not a bank, changes
in its account balance result in corresponding changes in the supply of reserves. Treasury
balances, the foreign repo pool, and the oat are the autonomous factors most dicult to
predict on a daily basis (Hilton 2008).
Each day, the Desk compared forecasts of the supply of reserve balances from auton-
omous factors with its projections of demand for reserves and determined the need for
OMOs.
10
In addition to forecasting daily changes in autonomous factors, the Desk also
forecasted longer-term trends, such as seasonal growth in currency in circulation (for
example, demand for currency tends to increase around anksgiving and Christmas) and
the long-term growth rate of currency. If these longer-term projections indicated that the
supply of reserves was likely to be low for several weeks, then outright purchases of Treasury
securities or long-term repos might be needed.
11
Outright holdings of Treasury securities
were preferred both for operational considerations and to limit direct credit extensions to
private market participants (Hilton 2008).
In practice, the Desk generally relied on temporary OMOs to achieve the daily changes
in reserves required to keep the fed funds rate near its target. ese operations typically
involved conducting repos and reverse repos (generally of overnight duration) to increase
and decrease, respectively, the supply of reserves with primary dealers.
12
For example, in
2004 the Desk conducted 299 repo operations for about $1.9 trillion and purchased outright
$50 billion of securities (Board of Governors 2005). Using repos allowed the Desk to easily
expand or contract the level of reserves with minimal disruption to the functioning of the
market in which the underlying securities were traded. Further, repo transactions reduced
the need to make frequent temporary downward adjustments to outright holdings. (Box 2
describes how the Fed conducted repo operations when dealer inventories were low.)
F R B N Y E P R , . , O
T P-C M P I F
B
What Are Autonomous Factors?
e term “autonomous factors” refers to items on the Federal Reserve’s balance sheet that are
outside the control of the Open Market Desk of the Federal Reserve Bank of Ne w Yor k . e Desk
needs to forecast changes in autonomous factors because these changes aect the level of reserves
in the banking system. For example, when the Treasury’s account balance at the Fed increases,
reserves are eectively drained, since funds are de facto transferred from the private sector into
the Treasury’s Fed account. Conversely, when the Fed spends money—for example, on employee
salaries or remittances to the Treasury—the level of reserves in the system increases. Most daily
changes on the Fed’s balance sheet are too small to make a dierence in monetary policy imple-
mentation. However, changes in some balance-sheet categories were routinely large enough to
matter; that is, these types of balance-sheet changes routinely had a signicant impact on the
overall level of reserves and needed to be considered as the Desk developed its plans for the appro-
priate size of OMOs. We discuss the Desk’s routine forecasts of these balance-sheet categories, or
autonomous factors, below.
Currency in Circulation
Currency in circulation is typically the largest and most important autonomous factor to forecast.
When a bank places an order for currency with a Federal Reserve Bank, the latter lls the order
and debits the bank’s account at the Fed and total reserve balances decline. Currency is fungible
with reserves; a bank’s actions to withdraw (deposit) currency from its Fed account increases
(reduces) currency in circulation, thus reducing (increasing) reserves. e outstanding level of
currency in circulation varies with both seasonal and longer-term trends. Longer-term trends
include transactional demand for currency as well as foreign demand to hold U.S. dollars as a store
of value. As the demand for currency grows with the economy, reserves would decline and the fed
funds rate would rise if the Fed did not oset diminishing reserves by conducting repo operations
or by purchasing securities. e expansion of Federal Reserve notes in circulation is the primary
reason that the Fed’s holdings of securities grew over time during the pre-crisis period.
Float
Federal Reserve oat is created when credit to the account of the bank presenting a check for
payment occurs on a dierent day than debit to the account of the bank on which the check is
drawn. Float temporarily adds reserve balances when there is a delay in debiting the paying insti-
tution’s account; conversely, oat temporarily drains balances when the payer’s account is debited
before the payee receives credit. Float tended to be quite high and variable whenever the normal
check-delivery process was disrupted, such as during bad weather when travel delays could slow
down the delivery and processing of physical checks. e magnitude of dierences in the timing of
oat has decreased in recent years because more transfers are conducted electronically.
Treasury General Account (TGA) Balance
e Fed serves as scal agent for the U.S. Treasury and the TGA functions as the checking account
for the U.S. Treasury. e Treasury draws on this account to make payments by check or direct
deposit for all types of federal spending. Since the Treasury is not a bank, its payment to the public
reduces the TGA balance and increases reserve balances available to banks. Changes in the TGA
(C )
F R B N Y E P R , . , O
T P-C M P I F
balance tend to be less predictable following corporate and individual tax dates, especially in
the weeks following the April 15 deadline for federal income tax payments. Before the crisis of
2007-09, the Treasury could redirect funds from the TGA account to private banks through the
Treasury Tax and Loan program (https://www.newyorkfed.org/aboutthefed/fedpoint/fed21.html).
Doing so helped moderate the day-to-day volatility of this liability on the Fed’s balance sheet,
volatility that would complicate reserve forecasting.
Foreign Repo Pool
About 250 central banks and foreign ocial institutions have accounts with the Ne w Yor k Fed’s
Central Bank and International Account Services (CBIAS) division, which oers payment,
custody, and investment services to these accounts (see https://www.newyorkfed.org/aboutthefed/
fedpoint/fed20.html). CBIAS also oers an investment product, known as the foreign repo pool,
in which CBIAS accounts can invest overnight funds in a repo arrangement backed by SOMA
collateral. Funds that are held in CBIAS accounts at the Ne w Yor k Fed, whether in the foreign repo
pool or in transaction accounts, drain reserves from the banking system. (By denition, funds held
at the Fed reduce the supply of reserves held by the private sector).
Until the mid-1990s, the Desk sometimes directed CBIAS accounts to conduct repos with
private market participants as a means of fine-tuning the level of reserves in the system. e
Desk stopped this process in the mid-1990s by moving to a framework in which CBIAS accounts
were encouraged to keep consistent, albeit fairly low, balances in their foreign repo pool accounts.
CBIAS sta would counsel accounts to encourage stability in their holdings. Since stability
was encouraged in these accounts, the Desk would treat typical foreign repo pool balances as a
“permanent” reserve drain and the day-to-day uctuations in the foreign repo pool became a
signicant autonomous factor.
a
Before the crisis, balances in the foreign repo pool averaged around $40 billion. As of early
2018, foreign repo balances tend to be around $230 billion to $250 billion. Use of the pool has
increased over time as constraints placed on CBIAS customers have been removed (see https://
www.newyorkfed.org/newsevents/speeches/2016/pot160222). is new paradigm has increased
both the level and the variability of the foreign repo pool, but such variability no longer causes
issues with monetary policy implementation, since the Desk no longer actively manages reserve
balance levels.
CBIAS balances are published under the heading “Reverse Repurchase Agreements – Foreign
Ocial and International Accounts” in the H.4.1 Federal Reserve Statistical Release, which is
published weekly (https://www.federalreserve.gov/releases/h41/current/).
a
The Desk had another CBIAS investment product that helped control the day-to-day level of reserves. On a
voluntary basis, the Desk would sell fed funds as an agent for pooled CBIAS funds. This approach allowed the
accounts to earn a return on unexpected end-of-day balances while minimizing disruption to the supply
of reserves resulting from unexpected operational issues, such as a failure to receive delivery on the purchase
of Treasury securities.
B (C)
T P-C M P I F
F R B N Y E P R , . , O
B
How Did the Desk Avoid Conducting Undersubscribed Repo Operations?
As described in this article, the Desk generally conducted daily repo operations in order to change
the overall level of reserves in the system to match the Desk’s daily forecast of demand for these
reserves. e near-daily conduct of these operations, which typically settled on a T+0 basis (that
is, on the same day as the trade occurred), raises the question of how the Desk avoided having
undersubscribed operations, which would have resulted in supplying less reserves than intended.
is risk is not insignicant, because dealers submitting winning propositions in repo operations
must pledge unencumbered collateral to the Desk through their designated tri-party clearing agent
in order for the transaction to settle and for intended reserves to hit the banking system. What if
there isn’t much unencumbered collateral on dealers’ balance sheets?
In practice, the temporary OMOs were rarely undersubscribed, even though the typical
operation time of 9:30 a.m. came aer the time when most repo volume occurs on a daily basis.
e main reason is that the Desk’s typical take-down in short-term repo operations totaled about
$7.0 billion a day, much less than the typical overnight repo volumes that are conducted in the
private market. In addition, based on experience, the Desk could oen anticipate when collateral
shortages might develop, and it planned around them accordingly. e following table provides an
illustrative example of how the Desk arranged the tenor of repo operations to minimize the risk of
conducting undersubscribed operations ahead of the March 2006 quarter-end date.
a
From the table above, we observe that the Desk conducted two short-term repo operations that
not only provided reserves on the days they were conducted (Wednesday, March 29, and ursday,
March 30) but the tenor of these repo operations was such that they provided reserves over the
upcoming weekend, which included the quarter-end date of Friday, March 31. In this manner,
(C )
Open Market Trading Desk Repo Operations, March 2016
Date Term (Days)
Propositions Received
(Billions of U.S. Dollars)
Propositions Accepted
(Billions of U.S. Dollars)
Wednesday,
March 29, 2006
5 (spanned weekend) 37.9 5.50
ursday,
March 30, 2006
5 (spanned weekend) 22.95
5.00
ursday,
March 30, 2006
1 29.65 3.25
Friday,
March 31, 2006
3 (spanned weekend) 11.55 4.25
Source: Federal Reserve Bank of New York.
a
The table ignores the conduct of “longer-term repos,” which are discussed elsewhere in this article.
F R B N Y E P R , . , O
T P-C M P I F
2.3 Volatility of Rates during the Reserve Maintenance Period
While the reserve maintenance period allowed depository institutions greater exibility in
managing reserve balances, it also posed challenges to forecasting and interest rate control.
One concern was that reduced exibility toward the end of the maintenance period would
make the fed funds rate particularly sensitive to shocks, inhibiting the Fed’s ability to achieve
the target. is challenge is evident in the relatively high intraday standard deviation in the fed
funds market toward the end of reserve maintenance periods, consistent with the ndings of
Bartolini, Bertola, and Prati (2000) (see Chart 4). In the next section, we examine the Desk’s
ability to manage end-of-maintenance-period volatility.
3. E M P M P
O
How eective was the pre-crisis framework in meeting the primary objectives of monetary
policy? In this section, we focus on the Desk’s control of short-term rates and whether changes
in the policy rate were quickly transmitted from the fed funds market to other money markets.
First, we show that in spite of increasing intraday dispersion of the fed funds rate toward the end
of the maintenance period, the eective rate remained close to target levels. en, we demonstrate
that while the fed funds rate deviated from its target toward the end of quarters (when demand for
liquidity was high), it quickly reverted to normal levels within one day. Finally, we document that
policy rate changes were rapidly transmitted from the fed funds rate to other money market rates.
3.1 Control of the Policy Rate
Deviations from the target for the federal funds rate rarely exceeded 20 basis points, as the
le panel of Chart 5 shows. Moreover, the deviations do not appear to be persistent; instead,
larger deviations are generally followed by smaller ones. is was true even toward the end
the Desk added $10.5 billion of reserves that were outstanding over the quarter-end date. On
the quarter-end date itself, the Desk added another $4.25 billion, such that total short-term
repo operations increased reserve levels over the quarter-end date by $14.75 billion. Since the
Desk observed that dealers were more likely to be short of collateral over quarter-end dates,
this strategy enabled the Desk to successfully avoid an undersubscribed operation. Note that
total propositions submitted for the repo operation conducted on March 31, 2006, were only
$11.55 billion, suggesting that a straightforward, over-the-weekend operation with a desired
target amount of $14.55 billion would have been undersubscribed. e Desk frequently referred
to this approach as “layering in reserves.”
B (C)
F R B N Y E P R , . , O
The Pre-Crisis Monetary Policy Implementation Framework
of the maintenance period when the rates were more widely dispersed (Chart 4); indeed, the
EFFR did not dri signicantly farther from the target rate at the end of maintenance period
than it did on other days in the maintenance period. is small deviation was not the result
of banks borrowing heavily from the discount window to meet their demand for reserves. As
the right panel of Chart 5 shows, while depository institutions tended to borrow more from
the discount window on the last day of the maintenance period, the amount borrowed was
small in comparison with the amount of excess reserves. In other words, the low volatility of
fed funds during the end of maintenance periods cannot be attributed to banks smoothing
their demand for reserves through discount window borrowings. Rather, the evidence from
Chart 5 suggests that the Desk was successful in managing reserves throughout the mainte-
nance period and, in particular, the ends of maintenance periods did not signicantly impair
the Desk’s ability to implement monetary policy.
While the fed funds rate, on average, was close to its target, it could occasionally deviate
from that target. Typically, autonomous factors could experience movement that the
Desk would forecast imperfectly and the dierence would result in small supply-demand
Sources: Federal Reserve Bank of New York; authors’ calculations.
Notes: Using data on the daily standard deviation of the eective federal funds rate (EFFR) (https://apps
.newyorkfed.org/markets/autorates/fed%20funds) for the period July 3, 2000, through August 1, 2007, the chart
shows the rate’s distribution by day in the reserve maintenance period. Fifty percent indicates the
median level, and 25 percent and 75 percent indicate the 25th and 75th percentiles of the distribution,
respectively. The ends of the whiskers represent observations up to 1.5 times the interquartile range. R1/2 is the
rst/second Thursday of the maintenance period; F1/2 is the rst/second Friday of the maintenance period; M1/2
is the rst/second Monday of the maintenance period; T1/2 is the rst/second Tuesday of the maintenance period;
W1/2 is the rst/second Wednesday of the maintenance period; W2 is the settlement date.
C
Intraday Standard Deviation of Effective Federal Funds Rate during Maintenance Period
F1R1
M1
T1
W1 R2
F2 M2 T2 W2
1
0
20
30
0
75%
50%
25%
Basis points
T P-C M P I F
F R B N Y E P R , . , O
mismatches. Other deviations were generally predictable (and therefore could be anticipated
and partially oset by the Desk) and well understood. For example, large rate movements
could occur within a reserve maintenance period ahead of a widely anticipated rate change
by the FOMC; rates would typically fall on the rst Friday of each maintenance period and
typically increase on high payment ow days. More important, rates quickly reverted to the
target following such deviations.
To illustrate the resilience of the policy rates during periods of high volatility, we con-
sider the behavior of fed funds rates during quarter-ends (see Box 3 for further details).
Heightened volatility around quarter-end dates typically caused the fed funds rate to deviate
from the target. is deviation increased by an average of 6 basis points on the last day of
the quarter (day 60 in Chart 6, page 55) and by 8 basis points the following day (day 1 in
Chart 6, which is the rst day of the following quarter). By contrast, on more “typical” days
(that is, excluding the quarter-end date plus the two days before and aer it), the fed
funds rate was within a basis point of the target, on average. e fed funds rate sometimes
increased sharply at the end of months, which accounts for the spike on day 20, but volatility
on these days was not unusually high.
In order to stabilize fed funds rates around quarter-end dates, the Desk supplied extra reserves
to meet the surge in demand (see page 54 of Box 3). Moreover, the Desk planned to leave rela-
tively low levels of reserves on other days in the same reserve maintenance period. Otherwise,
the supply of reserves would have exceeded demand over the non-quarter-end days of the
Sources: Federal Reserve Bank of New York; authors’ calculations.
Notes: The left panel shows the distribution of the absolute deviation of the federal funds rate from the target
rate by day during the maintenance period, for the time frame from July 1, 2001, through August 1, 2007. The
right panel shows the distribution of discount window borrowings by day during the maintenance period, for
the time frame from January 1, 2003, through June 30, 2007. Fifty percent indicates the median level, and
25 percent and 75 percent indicate the 25th and 75th percentiles of the distribution, respectively. Ends of the
whiskers represent observations up to 1.5 times the interquartile range. R1/2 is the rst/second Thursday of
the maintenance period; F1/2 is the rst/second Friday of the maintenance period; M1/2 is the rst/second
Monday of the maintenance period; T1/2 is the rst/second Tuesday of the maintenance period; W1/2 is the
rst/second Wednesday of the maintenance period; W2 is the settlement date.
C
Absolute Deviation of Effective Federal Funds Rate from Target and Discount
Window Borrowing by Day of Maintenance Period
400
500
0
200
100
300
600
Discount Window Borrowing
Millions of dollars
F1R1 M1 T1 W1 R2 F2 M2 T2 W2
25%
50%
75%
10
15
0
5
Deviation from Target
Basis points
F1R1 M1 T1 W1 R2 F2 M2 T2 W2
25%
50%
75%
B :
Quarter-End Dynamics of Federal Funds Rates
Volatility of Fed Funds Rates at Quarter-Ends
Quarter-end volatility remained a feature of the fed funds markets in the years before the nancial
crisis. Chart 3A plots the intraday volatility of the fed funds rate for each day of the quarter,
averaged across quarters from the fourth quarter of 2004 to the second quarter of 2007. During
this time period, there was a clear trend of elevated intraday volatility on the quarter-end date.
Level of Fed Funds Rates during Quarter-Ends
Heightened volatility around quarter-end dates typically caused the fed funds rate to deviate from
the target. is deviation increased by an average of 6 basis points on the last day of the quarter
(day 60) and by 8 basis points the following day (day 1), as shown in Chart 6, page 18. By contrast,
on more “typical” days (excluding the quarter-end date plus the two days before and after it),
(C )
F R B N Y E P R , . , O
T P-C M P I F
C 3A
Volatility of Federal Funds Rate Spikes on the Last Day of the Quarter:
2004:Q4 to 2007:Q2
Source: https://apps.newyorkfed.org/markets/autorates/fed%20funds.
Notes: The chart shows, for each day t, the median of the intraday standard deviation of the
federal funds rate across quarters. Day 60 is the quarter-end date. Day 1 is the start of the quar-
ter. The quarters are standardized to sixty days by using the rst thirty days from quarter-start
and the last thirty days from quarter-end, excluding days in the middle for quarters with more
than sixty days. Rates are in basis points.
0
5
10
15
20
25
30
Basis points
5 10 15 20 25 30 35 40 45 50 55
Business days
Fed funds rate volatility
60
F R B N Y E P R , . , O
T P-C M P I F
the fed funds rate was within a basis point of the target, on average. e fed funds rate sometimes
increased sharply at the end of months, which accounts for the spike on day 20, but volatility on
these days was not unusual (as shown in Chart 3A).
Supply of Reserves during Quarter-Ends
In order to stabilize fed funds rates around quarter-end dates, the Desk supplied extra reserves to meet
the surge in demand. Moreover, the Desk planned to leave relatively low levels of reserves on other days
in the same reserve maintenance period (that is, the period over which banks’ required reserves are
calculated). Otherwise, the supply of reserves would have exceeded demand over the non-quarter-end
days of the maintenance period, pushing rates below the target once the quarter-end had passed.
e box-whisker plot of the distribution of excess reserves in Chart 3B shows that the Desk le
an average of more than $4 billion of excess reserves around quarter-end dates. In contrast, the Desk
on average le less than $0.5 billion of excess reserves on non-quarter-end days of the maintenance
period. e chart further indicates that the range of excess reserves was relatively narrow, between $3
billion and $6 billion on most quarter-end dates, suggesting that the Fed chose not to eliminate reserve
demand shocks completely, as was also found by Bartolini, Bertola, and Prati (2002).
B (C)
C 3B
Excess Reserves around Quarter-End: 2004:Q4 to 2007:Q2
12
-4
-2
0
2
4
6
8
10
14
t - 2 t - 1 t t + 1 t + 2
Billions of dollars
Source: Federal Reserve Bank of New York.
Notes: Day t (shaded) is the quarter-end date. The chart plots the distribution of excess reserves
for the ve quarter-end dates. For each date, the blue box includes values between the 25th and
75th percentiles of the distribution, with the median indicated by the orange box. The “whiskers”
indicate outliers beyond this range.
F R B N Y E P R , . , O
T P-C M P I F
maintenance period, pushing rates below the target once the quarter-end passed. Conse-
quently, the deviation of the fed funds rate from its target was short-lived, generally falling
back to the target rate on the second day aer quarter-end (Chart 6).
3.2 Transmission of the Policy Rate to Other Money Markets
e FOMC traditionally implements monetary policy by announcing a policy target rate for
the EFFR, with the expectation that its decisions will quickly be transmitted to all money
market rates. Because the Fed does not directly control market interest rates, it relies on arbi-
trage forces in money markets to transmit the change in the fed funds rate to other short-term
rates.
13
A variety of market participants can be arbitrageurs, including primary dealers that
operate in most short-term money markets and hedge funds that seek to prot from price dis-
crepancies in related markets. In this section, we examine the eectiveness of arbitrage before
the recent nancial crisis.
In the pre-crisis period, because banks active in multiple money markets could earn a
prot when money market rates were misaligned, arbitrage kept those rates aligned and
thus facilitated the transmission of monetary policy. As shown in Chart 7, the overnight
AA nancial commercial paper rate, the EFFR, and the overnight Treasury general collat-
eral (GC)
14
repo rate were highly correlated before the crisis, as would be expected with
Source: Federal Reserve Bank of New York.
Notes: The chart shows the median of the dierence between the federal funds rate and the target rate across
quarters for each day. Day 60 is quarter-end. Day 1 is the start of the quarter. The quarters are standardized
to sixty days by using the rst thirty days from quarter-start and the last thirty days from quarter-end, excluding
days in the middle for quarters with more than sixty days.
C
Federal Funds Rate Spikes around the End of Quarters: 2004:Q4 to 2007:Q2
-4
-2
0
2
4
6
8
10
Basis points
Business days
Fed funds rate minus target rate
5 10 15 20 25 30 35 40 45 50 55 60
F R B N Y E P R , . , O
T P-C M P I F
eective arbitrage.
15
Other short-term money market rates, such as Eurodollar rates (not
shown in the chart), were also tightly aligned with the EFFR.
On average, the EFFR and the Treasury repo rate generally remained close to the FOMC’s target
rate (Chart 8). Also, the repo rate was consistently below the fed funds rate, as should be expected
since repos are secured and fed funds are not. As a result, the average dierence between the EFFR
and repo rates (or the spread) was positive. Of note, the standard deviation of both rates is relatively
high compared with the respective means, in particular for the EFFR. To the extent that the vola-
tility is fundamental, the relatively high standard deviation may represent price discovery (that is,
discovery of the rate equilibrating the demand for and supply of reserves) occurring in the actively
traded fed funds market. We return to this issue in Section 4, where we discuss the advantages of
active trading in the money markets for monetary policy implementation.
In Box 4, we present a formal test of monetary policy transmission using Granger causality tests
and daily data. We show that past values of the EFFR “cause” (or predict) the current repo rate in
the pre-crisis period, a pattern one would expect if arbitrageurs operated to keep intermarket rates
aligned. In turn, the existence of arbitrage activity likely facilitated the transmission of changes in
the target rate to the repo market. e results further show that the repo rate also Granger-causes
the EFFR in the pre-crisis period, indicating two-way ows of information between the fed funds
and repo markets. is result indicates that neither market is dominant in an informational sense.
In addition to examining the fed funds market, we analyze transmission between the
Eurodollar market and the repo market. We nd that changes in the Eurodollar rate are also
transmitted to the repo rate (as might be expected, since the Eurodollar and the EFFR have
historically been tightly connected).
Sources: Federal Reserve Bank of New York; Bloomberg L.P.
Notes: All rates are of overnight tenor. GC is general collateral.
C
Overnight Money Market and Target Federal Funds Rates: July 2000–July 2007
0
200
400
600
800
2000 2001 2002
2003 2004 2005 2006 2007
Basis points
Effective federal funds rate
Treasury GC repo rate
AA-rated financial commercial paper rate
Federal funds target rate
12 px b to b
3 px
6 px b to b
T P-C M P I F
F R B N Y E P R , . , O
4. O E F M
F
While its primary task is controlling the fed funds target, a monetary policy framework can
also be evaluated with respect to objectives related to operational eectiveness and nancial
market functioning. In this section, we evaluate the pre-crisis framework’s operational
eectiveness by analyzing the eciency and transparency of the Desk’s day-to-day actions
and procedures in managing liquidity and by viewing the framework through the lens of
universality—whether the framework remains applicable in dierent states of the economy.
We assess the nancial objectives by examining the impact of the Fed’s collateral policy and
by exploring the monetary policy framework’s eect on money market and payment system
activity (Bindseil 2016).
4.1 Operational Objectives: Eciency and Transparency of Procedures
Before the crisis, the Desk’s procedures for controlling short-term interest rates were complex
and resource-intensive, resulting from the need to manage liquidity daily and, consequently,
to forecast reserve supply and demand conditions daily, a technically challenging task
(Bernanke 2005).
16
In addition, the Federal Reserve did not publish its forecasts, which made
the procedures rather opaque to market participants.
Notes: The chart shows the means and standard deviations, respectively, of the eective federal funds rate
(EFFR) and the repo rate, measured as deviations from the target federal funds rate during the pre-crisis period.
Also shown are the mean and standard deviation of the spread (that is, the EFFR minus the repo rate). GC is
general collateral.
C
Deviations from the Target Federal Funds Rate: January 2002–December 2006
-0.06
-0.04
-0.02
0
0.02
0.04
0.06
0.08
EFFR Treasury GC repo rate
minus target
EFFR minus Treasury
GC repo rate
Mean Standard deviation
minus target
Percent
F R B N Y E P R , . , O
T P-C M P I F
In a multiple-day reserve maintenance system, the daily distribution of reserves may, in
theory, be less important, since reserve requirements only have to be met by the end of the
period. But because total requirements were low relative to the daily volatility of autonomous
factors, the Desk had to evaluate reserve supply and demand conditions closely every morning.
Most days, the Desk conducted the following activities:
• forecast numerous autonomous factors over a multiday horizon;
• forecast reserve demand for multiday horizons and for dierent types of banks; and
• plan and execute the repo or reverse repo operations.
B
Testing Monetary Policy Transmission with Granger Causality Tests
To evaluate the strength of monetary policy transmission, we conduct a Granger causality test using
daily data. Past values of the eective federal funds rate “caused” (or predicted) the current repo rate
in the pre-crisis period (see table), indicating that the Fed’s monetary policy decisions were trans-
mitted to the repo market. Also, the results show that the repo rate Granger-caused the EFFR in the
pre-crisis period, showing a two-way ow of information between the fed funds and repo markets.
One concern with the analysis is that the reporting time of the data is not synchronized: the repo
rate is reported as of 9 a.m. EST, whereas the EFFR is an all-day rate. To address this issue, we
estimate the Granger causality between the one-day lagged value of the EFFR and the repo rate
and, further, between the General Collateral Finance Repo (GCF Repo
®
) Treasury rate (which is
reported at the end of the day) and the EFFR.
a
In both cases, we obtain a similar result: there is
bi-directional causality between the EFFR and the repo rate during the pre-crisis period.
We focus on the transmission from the EFFR to the repo rate because of the historical importance
of the fed funds market and the availability of a long time-series of data for the EFFR. However, in an
unreported analysis, we also nd that Eurodollar rate changes are transmitted to the repo rate (as might
be expected, since the Eurodollar and fed funds rate have historically been tightly connected).
a
Another alternative is to use a morning funds rate, such as the Broker’s Fed Funds Open. However, these
rates represent quotes and not transactions, and, moreover, they are not based on meaningful volumes.
Does the Federal Funds Rate Predict the Repo Rate in the Pre-Crisis Period?
Result
Does the fed funds rate predict the repo rate? Ye s
Does the repo rate predict the fed funds rate? Ye s
Notes: The table shows results from a Granger causality test for the period January 2002 to
December 2006. Rates are measured relative to the target federal funds rate.
F R B N Y E P R , . , O
T P-C M P I F
A description of a typical day in the life of the Desk (as is provided in Board of Gover-
nors [2005]) gives a sense of the resources required on a daily basis to conduct monetary
operations. e day would start with independent projections of the supply of and demand
for reserves by two groups of sta members, one at the Federal Reserve Bank of Ne w Yor k
and the other at the Board of Governors in Washington. en a conference call would be held
that included the manager of the System Open Market Account (SOMA) in Ne w York , sta at
the Board of Governors, and a Federal Reserve Bank president who was at the time a voting
member of the FOMC. Participants would discuss the day’s forecasts for reserves and nancial
market developments, especially in the federal funds market. Based on this information, a plan
for conducting OMOs would be formulated and a repo operation with primary dealers would
typically be executed at this time. Primary dealers would learn of the size of the repo operation
only aer it was concluded, and the desired level of reserves resulting from the operation was
never published. Longer-term repos would be arranged earlier in the morning, usually on a
specic day of the week. If an outright operation was also needed, it would be executed later in
the morning, aer the daily repo operation was completed.
Unlike at many other central banks, the Desk did not publish its forecasts (Hilton 2008),
so market participants sometimes had diculty interpreting the Desk’s actions. For example,
market participants would oen speculate that day-to-day changes in outstanding repo
operations matched the Fed’s estimate for daily changes in the demand for reserves. is
speculation was inherently awed, since it ignored the equally important impact of fore-
casted changes to autonomous factors, into which market participants had limited insight.
e Desk did not publish its daily targeted level of reserves on an ex post basis, and the
intended sizes of repo operations were not announced concurrent with the operations. Repo
market participants oen had only a vague idea of what the sizes of the repo operations
would be at the time they were announced and then had diculty interpreting the results
aer they were released.
Would an alternative framework have achieved the intended monetary policy goals with
less operational complexity? In theory, a symmetric corridor system, with the target rate in
the middle of the standing deposit and lending rates, might have required less daily interven-
tion by the Desk as long as banks were able and willing to access the central bank’s standing
facilities. is system ensures that expected rates in the maintenance period are around the
target rate, since it is equally likely that banks would be in reserve decit or surplus over the
maintenance period and, further, the costs of both outcomes are symmetric around the policy
rate (Hilton 2008; Bindseil 2014). However, the interest rate corridor in the United States was
not symmetric, since there was no standing interest-bearing deposit facility. Instead, the eec-
tive deposit rate was zero, since no interest was paid on reserves. Because banks’ opportunity
cost of holding reserves was zero, the cost of having surplus reserves was higher than the cost
of being in decit, resulting in a bias toward rates below the fed funds target (Hilton 2008).
Bindseil (2014) examines a number of alternative monetary policy frameworks with symmetric
corridors and shows that, during the pre-crisis period, they were all eective in meeting their
monetary policy objectives, suggesting that the Fed could have met its monetary policy objec-
tives using a simpler framework.
Why was the Desk able to meet its monetary policy objectives in spite of a complex and
opaque operating framework? Hilton (2008) suggests that one factor might have been the
Desk’s daily fine-tuning of the supply of reserves, whereby, when rates deviated from the
target, it responded by adjusting the daily supply of reserves to induce rate movements in the
reverse direction. is behavior may have helped to ensure that expected future rates remained
anchored around the target rate. Market participants responded appropriately to the Desk’s
fine-tuning because the Desk had built up a consistent record of success in forecasting reserve
demand and supply factors. e Desk’s forecasting ability and market condence were mutu-
ally reinforcing elements that anchored market expectations and ensured that the EFFR stayed
close to the policy target rate.
4.2 Operational Objectives: Universality
A universal (state-independent) framework is one that remains eective across dierent nan-
cial and macroeconomic conditions. All else equal, a more universal framework is desirable,
since it allows the central bank to avoid the xed costs of designing, testing, and implementing
new frameworks as conditions evolve. A more universal framework could also help avoid
unexpected shis in the operating framework if conditions change rapidly (for example,
during a crisis). Such sudden alterations to the operating framework could be suboptimal if
they are made under time constraints, as during a crisis.
One disadvantage of the pre-crisis framework vis-à-vis universality was that in order to
control the fed funds rate, the Desk needed to have control over the size of the Federal Reserve’s
balance sheet. But if the Federal Reserve needed to change the amount of reserves for a reason
other than altering the fed funds rate, the Desk could lose control of the policy rate. is lim-
itation became relevant in 2008 when the provision of large amounts of liquidity undermined
control of the interest rate, a topic we explore in the concluding remarks of this article.
A second disadvantage of the pre-crisis framework was that active daily management of
reserves, and the attendant forecasts of the conditions for reserve demand and supply, proved
to be particularly problematic during crisis periods. For example, during the early stages of the
nancial crisis (primarily in 2007 and 2008), demand for reserves proved particularly hard to
predict, resulting in large intraday swings in the federal funds rate (Hilton 2008).
4.3 Financial Market Functioning Objectives: Asset Eligibility Policy
Central banks aect market functioning through their asset eligibility framework, including col-
lateral eligibility for discount window borrowing. Assets eligible for use as collateral may benet
from increased liquidity and enhanced ability to obtain credit, compared with ineligible assets.
17
Further, to the extent that “haircuts” do not fully reect risks—for example, if they do not vary
by counterparty—the price of eligible assets might be distorted (a “haircut” being the dierence
between the market value of the asset pledged as collateral and the amount of the loan). ese
market eects are likely to be higher, the broader the set of collateral assets. For example, if the
central bank accepts a wide range of collateral assets, then banks may have an incentive to struc-
ture their balance sheets to maximize access to central bank credit (Bindseil 2014).
18
e Federal Reserve Act limits the types of assets that the Federal Reserve may acquire
through open market operations. In practice, the Fed accepts only high-quality assets in its
F R B N Y E P R , . , O
T P-C M P I F
F R B N Y E P R , . , O
T P-C M P I F
OMOs—namely, Treasury debt and debt issued or fully guaranteed by U.S. federal agencies,
which includes agency mortgage-backed securities. A wider variety of assets, including
government and private-sector securities, mortgages, and consumer and commercial loans,
are eligible to be pledged against discount window loans. Since discount window borrowings
were negligible in normal times, the eligibility criteria for OMOs were the binding con-
straints. e strictness of the OMO eligibility criteria likely reduced the distortionary eects
on asset prices, since the additional liquidity benets of being granted eligibility are likely
small for these types of assets.
19
An alternative view is that the central bank should actively use its collateral policy to
support an important asset market that is currently illiquid. Indeed, in the 1920s and 1930s,
the Fed took an active role in enhancing the liquidity of the U.S. Treasury bond markets, in
part by including these securities as collateral for its nascent OMOs (Garbade 2012). Later,
the U.S. Treasury bond markets developed into some of the most liquid asset markets, and
so the Fed no longer needed to actively support these markets through its collateral policy.
Under this view, inclusion of a broader range of assets for collateral eligibility, even if that
involves including illiquid assets, may be desirable.
4.4 Financial Market Functioning Objectives: Money Market Activity
A framework based on reserve scarcity is likely to encourage higher interbank trading
activity than one with reserve abundance. Indeed, the scarcity of reserve balances relative to
required and precautionary demand for reserves that was a feature of the pre-crisis frame-
work resulted in large volumes of trading between banks, since banks with more reserves
than necessary would have an incentive to trade with banks that had too few reserves. For
example, in the fourth quarter of 2006, brokered fed funds activity averaged $95 billion
per day. In contrast, under an abundant-reserve regime in the fourth quarter of 2015, the
volume of brokered fed funds averaged only $42 billion per day. (See Box 5 for a more
detailed discussion of changes in fed funds market activity since the crisis).
As a general matter, it is unclear whether supporting active money markets should be a
goal of a monetary policy framework. Active money markets may promote the transmission
of changes in policy rates to the broader market by facilitating arbitrage, enabling price dis-
covery, and promoting market discipline. However, alternative markets (such as short-term
funding markets) may be available for providing these benets. e potential signaling
benets from money markets are also hard to quantify. Changes in trading volumes may not
be driven by fundamentals but rather by idiosyncratic payment shocks. Further, participants’
eorts to monitor the credit quality of counterparties vary considerably, and it may be di-
cult to value the social good that results from such monitoring, given that contagious credit
and liquidity shocks may force lenders to withdraw funding broadly.
20
In the particular circumstances of the pre-crisis period, however, activity in the fed funds
market likely provided some benets. An active fed funds market probably promoted rate
discovery, which facilitated the quick transmission of changes in the EFFR to other money
market rates (see Section 3).
F R B N Y E P R , . , O
T P-C M P I F
B
Federal Funds Market Activity before and after the Crisis
e pre-crisis period was characterized by signicant interbank trading. Banks would trade fed
funds for a variety of reasons, including avoiding overnight overdras, smoothing daily balances
emanating from day-to-day uctuations in both assets and liabilities, and meeting reserve require-
ments over the two-week reserve maintenance period. In addition, because the yield curve was
typically upward-sloping, some banks established a “structural short” position wherein they would
eectively fund longer-term assets through consistent borrowing in the fed funds market.
Along with the shi to abundant reserves following the crisis of 2007-09, fed funds trading
volume declined sharply. is decrease is evident in the chart below, which shows a roughly 50
percent drop in the volume of brokered fed funds aer the crisis.
Activity in the fed funds market is currently dominated by investors that were ineligible to receive
interest on excess reserves (IOER) interacting with mostly foreign banking organizations that gen-
erally leave the borrowed proceeds at the Fed to earn IOER, in a trade known as IOER arbitrage. As
a result, the fed funds market now diers fundamentally from what it was pre-crisis. Most, if not all,
of the pre-crisis motivations for borrowing and selling fed funds have changed signicantly, and new
Basel III regulations discourage banks from funding longer-term assets with short-term liabilities. As
a consequence, fed funds trading volumes are now persistently lower than they were pre-crisis.
Brokered Federal Funds Rate Volume: October 2006–February 2016
Source: Federal Reserve Bank of New York.
Note: The data are the result of aggregating daily total volumes voluntarily supplied by federal
funds brokers.
0
50
150
200
2007 2008 2009 2010 2011 2012 2013 2014 2015
100
Billions of dollars
F R B N Y E P R , . , O
T P-C M P I F
4.5 Financial Market Functioning Objectives: Payment System Activity
In the pre-crisis period, banks relied on substantial provisions of intraday, or daylight
overdra, credit from the Fed, because the level of reserves was insucient to cover the
clearing needs of the payment system. Since the Fed charged low fees for intraday credit,
banks relied on the Fed as a source of intraday funding and did not necessarily borrow in
the wholesale funding markets to address shortfalls. With average daily Fedwire volume of
$2.28 trillion and average reserve balances of just $9 billion held at Federal Reserve Banks
in 2006, large daylight overdras were a likely consequence of the low levels of reserves in
the system. (Box 6, page 64, discusses to what extent private solutions to the problem of
intraday credit existed.)
Peak daylight overdras (the largest total amount of credit outstanding at any time)
as well as average overdras by maintenance period over the pre-crisis period are shown
in the right panel of Chart 9 (above). In 2006, the use of intraday overdras averaged
roughly $51 billion during operating hours, and, on average, $140 billion was outstanding
at peak use over a maintenance period (roughly 6 percent of average payment volume over
Fedwire). Peak overdra use steadily increased in the pre-crisis period starting in 2000.
While daylight overdras facilitate payments, they also expose the Fed to the potential
for loss, should the institutions incurring negative balances fail to replenish their funds.
(See Box 6 for more discussion of the evolution of the Fed’s Payment System Risk policy).
C
Time Series of Fedwire Volume and Daylight Overdrafts by Maintenance Period
1,000
1,400
1,800
2,200
2,600
3,000
2000 2001 2002 2004 2005 2006 2008
Billions of dollars Billions of dollars
Average Fedwire Volume
by Maintenance Period
0
40
80
120
160
200
2000 2002 2004 2006
Average Daylight Overdrafts
by Maintenance Period
Peak daylight overdrafts
Average daylight overdrafts
2001 2003 2005 2007
Sources: Federal Reserve Board; Federal Reserve Bank of New York; authors' calculations.
Note: Both panels reect data for the period from January 1, 2000, through August 1, 2007.
F R B N Y E P R , . , O
T P-C M P I F
B
Changes to the Payment System Risk Policy Regarding Overdrafts
The Federal Reserve’s Payment System Risk Policy
e extension of intraday credit by Federal Reserve Bans is governed by the Fed’s Payment
System is (PS) policy. e policy addresses the riss that payment, clearing, and settlement
activities present to the Fed and the nancial system as a whole, and it also governs eserve
Bans’ provision of intraday credit to accountholders. (For more information, see https://www
.federalreserve.gov/paymentsystems/psr_about.htm.)
e policy was rst written in 1985 and has been amended multiple times since then. It was
modied in 1994 to charge bans fees for their use of intraday credit. In 2001, changes to the
PS policy allowed institutions meeting certain criteria to obtain collateralized overdras above
their authorized capacity, nown as net debit caps. In 2006, the policy was revised to require
government-sponsored enterprises (GSEs) and international organizations to pre-fund Fedwire
payments of principal and interest due on their outstanding debt, thus precluding the Fed
from granting intraday credit to these institutions. In 2008, the policy was once again revised
by setting the fee for collateralized overdras at zero and raising the fee for uncollateralized
overdras to 50 basis points. is policy change was intended to improve the eciency of
the payments system while also limiting the credit exposure of Federal Reserve Banks. ese
changes went into eect in 2011.
Changes in Peak Intraday Overdraft Activity
e chart on page 65 presents peak intraday overdra activity by quarter since 1986. We observe
a temporary slowdown in the growth of intraday credit around the time of the implementation of
the GSE restriction. However, it is dicult to attribute causation, since other factors may have been
responsible for the slowdown. Peak intraday credit resumed growing later in 2006 and ultimately
crested sometime in 2008. (See https://www.federalreserve.gov/paymentsystems/psr_data.htm.)
e most striking feature of the chart is the dramatic decline in the use of intraday overdra
credit once the Fed’s balance sheet expanded in the fourth quarter of 2008. Excess reserve balances
grew dramatically during this time, rst from draws on various liquidity facilities and then from the
Fed’s large-scale asset purchase programs. As excess reserve levels have remained high, demand from
banks to borrow funds from the Fed on an intraday basis has remained low for many banks.
Private Solutions to the Need for Intraday Overdrafts
During the period of high intraday credit use, no private market for providing this specic type
of credit materialized. However, there were some informal market-oriented solutions to provide
credit on an intraday basis. One solution was for large GSEs to obtain a collateralized intraday
credit line from a large money center bank. e other solution for borrowers that were facing
large intraday credit needs was to borrow overnight money in the brokered fed funds market on
a “ll or kill” basis. With a ll or kill order, the lender agrees to send the funds promptly, say, in a
een-minute window, and the borrower returns the funds within a twenty-three-hour window
by market convention. e borrower is willing to pay a little more to borrow funds in this manner
and thereby receives immediate relief on intraday credit issues by receiving the funds promptly.
is solution was widely used by market participants in the pre-crisis period.
(C )
F R B N Y E P R , . , O
T P-C M P I F
Peak Daylight Overdrafts
0
20
40
60
80
100
120
140
160
180
200
201820142010200620021998199419901986
Billions of dollars
Source: Federal Reserve Board.
Note: Amounts are based on a 21.5-hour Fedwire day.
B 6 (C)
5. C R
Control over the federal funds rate represented the primary policy tool of the Fed’s pre-crisis
operating framework. In order to exert this control, the Fed relied on a scarcity of reserves to
ensure that the fed funds rate would be sensitive to the level of reserves in the system. To adjust
the rate, the Desk forecasted the demand for and supply of reserves daily, and then typically
increased or decreased the amount of reserves available to banks relative to forecasted demand
by changing the size of daily repo operations. e aggregate amount of reserves was distributed
among banks by means of trading in the fed funds market. Arbitrageurs aided the framework
by transmitting changes in the fed funds rate to other short-term interest rates and to the real
economy more broadly. In this article, we discussed the desirability of the pre-crisis framework
in the context of meeting monetary policy objectives while conducting monetary operations in
an ecient and transparent manner, such that nancial market functioning was not impaired.
e pre-crisis framework was eective at meeting monetary policy objectives. First, the
Desk successfully maintained the EFFR close to the target rate set by the FOMC, even during
F R B N Y E P R , . , O
T P-C M P I F
periods of signicant volatility in banks’ demand for reserves. Furthermore, deviations were
not persistent—the Desk generally corrected any short-term movements in the fed funds rate.
Finally, changes in the target rate were quickly transmitted to other money market rates.
e pre-crisis framework receives mixed reviews in terms of unimpaired nancial market
functioning. Having a relatively restricted set of collateral eligible for open market operations
ensured that the Desk’s operations did not signicantly aect the relative pricing of risk. e
scarcity-of-reserves paradigm also ensured relatively active trading in the interbank market.
However, this same paradigm placed strains on the interbank payment system, leading to
heavy use of daylight overdra credit from the Fed.
e pre-crisis operational framework scores less well in terms of meeting monetary
objectives eciently and transparently. Implementing the operating procedures was complex
and the procedures were opaque, which meant that market participants generally found them
dicult to understand. In addition, the framework lacked universality in that large changes in
aggregate reserve balances could undermine the Desk’s ability to control the policy rate. is
critique became relevant in 2008 when the Federal Reserve implemented emergency lending
programs to combat the eects of the nancial crisis. ese programs expanded the aggregate
amount of reserve balances for reasons other than monetary policy, causing the Desk to
lose control over the policy rate. Finally, the daily operational procedures did not adapt well
to crisis market conditions as reserves demand became dicult to predict, resulting in high
intraday volatility.
In order to regain control of the policy rate, the Fed abandoned the pre-crisis framework
in favor of a framework that would allow the Desk to continue to carry out FOMC objectives
regardless of the amount of reserves in the banking system. Unlike the pre-crisis framework,
the current monetary policy framework is one of reserve abundance, whereby, through the use
of administered rates (those rates set by the Fed and not determined in the markets), the fed
funds rate is kept within a range set by the FOMC. Using this new framework, the Desk has
continued to maintain the policy rate within the target range set by the FOMC. is success
demonstrates that while the pre-crisis framework oered eective monetary control, this
control was not unique to that paradigm.
e abundant-reserves framework has resulted in changes to other aspects of the pre-crisis
framework. With the abandonment of the scarcity-of-reserves approach, banks no longer
rely heavily on overdra credit from the Federal Reserve. As a further by-product, the
abundant-reserves framework has diminished banks’ need to transact in the fed funds market,
causing a reduction in volume (see Box 5). However, the benets of active money markets are
debatable and must be weighed against reduced volatility of the EFFR (Potter 2016).
While the current framework will likely evolve as the FOMC considers its appropriateness
in meeting future monetary policy challenges, a return to the pre-crisis framework is not
necessarily desirable. As we have shown in this article, the pre-crisis framework had several
shortcomings that can probably be addressed.
T P-C M P I F
N
1
In addition, the mandate requires the Fed to maintain moderate levels of the long-term interest rate (Board of
Governors of the Federal Reserve System 1994).
2
In this article, we use “bank” to mean “depository institution.” Technically, they are not equivalent, since some
nonbank intermediaries, such as credit unions or savings and loan associations, can also take deposits.
3
There is no consensus in the literature regarding the appropriate goals of a monetary policy framework. Our explication
is loosely based on Bindseil (2014), who also discusses additional objectives that we do not consider. For example,
Bindseil (2014) includes adequate risk-adjusted financial returns on central bank assets as part of financial objectives.
4
In a symmetric corridor system, the target rate lies in the middle of the rates for the standing lending and borrowing
facilities. The benefits of such a system for liquidity management are discussed in Section 4.
5
Since the market fed funds rate varied from trade to trade depending on the creditworthiness of borrowers and
other factors, the Fed used a weighted average of market rates as its policy target. Prior to March 1, 2016, the EFFR
was calculated as a weighted average based on fed funds transactions as reported to the Desk by fed funds brokers.
Effective March 1, 2016, the EFFR calculation was changed from a weighted average mean to a volume-weighted
median and the source data were changed to the FR 2420, Report of Selected Money Market Rates. The EFFR
is published by the Desk on the morning of the business day following the day of the report.
6
Under Regulation A, Extensions of Credit by Federal Reserve Banks, as of January 9, 2003, financially strong
and well-capitalized banks can borrow under the Fed’s primary credit program at a penalty rate above the target fed
funds rate (rather than at a subsidized rate, as was the case prior to this regulation).
7
In reality, the stigma associated with borrowing from the Fed deters banks from using the facility, resulting in
some borrowing at market rates in excess of the primary credit rate (Armantier et al. 2015; Furne 2001). Prior to
2003, discount window borrowers had to satisfy the Fed that they had exhausted private sources of funding and that
they had a genuine business need for the funds, which likely contributed to the stigma. Since 2003, the discount
window has been a “no questions asked” facility but the stigma has continued to exist.
8
Theoretical models that incorporate a reserve maintenance period include Gaspar and Rodrigues-Mendizabal (2004)
and Ennis and Keister (2008).
9
In practice, the Desk managed reserve levels to meet required operating balances, which were equal to
reserve requirements plus contractual clearing balances. These were the amounts that some banks voluntarily
held at their Reserve Banks to defray the cost of Federal Reserve services (described in more detail in Section 2.2).
For simplicity, we refer to “required operating balances” as “reserve requirements” for the remainder of this article.
10
The Desk also forecast the supply of reserves from discount window lending, but, as previously mentioned,
these amounts were generally small. See Meulendyke (1998) for further discussion of the Desk’s daily operations.
11
The Desk may also sell Treasury securities outright, but such transactions are extremely rare. In addition, the Fed
may transact with foreign officials and international customers at market prices to make small adjustments to
its portfolio without entering the market (see Box 1).
12
Certain broker-dealers are designated primary dealers. These institutions must meet certain standards and serve
as trading counterparties to the Federal Reserve Bank of New Yor k in carrying out monetary policy. They also
participate in auctions of government securities and make markets for these instruments.
13
See Bernanke (2005).
14
General collateral Treasury securities are those that have no special features, such as unusually high market demand.
15
The repo rate is the Desk’s 9 a.m. Primary Dealer Survey Treasury GC Repo rate. Prior to March 1, 2016, the EFFR
was calculated by the Desk from broker submissions. Since then, the EFFR has been calculated based on borrowing
data submitted by banks in FR 2420, Report of Selected Money Market Rates.
16
Indeed, it could be challenging to explain the daily monetary operations even to experts, as implied by the
Board’s own description of the open market policy process (Board of Governors 1963).
F R B N Y E P R , . , O
F R B N Y E P R , . , O
T P-C M P I F
N (C)
17
See the European Central Bank Monthly Bulletin (2007).
18
It is possible that, with a wider range of assets, more asset prices are distorted but there is less distortion for each asset.
19
The ECB, in contrast, accepts a broad range of illiquid collateral but, to avoid distorting prices, uses objective
and publicly available criteria in its asset selection and ensures that assets with similar properties are treated in
a similar manner. See the European Central Bank Monthly Bulletin (2007).
20
See Potter (2016) for a discussion of these issues.
F R B N Y E P R , . , O
T P-C M P I F
R
Armantier, O., E. Ghysels, A. Sarkar, and J. Shrader. 2015. “Stigma in Financial Markets: Evidence from
Liquidity Auctions and Discount Window Borrowing during the Crisis.” J F
E , no. 2: 317-35.
Bartolini, L., G. Bertola, and A. Prati. 2002. “Day-to-Day Monetary Policy and the Volatility of the Fed
Funds Rate.” J M, C, B , no. 1: 137-59.
Bech, M. L., and E. Klee. 2011. “e Mechanics of a Graceful Exit: Interest on Reserves and Segmentation
in the Fed Funds Market.” J M E , no. 5 (July): 415-31.
Bernanke, B. 2005. “Implementing Monetary Policy.” Remarks presented at the Redening Investment
Strategy Education Conference, University of Dayton, Dayton, Ohio, March 30.
Bindseil, U. 2004. M P I. Oxford: Oxford University Press.
Bindseil, U. 2014. M P O F S. Oxford: Oxford
University Press.
Bindseil, U. 2016. “Evaluating Monetary Policy Operational Frameworks.” European Central Bank
working paper, August.
Board of Governors of the Federal Reserve System. 1963. “e Federal Reserve System: Purposes
and Functions.”
———. 1994. “e Federal Reserve System: Purposes and Functions.”
———. 2005. “e Federal Reserve System: Purposes and Functions.”
Carpenter, S., and S. Demiralp. 2006a. “Anticipation of Monetary Policy and Open Market Operations.”
I J C B 2.
Carpenter, S., and S. Demiralp. 2006b. “e Liquidity Eect in the Fed Funds Market: Evidence from
Daily Open Market Operations.” J M, C, B : 901-20.
European Central Bank. 2007. “e Collateral Frameworks of the Federal Reserve System, the Bank of
Japan, and the Eurosystem.” European Central Bank M B (October).
Ennis, H. M., and T. Keister. 2008. “Understanding Monetary Policy Implementations.” Federal Reserve
Bank of Richmond E Q 94, no. 3 (Summer): 235-63.
Furne, C. 2001. “e Reluctance to Borrow from the Fed.” E L 72, no. 2: 209-13.
Garbade, K. D. 2012. B M. Cambridge, Mass.: The MIT Press.
F R B N Y E P R , . , O
T P-C M P I F
R (C)
Gaspar, V., G. Perez-Quiros, and H. Rodriguez-Mendizabal. 2004. “Interest Rate Determination in the
Interbank Money Market.” ECB Working Paper Series, no. 351.
Hamilton, J. 1997. “Measuring the Liquidity Eect.” A E R , no. 1: 80-97.
Hilton, S. 2008. “Recent Developments in Federal Reserve System Liquidity and Reserve Operations.”
Reserve Bank of Australia Conference“Lessons from the Financial Turmoil of 2007 and 2008,”
Sydney, Australia.
Ihrig, J. E., E. E. Meade, and G. C. Weinbach. 2015. “Monetary Policy 101: A Primer on the Fed’s Changing
Approach to Policy Implementation.” Finance and Economic Discussion Series 2015-047. Washington:
Board of Governors of the Federal Reserve System, http://dx.doi.org/10.17016/FEDS.2015.047.
Judson, R., and E. Klee. 2010. “Whither the Liquidity Eect: e Impact of Federal Reserve Open Market
Operations in Recent Years.” J M , no. 3: 713-31.
Keister, T., A. Martin, and J. McAndrews. 2008. “Divorcing Money from Monetary Policy.” Federal
Reserve Bank of Ne w York E P R , no. 2 (September): 41-56.
Logan, L. K. 2017. “Implementing Monetary Policy: Perspectives from the Open Market Desk.” Remarks
before the Money Marketeers of Ne w Yo rk University, Ne w York City.
Meulendyke, A-M. 1998. U.S. M P F M. Monograph, Federal
Reserve Bank of Ne w York .
Potter, S. 2016. “Discussion of ‘Evaluating Monetary Policy Operational Frameworks’ by Ulrich Bindseil.”
Remarks at the 2016 Economic Policy Symposium at Jackson Hole, Wyoming.
e views expressed are those of the authors and do not necessarily reect the position of
the Federal Reserve Bank of New York or the Federal Reserve System. e Federal Reserve
Bank of New York provides no warranty, express or implied, as to the accuracy, timeliness,
completeness, merchantability, or tness for any particular purpose of any information
contained in documents produced and provided by the Federal Reserve Bank of New York
in any form or manner whatsoever.
