View original document

The full text on this page is automatically extracted from the file linked above and may contain errors and inconsistencies.

Finance and Economics Discussion Series
Division of Research and Statistics
Division of Monetary Affairs
Federal Reserve Board, Washington, D.C.

Working Studies 1: Part 1

OPERATING PROCEDURES AND THE CONDUCT OF
MONETARY POLICY: CONFERENCE PROCEEDINGS
Special issue Editors:
Marvin Goodfriend and David H. Small
March 1993
NOTE: Working Studies are collections of staff studies organized around specific themes and issued as occasional
supplements to the Finance and Economics Discussion Series. Working Studies are preliminary materials circulated
to stimulate discussion and critical comment. The analysis and conclusions set forth are those of the authors and do
not indicate concurrence by other members of the research staffs, by the Board of Governors, or by the Federal
Reserve Banks. Upon request, single copies of the paper will be provided. References in publications to the Finance
and Economics Discussion Series (other than acknowledgement by a writer that he has access to such unpublished
material) should be cleared with the author to protect the tentative character of these papers.

Operating Procedures and the Conduct of Monetary Policy:
Conference Proceedings

ABSTRACT
The Federal Reserve System, through its Committee on Financial
Analysis, sponsored a conference on monetary policy operating
procedures and strategies which was hosted by the Federal Reserve Bank
of St. Louis on June 18-19, 1992. An update on the academic and
System staff views of such issues seemed desirable because it had
been a decade since the last broad examination of operating
procedures, which appeared in "New Monetary Control Procedures".
Since that time, the Federal Reserve has shifted from reserve-based
operating procedures tied closely to movements in transactions money
to discretionary changes in reserve conditions keyed to a variety of
indicators of developments in broad money and credit, financial
markets more generally, and the economy. Meanwhile, the operating
procedures of foreign central banks have also been evolving, perhaps
with implications and lessons for the Federal Reserve. The conference
was intended to review these developments and relate them to
achievement of longer-term objectives for the United States economy,
and to stimulate further thinking and research on topics related to
the design and execution of monetary policy.
To these ends, System economists prepared and presented the
papers in this volume, which were reviewed by discussants. Professor
Bennett McCallum of Carnegie-Mellon University and Professor John
Taylor of Stanford University were invited to discuss specific papers
and provide overviews of the entire conference proceedings. Their
overviews appear at the end of this volume.
The conference was organized by Al Broaddus (then Director of
Research and currently President) and Marvin Goodfriend of the Federal
Reserve Bank of Richmond and by David Lindsey (Deputy Director) and
David Small of the Division of Monetary Affairs of the Board of
Governors of the Federal Reserve System.

pr7'*»*jmf<*.

/7**L

Donald L. Kohn
Director
Division of Monetary Affairs
Board of Governors of the Federal Reserve System
April 2, 1993

1. Board of Governors, 1981. A more narrowly focused study was
conducted by the Federal Reserve Bank of New York in "Intermediate
Targets and Indicators for Monetary Policy: A Critical Survey", 1990,




TABLE OF CONTENTS
VOLUME 1

Session 1. Historical Overview
Ann-Marie Meulendyke
"Federal Reserve Tools in the Monetary Policy Process in Recent
Decades"
Comments: Robert L. Hetzel
Marvin Goodfriend
"Interest Rate Policy and the Inflation Scare Problem: 1979-1992"
Comments: R. Alton Gilbert

Session 2. International Comparisons
John Morton and Paul Wood
"Interest Rate Operating Procedures of Foreign Central Banks"
Bruce Kasman
"A Comparison of Monetary Policy Operating Procedures in Six
Industrial Countries"
Comments (on both John Morton and Paul Wood and on Bruce Kasman)
Stephen A. Meyer
Robert B. Kahn and Linda S. Kole
"Monetary Transmission Channels in Major Foreign Industrial
Countries"
Comments: Craig S. Hakkio

Session 3: Time Series Econometric Issues
William Roberds, David Runkle, and Charles H. Whiteman
"Another Hole in the Ozone Layer: Changes in FOMC Operating
Procedure and the Term Structure
Comments: Glenn D. Rudebusch
Charles Evans, Steven Strongin, and Francesca Eugeni
"A Policymaker's Guide to Indicators of Economic Activity"
Comments: Richard W. Kopcke




-2-

Session 4: Operating Issues Related to Banking
Wilbur John Coleman II, Christian Gilles, and Pamela Labadie
"Discount Window Borrowing and Liquidity"
Comments: Michael Dotsey
Kenneth N. Kuttner
"Credit Conditions and External Finance: Interpreting the Behavior
of Financial Flows and Interest Rate Spreads"
Comments: David Wilcox

VOLUME 2
Session 5: Reserve Targeting, Interest Rate Targeting, and Term
Structure Volatility
Joseph E. Gagnon and Ralph W. Tryon
"Price and Output Stability Under Alternative Monetary Policy
Rules"
Comments: Satyajit Chatterjee
Steven Russell
"Monetary Policy Experiments in a Stochastic Overlapping
Generations Model of the Term Structure"
Comments: Eric M. Leeper
Jeffrey Fuhrer and George Moore
"Inflation Persistence"
Comments: John B. Taylor
Session 6: Adapting to Regulatory Change
Allan D. Brunner and Cara S. Lown
"Implementing Short-Run Monetary Policy with Lower Reserve
Requirements"
Comments: Edward J. Stevens
John Wenninger and William Lee
"Federal Reserve Operating Procedures and Institutional Change"
Comments: Daniel L. Thornton




-3-

Session 7: Feedback Rules for Monetary Policy
John P. Judd and Brian Motley
"Controlling Inflation with an Interest Rate Instrument"
Comments: Evan F. Koenig
Gregory D. Hess, David H. Small, and Flint Brayton
"Nominal Income Targeting with the Monetary Base as Instrument: An
Evaluation of McCallum's Rule"
(with an appendix by Richard D. Porter)
Comments: Bennett T. McCallum

Summary and Overview
John B. Taylor
"New Directions in Monetary Policy Research: Comments on the
Federal Reserve System's Special Meeting on Operating Procedures"
Bennett T. McCallum
"Concluding Observations"




LIST OF CONTRIBUTORS
CONTRIBUTORS

AFFILIATION

Brayton, Flint

Board of Governors

Broaddus, Al

Federal Reserve Bank of Richmond

Brunner, Allan D.

Board of Governors

Chatterjee, Satyajit

Federal Reserve Bank of Philadelphia

Coleman, Wilbur John, II

Board of Governors

Dotsey, Michael

Federal Reserve Bank of Richmond

Eugeni, Francesca

Federal Reserve Bank of Chicago

Evans, Charles

Federal Reserve Bank of Chicago

Fuhrer, Jeffrey

Federal Reserve Bank of Boston

Gagnon, Joseph E.

Board of Governors

Gilbert, R. Alton

Federal Reserve Bank on St. Louis

Gilles, Christian

Board of Governors

Goodfriend, Marvin

Federal Reserve Bank of Richmond

Hakkio, Craig S.

Federal Reserve Bank of Kansas City

Hess, Gregory D.

Board of Governors

Hetzel, Robert L.

Federal Reserve Bank of Richmond

Judd, John P.

Federal Reserve Bank of San Francisco

Kahn, Robert B.

Board of Governors

Kasman, Bruce

Federal Reserve Bank of New York

Koenig, Evan F.

Federal Reserve Bank of Dallas

Kohn, Donald L.

Board of Governors

Kole, Linda S.

Board of Governors

Kopcke, Richard W.

Federal Reserve Bank of Boston

Kuttner, Kenneth N.

Federal Reserve Bank of Chicago




Labadie, Pamela

Board of Governors

Lee, William

Federal Reserve Bank of New York

Leeper, Eric M.

Federal Reserve Bank of Atlanta

Lindsey, David E.

Board of Governors

Lown, Cara S.

Federal Reserve Bank of New York

McCallum, Bennett T.

Carnegie-Mellon University

Meulendyke, Ann-Marie

Federal Reserve Bank of New York

Meyer, Stephen A.

Federal Reserve Bank of Philadelphia

Moore, George

Board of Governors

Morton, John

Board of Governors

Motley, Brian

Federal Reserve Bank of San Francisco

Porter, Richard D.

Board of Governors

Roberds, William

Federal Reserve Bank of Atlanta

Rudebusch, Glenn D.

Board of Governors

Runkle, David

Federal Reserve Bank of Minneapolis

Russell, Steven

Federal Reserve Bank of St. Louis

Small, David H.

Board of Governors

Stevens, Edward J.

Federal Reserve Bank of Cleveland

Strongin, Steven

Federal Reserve Bank of Chicago

Taylor, John B.

Stanford University

Thornton, Daniel L.

Federal Reserve Bank of St. Louis

Tryon, Ralph W.

Board of Governors

Wenninger, John

Federal Reserve Bank of New York

Whiteman, Charles H.

University of Iowa

Wilcox, David

Board of Governors

Wood, Paul

Board of Governors




FEDERAL RESERVE TOOLS IN THE MONETARY POLICY PROCESS
IN RECENT DECADES
Ann-Marie Meulendyke1
Most students of money and banking in the United States would identify
open market operations, reserve requirements, and the discount rate as
the basic tools of monetary policy.

They would add that open market

operations are the primary, most actively employed tool because of their
flexibility and ease of use. The historical roles of open market
operations in the conduct of monetary policy under the guidelines
established by the Federal Open Market Committee (FOMC) were examined in
some detail in an earlier article by the author.2

This article

provides parallel treatment for reserve requirements and the discount
window.

Both articles focus on the years since the 1951 Treasury-

Federal Reserve Accord, an agreement that freed the Federal Reserve from
the obligation to peg interest rates on U.S. Treasury debt and enabled
it to resume an independent monetary policy.
Before beginning the review of reserve requirements and the
discount window, it may be helpful to summarize the main findings on
open market operations.

Since the Accord, the FOMC has used various

money and credit measures, as well as assessments of the underlying
economic and price picture, as intermediate objectives to guide the
settings of its operating instruments.

Reserve measures and interest

rates have alternated as the FOMC's primary guide for day-to-day
operations.
In the first two decades after the Accord, the Trading Desk at
the New York Federal Reserve Bank carried out the FOMC's instructions

1. Manager and Senior Economist, Open Market Department, Federal
Reserve Bank of New York. Ted Tulpan provided excellent research
assistance. The author wishes to thank Peter Sternlight, Betsy White,
John Wenninger, Bruce Kasman, and Spence Hilton of the New York Federal
Reserve and Robert Hetzel of the Richmond Federal Reserve for helpful
comments on an earlier draft.
2. Ann-Marie Meulendyke, "A Review of Federal Reserve Policy
Targets and Operating Guides in Recent Decades," Intermediate
Targets
and Indicators
for Monetary Policy: A Critical
Survey, Federal Reserve
Bank of New York, July 1990. Reprinted from Federal Reserve Bank of New
York Quarterly Review, vol. 13, no. 3 (Autumn 1988), pp. 6-17.




Meulendyke

for achieving the desired average behavior of various measures of bank
credit.

Operating decisions were keyed to free reserves--reserves in

excess of those needed to meet reserve requirements less reserves
borrowed at the discount window--and to the tone and feel of the money
markets.

By the 1970s, the monetary aggregates had replaced credit

measures as intermediate targets and the day-to-day emphasis shifted
toward controlling the overnight interbank rate, called the federal
funds rate.
During the 1970s, adjustments to the federal funds rate were
generally small, and at times there was a reluctance to make necessary
increases in the rate.

Partly as a result, money growth persistently

exceeded its targets, and inflationary pressures reached clearly
unacceptable levels by the latter part of the decade.
changed its approach to policy.

In 1979, the FOMC

Under the new procedures, it targeted

levels of nonborrowed reserves, a measure that was closely linked
through reserve requirement ratios to desired growth rates of a narrowly
defined measure of money, Ml.

In addition, it allowed the federal funds

rate to move over a much wider range than before to increase the
likelihood that money growth would be brought under control.

Although

these procedures contributed to increased fluctuations in both money and
interest rates, they did help to bring down average money growth and
inflation.
At the same time, however, the creation of money substitutes and
the deregulation of interest rates were making Ml a less reliable guide
to future behavior of economic activity and prices.

Consequently, the

FOMC changed procedures once again late in 1982, adopting a borrowed
reserve procedure resembling the free reserve technique of the 1960s.
The degree of reserve pressure--defined as the volume of reserves that
banks as a group were forced to borrow at the discount window--was
adjusted judgmentally when developments in the economy, money, or prices
suggested that a change was appropriate.

Over time, the borrowing

relationship that underpinned this approach has become less dependable.
Consequently, the Desk has once again come to rely more closely on the
behavior of the federal funds rate, although the rate has not become a
formal target.




-2-

Meulendyke

RESERVE REQUIREMENTS
This section reviews the various roles of reserve requirements in the
monetary policy process.

It describes how the monetary authorities,

charged with determining appropriate reserve requirements, have
responded to the distinct and sometimes conflicting interests of the
Federal Reserve, the banks, and the Treasury.
Particular attention is given to the different parties' views of
the optimal level of reserve requirements.
sought to minimize reserve requirements.

Historically, banks have
Because the reserves that

banks must hold against their deposits do not pay interest, the
requirements act as an implicit tax on deposit creation.

By contrast,

the Treasury has sometimes resisted efforts to lower requirements
because reserves provide it with an indirect source of revenue.

The

effective tax is sensitive to both the level of required reserves and of
interest rates and has consequently been subject to considerable
variation over time.
The Federal Reserve, approaching the issue from a somewhat
different perspective than either the Treasury or the banks, has viewed
requirements as a mechanism that can help to stabilize the demand for
reserves.

It has sought to make them high enough to promote that

stability but low enough to minimize the distortions in resource
allocation that inevitably accompany any tax.

The Board's most recent

cuts in requirements were intended to reduce the implicit tax on
banking.

The lowered requirements reduced the effective tax to less

than $1 billion; it helped depositories improve earnings and deal more
effectively with both strains on their capital and dramatically
increased insurance premia.

Along with their desirable effects,

however, the recent reductions brought required reserves to levels that
no longer met many banks' reserve needs for clearing purposes.
Consequently, the total demand for reserves became more difficult to
predict, and the use of open market operations became more complicated.
The history of reserve requirements since the 1951 Accord
encompasses numerous regulatory changes and legislative initiatives that
attempted to address these conflicting interests.

Effective required

reserve ratios have been cut substantially on balance over the years,




-3-

Meulendyke
both to reduce the distorting impact of the implicit tax on the behavior
of banks and their customers and to change reserve pressures.

Reserve

requirements since the Accord are shown in Chart 1. Required reserve
balances at the Federal Reserve are currently very similar in level to
those of the early 1950s despite the massive growth in deposits over the
intervening decades.
The Roles of Reserve Requirements
Over the years, analysts have attributed several different roles to
reserve requirements in the policy process. The literature since World
War II has most commonly cited two--money control and revenues for the
Treasury.3

Reserve requirements could affect the process of monetary

control by their existence and through changes in the mandated ratios of
reserves to deposits. The existence of requirements provides the
linkage that allows changes in reserve levels, accomplished through open
market operations, to encourage a change in monetary deposits. In
theory, in a system where required reserves are a specified fraction of
deposits, an increase in the amount of reserves provided to the banking
system should be associated with an increase in reservable deposits in
an amount that is a multiple of the reserve increase.

The size of the

multiple would be the inverse of the required reserve ratio, as in the
classic textbook reserve multiplier process.

In practice, the

relationships linking reserves and deposits are far from precise, partly
because not all deposits are subject to the same reserve requirement
ratios and partly because excess and borrowed reserve levels can vary.
The primary direction of causality linking deposits and reserves
will depend upon the Federal Reserve's guidelines for reserve provision.
Regardless of its operating procedures, the Fed has found the existence
of reserve requirements to be a valuable tool of monetary policy because

3. See Marvin Goodfriend and Monica Hargraves, "A Historical
Assessment of the Rationales and Functions of Reserve Requirements,"
Federal Reserve Bank of Richmond Review, March-April 1983, for an
excellent review of the rationales for reserve requirements.




-4-

Meulendyke

requirements contribute to a stable demand for reserves.4

A number of

observers have argued that reserve requirements are not essential
because banks would demand reserves in any case to settle transactions
with other banks and to avoid overdrafts.5

Many Federal Reserve

commentators have rejected this claim, contending that the voluntary
demand for reserves would probably not be stable in the absence of
requirements because the banks would always be trying to minimize excess
reserves but would have varying degrees of success depending on each
period's reserve flows.6
The Board of Governors of the Federal Reserve System may also
change reserve requirement ratios to influence monetary policy.

To

force a contraction in deposits, the Board can raise requirements; to
encourage more expansion, it can lower requirements.

Although such

measures may accomplish desired adjustments in reserve availability,
they tend to be a blunt instrument, not well suited to fine tuning.

The

Federal Reserve discovered that problem in the 1930s, when legislation
first gave it the power to change reserve requirements.

In recent

decades, it has generally used open market operations to cushion the
immediate impact of a reserve requirement change.
As noted earlier, reserve requirements have also been seen as a
source of revenue for the Treasury since they represent an implicit tax

4. Gordon H. Sellon, Jr., "The Instruments of Monetary Policy,"
Federal Reserve Bank of Kansas City Economic Review,
May 1984, pp. 3-20,
discusses this issue.
5. For examples, see Deane Carson, "Is the Federal Reserve System
Really Necessary?" Journal
of Finance,
vol. 19, no. 4 (December 1964),
pp. 652-61; and Robert E. Hall, "A Free Market Policy to Stabilize the
Purchasing Power of the Dollar," in Barry Seigel, ed., Money in
Crisis:
The Federal
Reserve,
The Economy, and Monetary Reform, Pacific Studies
in Public Policy (Cambridge, Massachusetts: Ballinger, 1984), pp. 30321. Thomas Mayer, Monetary Policy
in the United States
(New York:
Random House, 1968), pp. 39-43, discusses the theoretical arguments
against requirements but concludes that they are useful, giving reasons
similar to those cited in the text.
6. Richard D. Porter and Kenneth J. Kopecky, "The Role of Reserve
Requirements as a Public Policy Tool," Conference
on
Reserve
Requirements
and the Role of the Federal
Reserve
System,
Washington,
D.C., January 18-19, 1979.




-5-

Meulendyke
on deposit creation.

Required reserves on which no interest is paid

reduce bank earnings--at least to the extent that the level of reserves
exceeds what banks would hold voluntarily.

They also enhance the

revenues of the Federal Reserve because the Fed buys interest bearing
Treasury debt when it supplies the reserves. The Treasury benefits
indirectly because the Federal Reserve turns its profits over to the
Treasury.

How burdensome a given level of requirements will be for

banks depends on several factors, but especially on the level of nominal
interest rates: the higher the rates, the greater the earnings forgone.
Mindful of the "tax" effects of increasing reserve requirement ratios,
the Federal Reserve has often turned to other tools when it wanted to
tighten policy.
Policy Responses to Conflicts between Treasury Revenues and Money
Control.

Federal Reserve and government policies toward reserve

requirements from the end of World War II through 1980 were
significantly influenced by ongoing strains arising from the different
reserve objectives of the government, the Federal Reserve, and the
banks.

Membership in the Federal Reserve was voluntary for state-

chartered banks, so they could escape the tax by dropping their
membership.

(State requirements were lower and generally could be met

by maintaining balances at other banks, for which services were
provided, and sometimes by holding Treasury bills, which paid interest.)
The Federal Reserve wanted reserve requirements to be broad based enough
to facilitate money control.7 The Fed believed that reserve
requirements could be set in a way that would strengthen the linkages
between reserves and money and between reserves and short-term interest
rates.

The existing structure encouraged departures from Federal

Reserve membership that weakened those linkages.
The Federal Reserve proposed two solutions to this conflict
during the 1970s.

First, it called for universal membership so that all

banks would be subject to the Fed's reserve requirements.

Second, it

proposed paying interest on required reserves to offset the banks'

7. G. William Miller, "Proposals on Financial Institution Reserve
Requirements and Related Issues," testimony before the House Committee
on Banking, Finance and Urban Affairs, July 27, 1978.




-6-

Meulendyke

revenue loss and to make membership in the Federal Reserve System
attractive.8

The generally high nominal interest rates prevailing

during the 1970s made requirements particularly onerous and increased
the incentive to surrender membership.

Negotiations to address these

issues culminated in the Monetary Control Act of 1980 (MCA).

The act

extended reserve requirements to all depository institutions while
allowing membership to remain voluntary.

It also lowered required

reserve ratios to reduce the implicit tax on member banks.
Although the lower requirements helped to ease the implicit tax
on banks, the exceptionally high interest rates of the early years of
the 1980s lifted the implicit tax so that the potential earnings of many
depositories were significantly diminished and their ability to pay
competitive rates thereby constrained.

Wide spreads between market

rates and deposit rates encouraged depositors to move funds into
instruments exempt from reserve requirements.

The Federal Reserve

continued to ask for the right to pay interest on required reserve
balances (in conjunction with allowing interest on demand deposits) but
its appeals were not successful.9
The eight-year phase-in period for the new reserve requirement
structure mandated by the MCA discouraged the Fed from making changes in
requirements for monetary policy purposes.

The role of requirements in

money control continued to be discussed; it was especially important
between 1979 and 1982 when the Fed was seeking to control Ml by
adjusting nonborrowed reserves.10

Thereafter, as the Fed moved away

8. Both the Federal Reserve's proposals for legislation and some
alternative proposals appear in Miller, "Proposals on Financial
Institution Reserve Requirements."
9. See statement by J. Charles Partee before the Subcommittee on
Financial Institutions Supervision, Regulation and Insurance of the
House Committee on Banking, Finance and Urban Affairs, October 27, 1983,
reported in the Federal
Reserve
Bulletin,
November 1983, pp. 850-51.

10. To improve the linkage between reserves and deposits, the
Federal Reserve did switch from lagged reserve accounting to almostcontemporaneous reserve accounting, a change that was announced in 1982
but not put into effect until 1984.




-7-

Meulendyke
from Ml control, the reserve-Ml linkage received less attention.
Nevertheless, even now the linkage is used to forecast required reserves
and banks' demand for reserves.
Required Reserves and Their Role in Bank Liquidity.

In the nineteenth

and early twentieth centuries, most analysts believed that an important
function of required reserves was providing liquidity to the banks.
Most postwar commentary on reserve requirements has, however, downplayed
the idea. Many writers have pointed out that if banks have to hold
reserves to meet requirements, they cannot simultaneously use those
reserves to make loans or handle unexpected withdrawals.11 That
conclusion is almost certainly appropriate when the object is to provide
liquidity over time.
Nonetheless, reserve balances do provide a very important form
of liquidity for periods shorter than the time interval over which
requirements must be met on average (one or two weeks in recent
decades).

These balances constitute a clearing mechanism for interbank

check and wire transfers.

Far from being sterile balances sitting idly

at the Federal Reserve, as they are described in many textbooks,
reserves actually flow from one depository institution's account to
another's many times a day.
The short-run liquidity role of reserve requirements garnered
some attention within the Federal Reserve during the 1980s. At that
time, the Fed was seeking an explanation for observed increases in
excess reserves.12 Understanding the importance of the Fed's findings

11. Before the founding of the Federal Reserve, there was no
regular mechanism to produce extra reserves to meet seasonal credit
needs. Small banks kept part of their reserves in the form of deposits
at large banks and used those reserves to meet their seasonal needs.
The withdrawal of interbank deposits from the large cities actually
extinguished reserves, forcing interest rates to climb sharply higher at
those times. These liquidity problems have been widely discussed. See,
for instance, Thomas Mayer, James S. Duesenberry, and Robert Z. Aliber,
Money, Banking, and the Economy, 3d ed. (New York: W.W. Norton and
Company, 1987), pp. 28-29.
12. The large volumes of daylight overdrafts also alerted the
Federal Reserve to some banks' heavy dependence on reserve balances for
clearing activities.




-8-

Meulendyke

requires a brief review of the composition and uses of required
reserves.13
Since 1959, banks have been able to satisfy reserve requirements
by holding vault cash and reserve balances at the Federal Reserve.
Beginning in 1968, the vault cash applied to meeting reserve
requirements in the current period was the vault cash banks had held in
an earlier period.

Consequently, vault cash could not play a role in

meeting the banking system's marginal reserve requirements once a
reserve maintenance period began.

Since the reserve requirement

restructuring of the 1980s, many depository institutions, including
small commercial banks, thrifts, and credit unions, were able to meet
their reserve requirement with vault cash alone.

It does not appear,

however, that the requirements determine the institutions' holdings of
vault cash; instead these institutions base their holdings on
anticipated customer demands for currency and a strong preference not to
be embarrassed by shortages of cash.

For institutions that consistently

meet or more than meet their reserve requirements with vault cash
("nonbound" institutions), reductions in the level of the requirements
are of no consequence.14
Those medium and large depository institutions that do not cover
their whole requirement with vault cash ("bound" institutions) have to
hold on average during each reserve maintenance period sufficient
reserve balances at the Federal Reserve to meet the remainder of their
requirement (called required reserve balances).

But those reserve

balances also serve as the means of payment for the clearing and
settlement process.

Any depository that does even a portion of its own

clearing of checks or funds wires has to maintain a reserve balance to
facilitate that clearing.
The volume of transactions executed each day using reserve
accounts as a means of payment has long been high relative to the

13. The following discussion draws heavily from Ann-Marie
Meulendyke, "Monetary Policy Implementation and Reserve Requirements,"
internal working paper, September 1992, pp. 3-5.
14. The Federal Reserve excludes surplus vault cash from its
measures of total and nonborrowed reserves.




-9-

Meulendyke

balances held in the accounts.
turn over many times a day.

For many depositories, reserve balances

That turnover rate has had an upward trend.

The trend reflects cuts in reserve requirements that occurred between
1980 and 1984, and again in 1990 and 1992, and increases in the volume
of transactions being processed by the Federal Reserve.15
3 show recent patterns in these measures.16

Charts 2 and

The daily flows have a

large predictable component, but considerable potential for surprise
remains.

The Federal Reserve generally processes instructions to pay

out reserve balances even if the action puts the sending bank into
overdraft.

The Fed imposes a penalty charge on any institution that

ends the day overdrawn.

Consequently, depository institutions have to

aim for a significant positive end-of-day balance to minimize the risk
of an inadvertent overdraft, regardless of their reserve requirements.
When reserve requirements were reduced, it became more common for
precautionary needs to exceed required reserve balances.
Depository institutions can deal with these additional
precautionary reserve needs by holding excess reserves, but this
strategy is costly since no interest is paid on reserves.

When required

reserve balances declined in the early 1980s and again at the end of
1990, depositories continued to try to minimize excess reserve holdings,
but they were restricted in their ability to do so as the difficulties
in avoiding overnight overdrafts became more severe.

If they ended up

with excess reserves, they might not be able to work them off later in
the same maintenance period without risking being overdrawn.

In trying

to cope with the narrowing of ranges of reserve balances that were

15. Since 1980, depositories have been able to establish required
clearing balances to provide some reserve management flexibility. These
are additional reserve balances that depositories agree in advance to
hold. In return, they receive credits to pay for priced Federal Reserve
services. The level of priced services used by a depository provides an
effective maximum demand for required clearing balances. Required
clearing balances were fairly small until after the 1990 cut in reserve
requirements, when many large banks started to hold them.
16. Fedwire transactions have the largest impact on reserve
balances, but other wire transfer operations and check processing
transactions also lead to reserve transfers. These other transactions
raise the turnover rate for reserve balances even further.




-10-

Meulendyke

acceptable in the management of reserves, depositories devoted
considerable resources to monitoring internal reserve flows.

In the

process, they became less tolerant of excess reserves early in
maintenance periods because of their diminished ability to work them off
in subsequent days.

These developments restricted the depositories'

day-to-day flexibility in managing reserves, caused more frequent bulges
in excess reserves, and added to end-of-day volatility in the federal
funds rate.
Reserve Requirements in the 1950s and Early 1960s
At the time of the Treasury-Federal Reserve Accord of 1951, reserve
requirement ratios on demand deposits of Federal Reserve member banks
were 24 percent for banks located in "central reserve cities" (New York
and Chicago), 20 percent for member banks in "reserve cities" (other
cities with Federal Reserve Banks or branches), and 14 percent for
"country banks" (the term for all other member banks).

The reserve

ratio for time and savings deposits was 6 percent for member banks in
all locations.
During the fifteen years between 1951 and 1966, requirements
were raised on five occasions and were lowered ten times.17

The

changes in reserve requirements were sometimes made in conjunction with
complementary changes in the discount rate, while at other times the
moves were made independently.

Open market operations were used to

cushion the changes in reserve requirements, so that hardly any of the
immediate impact of the reserves released or absorbed was felt as a
change in excess or borrowed reserves.
In those years, the Federal Reserve formally described reserve
requirements as a policy tool used to make reserves more or less
plentiful so as to alter credit availability and money market interest

17. Reserve requirement ratios were changed for several reasons
over these years. Although many of the changes were undertaken to make
reserves more or less costly as part of the monetary policy process,
changes were also made to meet seasonal reserve demands and to implement
the 1959 legislation aimed at equalizing reserve ratios at central
reserve and reserve city banks. In addition, ratios were slightly
modified in 1966 when tranches were introduced for both demand and time
deposits. At the same time, savings accounts were separated from time
deposits for required reserve calculations.




-11-

Meulendyke

rates--the near-term policy goals of the time.18

Its decisions about

reserve requirements were, in practice, constrained by the exodus of
small banks from the Federal Reserve System in the 1950s.

Legislation

passed in 1959 addressed an apparent inequity between large and small
banks in an attempt to make membership more attractive for the small
banks.

Country banks had lower nominal reserve requirements, but they

often had to tie up relatively large sums in non-interest-earning
balances that did not serve any other purpose.
generally handled payment clearing for them.)

(A reserve city bank
Because of their customer

bases, most country banks had to hold relatively high amounts of vault
cash, but they could not use these holdings to satisfy requirements.
The 1959 act permitted the Fed to count vault cash toward meeting
reserve requirements.

That change--implemented in three steps during

1959 and 1960--reduced effective requirements, especially for country
banks.

It was hoped that the lower requirements would encourage those

banks to remain members of the Federal Reserve.
Contemporary Views of Reserve Requirements.

A commonly held view about

reserve requirements was expressed by a presidential commission
appointed in 1963 to study financial institutions.

The commission

concluded that "there is, within broad limits, little basis for judging
that in the long run one level [of reserve requirement ratios] is
preferable to another in terms of facilitating monetary policy."19

The

commission felt that the effects of requirements on bank earnings and
Treasury revenues should be the primary factor considered in choosing
reserve ratios.

While it saw the advantages to bank profitability of a

significant cut, it believed that the cost to the Treasury would be too
great.
Some academic literature of the time offered other views on
reserve requirements and monetary control.

Several articles and books

dealt with the concept of fractional reserve requirement ratios and

18. Board of Governors of the Federal Reserve System,
Report,
various years.

Annual

19. Report of the Committee on Financial Institutions to the
President of the United States. Walter W. Heller, Chairman.
Washington, D.C.: Government Printing Office, 1963, p. 12.




-12-

Meulendyke

described the strengths and weaknesses of that structure.
analyzed the tax implicit in reserve requirements.

20

Tolley

He suggested that

the level of reserve requirement ratios and hence of the amount of the
tax had come about by accident.
for such a tax.

He then tried to establish a rationale

He believed that under a gold standard, a system in

which real resources had to be devoted to producing money, a fee was
appropriate to encourage people to economize on the use of money.

But

when the cost of producing money is trivial, as it is with fiat money,
the only justification for a charge is that the government could benefit
from the revenues arising from the Federal Reserve's provision of
reserves.

Tolley went on to observe, however, that the government's

gains would cause misallocation of resources as banks took actions to
reduce the effect of the tax.
low reserve requirements.

Such a distortion would argue for very

But Tolley thought very low requirements

might make monetary control difficult because shifts between currency
(which is effectively subject to a 100 percent reserve requirement) and
deposits would have a large impact on the amount of money created, as
would mistakes in estimating reserve provision.

Hence, he recommended

that interest be paid on required reserves so that requirements would
not need to be reduced.
Friedman also discussed how shifts in preferences between
currency and deposit holdings could ease or tighten reserve
conditions.21

He reiterated the arguments from the 1930s for 100

percent reserve requirements.

Such requirements had been proposed as a

solution to the unpredictable multiplier effects of fractional reserve
accounting arising from the differential treatment of deposits and
currency.

Friedman also recognized the undesirable tax effect of 100

percent requirements and described the inevitable incentive for money
and credit provision to move outside the regulated area of banking.
combat that problem, he recommended paying interest on reserves.

To

Later,

20. George S. Tolley, "Providing for Growth of the Money Supply,"
Journal
of Political
Economy, December 1957, pp. 477-85.

21. Milton Friedman, A Program for Monetary
Fordham University Press, 1959), pp. 65-76.




-13-

Stability

(New York:

Meulendyke
the Federal Reserve seriously considered the proposal to pay interest on
reserves; it has periodically requested authority to do so from the
Congress.
Reserve Requirements in the Latter Part of the 1960s and 1970s
Reserve requirements continued to be raised and lowered to reinforce
tightening or easing moves implemented with other tools during the rest
of the 1960s and 1970s. Requirements were increased four times and
decreased seven times during these years.22 Sensitivity to the
membership problem sometimes made the Federal Reserve Board hesitant to
raise requirements.

On occasion, the Board raised them just on large

time deposits--deposits mostly issued by the large banks, which were the
least able to give up the services provided by Fed membership.

The

combination of higher inflation and higher interest rates that emerged
during these years drew increasing attention to the tax burden of
reserve requirements and the related question of differential treatment
of member and nonmember banks.
The Federal Reserve appointed a study group headed by Robert
Black to review reserve requirement ratios. The group reported its
recommendations in 1966.23

The primary result of that study was the

decision to move from near-contemporaneous reserve requirements with
one-week reserve maintenance periods for reserve city banks and two-week
periods for country banks to weekly reserve periods for all member banks
with a two-week lag between the computation and maintenance periods.
This change was believed to make calculating requirements easier for the
banks and the New York Fed's Trading Desk.24

22. The count does not include the 1972 restructuring that raised
requirements for some banks and lowered them for others, as described
later in the text.
23.

Proposals,

Robert P. Black, Report

of

the Ad Hoc Subcommittee

on

Reserve

May 13, 1966.

24. The other change was to permit banks to carry forward reserve
excesses up to 2 percent of required reserves for one reserve period.
(Banks already had the authority to carry forward 2 percent of reserve
deficiencies.)




-14-

Meulendyke

Lagged reserve requirements weakened the direct linkage between
reserves and money, making it harder, in theory, to manipulate reserves
as a means of controlling money.

For the most part, the Federal Reserve

did not see any reason to be concerned because it was not attempting to
control money in this way.

Instead, the Fed was attempting to affect

money growth indirectly by influencing the demand for money.

It altered

the cost of obtaining reserves and hence the cost at which credit was
provided.25
In 1972, another Federal Reserve reform addressed the problem of
retaining member banks.

For both reserve city and country banks,

reserve requirement ratios were to be graduated on the same schedule by
volume of deposits.

The change represented a significant cut in reserve

requirements for small banks in Federal Reserve cities and caused some
large banks outside of Federal Reserve cities to face higher
requirements.

The series of graduated steps in the required reserve

schedule further weakened the relationship between required reserves and
monetary deposits, an outcome that distressed those economists who
wanted to see the Federal Reserve control reserves in order to control
money growth.

At the time, the Federal Reserve was targeting the

federal funds rate and reserve requirements were lagged, so the concerns
were not immediately relevant to operations.26
Nonetheless, Federal Reserve membership continued to decline.
The Federal Reserve proposed paying interest on reserves on a couple of

25. Lyle E. Gramley and Samuel B. Chase, Jr., "Time Deposits in
Monetary Analysis," Federal
Reserve
Bulletin,
October 1965, pp. 13801404.
26. Nonetheless, shortly afterwards the Federal Reserve did take
limited steps to use reserve targeting when it experimented with
reserves on private deposits. See Ann-Marie Meulendyke, "A Review of
Federal Reserve Policy Tatgets and Operating Guides in Recent Decades,"

Intermediate
Survey,




Targets

and Indicators

for Monetary Policy:

A

Critical

Federal Reserve Bank of New York, July 1990, pp. 463-64.
-15-

Meulendyke
occasions in the 1970s to halt the decline, but the revenue loss to the
Treasury engendered strong congressional opposition.27
The Monetary Control Act and Reserve Requirements in the 1980s
At the end of the 1970s, the Federal Reserve once again tried to achieve
universal membership.

Although it did not literally accomplish that, it

did achieve, through the 1980 MCA, the most important goal associated
with expanded membership: the extension of reserve requirements to all
depository institutions.

Furthermore, the Fed was permitted to collect

deposit data on an ongoing basis from all but the smallest depositories,
enabling it to improve both estimates of actual money and forecasts of
future money.

Reserve requirement ratios for member banks were cut over

a four-year period from a top rate of 16 1/4 percent to a top rate of 12
percent on transactions deposits. A low reserve tranche was also
established of 3 percent on the first $25 million of deposits, with the
amount allowed to rise over time.28 Nonmember banks and thrifts that
faced the increases in requirements were given an eight-year phase-in
period to reach the final levels of requirements specified in the act.
The Federal Reserve Board retained the option to adjust reserve ratios
within specified bands.
The MCA was directed toward improving the Fed's ability to
control money.

It focused on deposits in Ml, the primary intermediate

policy variable at the time.

It did not, however, provide any scope for

using reserves to control M2, a secondary target at the time the act was
passed but the primary monetary target later in the decade. Money
market mutual fund balances remained exempt, and MCA actually took away
from the Federal Reserve the power to impose reserve requirements on
personal time and savings deposits.
Aside from the changes to reserve requirements mandated by the
legislation, only minor modifications were made to reserve requirements

27. Specific proposals to pay interest on reserves were introduced
in the Congress in 1977 and 1978. See Stuart E. Weiner, "Payment of
Interest on Reserves," Federal Reserve Bank of Kansas City Economic
Review, January 1985, pp. 20-21.
28. In 1982, the Gam-St Germain Act modified the reserve
requirement structure further to introduce a zero requirement tranche.




-16-

Meulendyke
during the 1980s.29

Because the structure of requirements had been set

within specified limits by the MCA, it was generally felt that there was
little point to considering policy-related changes in the ratios.

Such

changes would have been difficult to implement during the eight-year
phase-in period.

Since the legislation had not given the Federal

Reserve the option to pay interest on reserve balances, the Board might
have hesitated to raise requirements because of the implied increase in
the tax burden.30

Furthermore, the Federal Reserve believed it could

achieve its objectives just as well through open market operations and
discount window policy.
Excess Reserve Behavior and Potential Problems with Reserve
Requirements.

The Federal Reserve saw increasing evidence during the

1980s that depository institutions were having difficulty managing
reserves.

These observations suggested that reserve requirements might

be inadequate for smooth monetary operations.

Normal levels of excess

reserves rose fairly steadily in the years following passage of the MCA.
Some of the increase was the inevitable result of extending reserve
requirements to nonmember depository institutions.31

But member bank

excess reserves were also rising, in a pattern that contrasted with
their behavior during much of the 1970s, when they generally hovered in
a range near $200 million.
discoveries.

The search for explanations led to several

It was observed that excess reserves tended to move

inversely to required reserves not met by vault cash, both period to

29. In March 1983, the Board eliminated reserve requirements on
time deposits with an initial maturity of two and one-half years or
more. In September 1983, it reduced the minimum maturity for exemption
from requirements to eighteen months.
30. The MCA did provide for payment of interest on supplemental
reserve requirements if such requirements were needed for monetary
control. The provision has not been used.
31. At some point during the phase-in period, vault cash no longer
met all of the larger nonmember institutions' requirements, and they
opened reserve accounts at the Federal Reserve. Only then could these
institutions have excess reserves. (Previously, they may have had
excess reserves from their own perspective in the form of surplus vault
cash and deposits at correspondents, but the Federal Reserve does not
count these in its reserve measures.)




-17-

Meulendyke

period and over time, as balances held at Federal Reserve Banks trended
lower.32

The sharp drop in required reserve balances between 1980 and

1984 occurred as lower reserve requirements were being phased in for
member banks under MCA and the spread of automatic teller machines was
encouraging rapid expansion of vault cash holdings (Chart 1 ) .
Average required reserve balances rose again in the next few
years, but excess reserves continued to expand at member banks as well
as at nonmember banks.

Conversations with officials at a number of

banks underscored the growing role of large payments flowing through
their reserve accounts.

The volume of wire transfers over Fedwire--the

Federal Reserve's wire transfer system--grew rapidly (Chart 3), making
it increasingly difficult for banks to predict reserve balances.

Since

the Federal Reserve penalized end-of-day overdrafts, banks had to be
careful not to aim for too low a reserve balance lest an unexpected late
day outflow (or an expected receipt that did not arrive) should leave
them overdrawn.

These discoveries suggested that for a number of banks,

reserve balances needed to meet requirements were not very different in
size from those needed to manage clearing and settlement and to avoid
overdrafts.
These factors were taken into account by the Federal Reserve in
estimating the aggregate demand for excess reserves.33

But they did

not lead to serious discussions of the structure of reserve requirements
during the 1980s.
Cuts in Reserve Requirements in the 1990s
The Federal Reserve Board eliminated reserve requirements on
nontransaction deposits at the end of 1990.

In explaining its action,

the Board indicated that the existing structure had been designed

32. David Jones, "Excess Reserves under MCA," November 10, 1983,
and David Small and Brian Madigan, "An Analysis of Excess Reserves,"
July 1, 1986, internal memoranda, Board of Governors of the Federal
Reserve System.
33.
Markets,




Ann-Marie Meulendyke, U.S. Monetary Policy
and
Financial
Federal Reserve Bank of New York, 1990, chap. 6.

-18-

Meulendyke

"primarily to permit greater precision of monetary control when
policy focused on reserve aggregate targeting."

It went on to

describe the changing conditions that had prompted its move:
In subsequent years, as the Federal Reserve, moved away
from the procedures in effect in the early 1980s, which
required a broad reserve base, reserve requirements on
nonpersonal time accounts have become somewhat of an
anachronism. Moreover, the current 3 percent requirement
has placed depository institutions at a disadvantage
relative to other providers of credit, spawning efforts to
circumvent the requirement.
The Board took action at this time also in response
to mounting evidence that commercial banks have been
tightening their standards of creditworthiness [a
development that] has in recent months begun to exert a
contractionary influence on the economy. . . . Lower
reserve requirements at any given level of money market
interest rates will reduce costs to depository
institutions, providing added incentive to lend to
creditworthy borrowers.3A
The reduction in reserve requirements boosted earnings for some
depository institutions but, as indicated earlier, it had the
undesirable side effect of complicating reserve management for many
institutions.

With lower routine levels of required reserve balances,

their ability to accept reserve variability from day to day within a
two-week reserve maintenance period without either incurring an
expensive overdraft or being stuck with unusable excess reserves was
reduced.

Depositories found they had to use considerable resources to

hold down excess reserves.

The action also complicated operations of

the Open Market Trading Desk at the New York Federal Reserve Bank.
Relatively modest reserve excesses often inspired sharp declines
in the federal funds rate, even on days that were not the ends of
maintenance periods.

Depositories had less ability to absorb and make

use of the excess reserves because they could not run large deficiencies
in subsequent days without ending overdrawn.

When a number of

depositories discovered toward the end of a day that they had excess
reserve positions and tried to sell the funds into the interbank federal

34.




Federal

Reserve

Bulletin,

February 1991, p. 95.

-19-

Meulendyke
funds market, their efforts often pushed the funds rate down sharply,
sometimes almost to zero. At that time of day, it is too late for open
market operations to be undertaken to affect that day's reserves.
Hence, depositories as a group could not eliminate the excesses except
by repaying discount window loans.

In 1991, routine borrowing from this

source was already at very low levels, so little could be repaid.
Low reserve balances also increased the likelihood of an
incipient overdraft.

Depositories that discovered they were overdrawn

late in the day generally tried to cover the overdrafts by borrowing in
the federal funds market.

If funds were scarce systemwide, sufficient

reserves might not be available.

Depositories could obtain reserves

from the discount window, but in the early months of 1991, many banks
were unusually reluctant to borrow for fear that such a step could be
read as a sign that they were in trouble.

That reluctance to borrow

often caused federal funds to be bid to very high levels before some
banks finally turned to the window to cover the shortages.
The Desk's Approach to Managing Reserves in this Environment.

At the

time of the 1990 reserve requirement cut, the Desk was formally
targeting borrowed reserves.35

Because the relationship between

borrowing and the funds rate remained unreliable, however, the Desk was
also taking considerable guidance from the federal funds rate.

The Desk

still attempted to achieve the levels of nonborrowed reserves believed
consistent with demands and the desired degree of reserve pressure, but
demands became harder to gauge after the cut in requirements.36

In

choosing its reserve management strategy, the Desk had traditionally
focused on two-week average reserve levels that banks had to hold over

35. Ann-Marie Meulendyke, "A Review of Federal Reserve Policy
Targets and Operating Guides in Recent Decades," Intermediate
Targets
and Indicators
for Monetary Policy: A Critical
Survey, Federal Reserve
Bank of New York, July 1990, describes the formal procedures. Recent
modifications are discussed in "Monetary Policy and Open Market
Operations during 1990," Federal Reserve Bank of New York Quarterly
Review, Spring 1991, pp. 66-74.
36. See "Monetary Policy and Open Market Operations during 1991,"
Federal Reserve Bank of New York Quarterly Review, Spring 1992,
pp. 80-88.




-20-

Meulendyke

the maintenance period as a whole, although it had also made efforts to
avoid extreme movements in daily reserve levels.

Once reserve

requirements were reduced, the Desk had to pay increased attention to
daily levels because of the depositories' diminished tolerance for being
short or long relative to their requirements.

The Desk often found that

the funds rate in the morning was not a good guide to reserve
availability; the rate sometimes plunged or rose sharply late in the day
when the depositories finally discovered that reserves were plentiful or
scarce.
Because market participants judged the Fed's policy stance by
the behavior of the federal funds rate, the signaling of policy
intentions sometimes conflicted with the desired reserve management
strategy.

If, for instance, an estimated reserve shortage coincided

with a funds rate level below that perceived to be the target, the Desk
had to decide whether to meet the estimated reserve need.

If it met the

need, it would risk giving a misleading indication that the stance of
policy had been eased.

But not meeting the need would increase the

chances of a sharp rise in the funds rate late in the day, possibly
accompanied by heavy discount window borrowing.

Such greater than

desired reserve pressure imposed an unintended cost on the banks and
involved a risk that observers could be misled about policy.

Although

these conflicts had been a periodic feature of reserve management for
years, they increased in frequency once levels of required reserve
balances fell.
Reserve balances rose during 1991, helping to ease somewhat the
difficulties of reserve management.

However, another cut in reserve

requirement ratios in April 1992 once more lowered the range of
flexibility in day-to-day management of reserves, although typical
reserve balance levels remained above those of the early part of
1992.37

37. A series of papers prepared by the staff of the Federal
Reserve Bank of New York after the 1990 cut in required reserve ratios
considers the operational difficulties of low required reserve ratios
and evaluates possible solutions. Overall, the papers suggest that the
best solution to the reserve management problems encountered with low




-21-

Meulendyke
THE ROLE OF THE DISCOUNT WINDOW IN POLICY IMPLEMENTATION
Like reserve requirements, the discount window has played a supporting
role to open market operations in the monetary policy process. This
section describes the guiding principles for discount window borrowing.
It reviews the two main features of that borrowing, the rules that
govern the use of the facility and the rate or rates that are charged.
It then provides a chronological review of developments in the behavior
of borrowing from the 1950s to the present.
The Philosophy behind the Discount Window Mechanism
Federal Reserve views of the discount window's roles changed
considerably between the founding of the Federal Reserve in 1914 and the
1930s as open market operations gradually replaced discount window
borrowing as the primary source of Federal Reserve credit. Then,
between 1934 and 1950, the discount window fell into disuse, and there
was little consideration of the roles of the window as a policy tool.
The Federal Reserve's concept of the policy role of the discount
window was reexamined after the 1951 Accord and again in the latter half

reserve balances would be to pay interest on reserves so that
requirements could be increased without raising the costs to depository
institutions.
The collection of papers also evaluates other alternatives. A
return to more routine use of the discount window would provide the
banking system with valuable flexibility, but overcoming the current
strong reluctance to borrow appears to be a difficult challenge.
In the absence of such changes, only one of the other
alternatives could provide more than modest help to the reserve
management process: permitting banks to end the day overdrawn.
Nonetheless, permitting overdrafts would have significant drawbacks. If
this approach were to be seriously considered, permitted overdrafts
would have to be collateralized and made subject to a modest charge.
Even so, it seems to go against the thrust of efforts to reduce daylight
overdrafts and could be seen as weakening the essential discipline of a
reserve requirement structure.
Other approaches deserving consideration include expanding
reserve carryovers and shortening the vault cash lag, variants of which
have recently been introduced by the Board of Governors. These
approaches, however, would raise reserve management flexibility only
slightly.




-22-

Meulendyke

of the 1960s.

Both studies led to some modifications in the rules for

borrowing but did not change the underlying philosophy.

Most of the

rule changes since the early 1970s have been small and have addressed
specific concerns.
Since the Accord, the Federal Reserve's discount window policy
has discouraged persistent reliance on borrowing.

That stance has

ensured that borrowed reserves generally represent only a modest share
of total reserves.

The Fed believes that the discount window should

serve as a safety valve, a temporary source of reserves when they are
not readily available from other sources.38

The window in recent

decades has been available to healthy banks for occasional, but not
continuous, use.39

Borrowing has been rationed through a variety of

means that have encouraged a "reluctance to borrow."

The degree of

reluctance shown by the banks has varied considerably over the years,
even in the absence of changes in the guidelines for borrowing.
At the same time, the Fed has counted on there being some amount
of borrowing because borrowing is an element in the reserve adjustment
process.

In this context, the window has played a vital role in meeting

unexpected reserve needs.

Various open market operating procedures

depend on some degree of stability in the banks' demand for borrowed
reserves, but the administrative guidelines and changing bank attitudes
have made this stability difficult to achieve.

For much of the time

since the mid-1960s, the discount rate has been below competing market
rates, in particular the overnight federal funds rate.

Consequently,

administrative restrictions rather than the rate have had the biggest
role in limiting the amount of borrowing.

Banks have responded to the

profit incentive to borrow, but in doing so they have had to factor in

38. All borrowing from the Federal Reserve must be fully
collateralized.
39. At times, the Fed also provides extended credit at market-based
rates to banks whose financial difficulties have cut them off from
regular sources of financing. Banks using the facility must work with
their regulators toward a solution. That type of borrowing is not a
monetary policy tool, and thus is not a focus of this piece.




-23-

Meulendyke

some nonprice costs--such as potential loss of future access to the
window--that are difficult to estimate.
During the 1980s, increasing financial difficulties and bank
failures led banks to become more reluctant to borrow, even under
conditions that would formerly have led them to borrow.

The rise in

banking crises made many banks fearful that if they borrowed, rumors
would start that they were in financial trouble.

Thus, the demand for

borrowing became even less predictable, reducing the value of the
relationship between borrowing and the spread between the Federal funds
rate and the discount rate.
The direct cost represented by the rate charged for discount
window borrowing has also played some role in the policy process.
Changes in the rate have normally attracted general attention to the
state of monetary policy, giving rate changes the potential for an
announcement effect.

The extent of the announcement effect has varied

over time, depending on the verbal message given with the rate change
and the way borrowing was being used in the policy process.

Sometimes

the Fed has sought to signal policy changes when it changed the rate.
At other times it deliberately downplayed the significance of the move.
Changes in the discount rate are voted by the Boards of
Directors of the twelve Federal Reserve Banks and approved by the Board
of Governors.

The governors generally approve changes in the rate when

they want to signal a change in the stance of policy or when market
rates have moved significantly away from the discount rate, so that the
discount rate is "catching up" with the changes.

Rate changes have

normally complemented the guidelines established by the FOMC for the
conduct of open market operations.
The discount rate per se has not, in the post-Accord period,
been regarded as a primary means of influencing the amount of discount
window borrowing.

Indeed, because short-term interest rates have

frequently exceeded the discount rate since the mid-1960s, rationing of
the use of the window has had to be accomplished through means other
than the rate.

There have been numerous recommendations over the years

that the rate be given the primary role in rationing credit, either
because the approach was more straightforward and less arbitrary than




-24-

Meulendyke

rationing administratively or because the use of a below-market rate
implied a subsidy.

The specifics of the relationship between the

discount rate and open market policy changed modestly when the
techniques of policy implementation were changed but have throughout
relied on administered disincentives to borrow.
The Discount Window in the 1950s Through the mid-1960s
Borrowing jumped dramatically in the early 1950s.

It rose from an

average of $130 million in 1950 to an average of $800 million in 1952.
By December 1952, it had reached $1.6 billion.

Interest rates rose

after the Accord, and the discount rate lagged behind.

(Chart 4 shows

borrowed reserves and their share of total, reserves between 1950 and
1965, along with the discount rate and short-term interest rates.)

The

cost structure made borrowing attractive for the first time since the
early 1930s.

An excess profits tax instituted in 1951 increased the

incentive to use the discount window because borrowings served as an
offset in computing the tax.
A Federal Reserve System committee was established in 1953 to
examine the history of the rationales for borrowing.

The committee

concluded that the established "tradition against borrowing" should be
encouraged because it contributed to the soundness of individual banks
and the banking system.40

The committee report served as the basis of

the 1955 revisions to Regulation A, the regulation governing use of the
window.A1
The report observed that the founders of the Federal Reserve had
expected the discount window to be the primary source of Federal Reserve
credit.

In the early years of the Federal Reserve, many member banks

borrowed a substantial portion of the reserves they needed from the
window; indeed, it was not unusual for a bank to borrow continuously.
By contrast, in the years before the founding of the Federal Reserve, a
bank that was heavily dependent on borrowed funds, rather than on its
own capital and deposits, was believed to be more vulnerable to failure.

40. System Committee on the Discount and Discount Rate Mechanism,
"Report on the Discount Mechanism," March 12, 1954.
41.




Federal

Reserve

Bulletin,

January 1955, pp. 8-14.
-25-

Meulendyke
The committee noted that the development of open market
operations during the 1920s as an alternative source of Federal Reserve
credit made possible a gradual move to discourage heavy borrowing.

Once

again, banks that borrowed persistently came to be seen as more likely
to fail, and this view was reinforced during the early 1930s when the
number of bank failures soared.

Mindful of this negative image, the

banks themselves became reluctant to borrow and instead built up
holdings of excess reserves during the latter part of the 1930s. This
course of action was simplified by the monetization of the vast gold
inflows inspired by the revaluation of gold in 1934 and by the approach
of war in Europe in the latter years of the decade.42
By the early 1950s, however, a decade and a half with low
numbers of bank failures had apparently reduced the banks' own
reluctance to borrow to such an extent that many banks were inclined to
return to the window when doing so became profitable.
felt this behavior should be discouraged.

The committee

It reiterated the belief that

a bank that used its own resources to meet increased demands for credit
was healthier than one that was dependent on borrowed funds.

In its

1954 report, the committee recommended that routine reserve provision be
accomplished almost entirely through open market operations.

The report

also recommended limiting the term of borrowing to fifteen days under
normal circumstances.

It noted that most banks had emerged from the war

with substantial portfolios of government securities that could be sold
to raise additional funds for seasonal or other purposes. The
regulations that were subsequently adopted guided discount officers in
distinguishing between appropriate and inappropriate borrowing.
Borrowing was considered inappropriate when the funds were used for
normal business activities.

In particular, the committee disapproved of

borrowing to profit from interest rate differentials.
The role of the discount window during the rest of the 1950s and
early 1960s generally followed the pattern set out by the committee's

42.

Markets,




Ann-Marie Meulendyke, U.S.

Monetary

1990, Chap. 2.
-26-

Policy

and

Financial

Meulendyke

guidelines.

There was some debate about whether the reluctance to

borrow was motivated by the banks' own caution or by Federal Reserve
restrictions.

Some banks almost never borrowed, suggesting an

internally generated reluctance.

Many banks, however, apparently took

account of the full cost of borrowing, including potential loss of
future access, and borrowed when it was profitable.
borrowing was rarely a large bargain.

In that context,

In fact, the discount rate was

often slightly above short-term Treasury bill rates, although both
borrowing and the incentive to borrow varied cyclically.

Normally,

borrowing was only a modest share of total Federal Reserve credit.
The Board of Governors approved periodic adjustments to the
discount rate and issued a statement of purpose with each adjustment.
Often the changes lagged market rates, and the Board explained its
action as an effort to catch up with market rates.

When the discount

rate was low relative to other short-term rates, borrowing often rose.
(The primary alternative rate was the Treasury bill rate in the 1950s;
the federal funds market grew in importance during the 1960s.)
Some academic economists criticized the discount mechanism.
They did not like the fact that banks were given mixed signals about
borrowing, with the relatively low discount rate often encouraging use
of the window while the administrative guidelines were discouraging it.
They felt that the rules made it difficult to judge whether policy was
tight or easy.43

The authors preferred a rate that was set above

market rates--a penalty rate--but urged that no administrative
restrictions be placed on borrowing.
Discount Window Policy in the Late 1960s and 1970s
Higher interest rate levels in the latter half of the 1960s, especially
the "tight money" episode of 1966, encouraged more borrowing (Chart 5).
The decline in membership was also garnering attention, and there was
concern the discount window was not sufficiently available to small

43. See Milton Friedman, A Program
York: Fordham University Press, 1959),

Reserves

and the Money Supply

for Monetary Stability
(New
pp. 38-41; A James Meigs, Free

(Chicago: University of Chicago Press,

1962); and Warren Smith, "The Discount Rate as a Credit-Control Weapon,"
Journal
of Political
Economy, April 1958, pp. 171-77.




-27-

Meulendyke

member banks.

A series of studies were undertaken during the late 1960s

under the guidance of a steering committee of Federal Reserve Governors
and Presidents.44

The studies reviewed the history of the discount

mechanism, compared the discount window with the tools and techniques of
foreign central banks, evaluated some of its problems, and presented
several possible reforms.

The steering committee endorsed the practice

of permitting banks to borrow only intermittently.

It wanted to

continue the administrative disincentives to frequent borrowing, but it
was troubled that some banks seemed to get little or no benefit from the
window.

The summary report recommended some changes to make borrowing

more convenient, especially for small unit banks with large seasonal
swings in loan demand and limited access to the national credit markets.
The report's recommendation of a special seasonal borrowing privilege
for small member banks was adopted in 1973 and remains in effect,
although it has been modified somewhat in recent years.45
The report also proposed that one form of adjustment credit
should consist of a basic borrowing privilege that would give all
(member) banks some access at reasonable cost to Federal Reserve credit
based on published guidelines for amount and frequency of borrowing.
Even the proposed basic borrowing privilege did not envision continuous
borrowing: if a bank needed additional credit, its borrowing would be
subject to scrutiny.

The approach was not adopted, although the

proposed frequency schedule did influence the informal guidelines used
by the discount officers in subsequent years.

Finally, the study

brought to light considerable inconsistencies in the administration of
the window by the different Federal Reserve Banks.

Efforts were made to

improve coordination in order to minimize those differences.
During the 1970s, Federal Reserve monetary policy focused on
adjusting the federal funds rate to respond to deviations in money

44.

Board of Governors of the Federal Reserve System,

of the Federal

Reserve

Discount

Mechanism,

Reappraisal

1971.

45. The seasonal borrowing privilege was extended to nonmember
banks under the MCA. In 1992, the Board began charging a market rate on
seasonal borrowing tied to the federal funds rate and certificate of
deposit rates.




-28-

Meulendyke

growth from desired ranges.

The discount window generally played a

subsidiary role in the process.46

Changes in the discount rate were

often motivated by changes in market rates, as they had been in earlier
decades, although occasionally changes were intended to create an
announcement effect.47

The amount of borrowing generally increased as

the federal funds rate rose relative to the discount rate, a
relationship that suggested that banks were seeking to maximize profits
through their borrowing decisions.

The Open Market Trading Desk took

that relationship into account when choosing how many nonborrowed
reserves to provide, since the amount of desired borrowing affected the
reserve levels consistent with the desired funds rate.
Relation between Discount Policy and Reserve Targeting from 1979 to
1982.
Borrowing took on increased importance after the October 1979
changes to reserve operating procedures.

Under the new procedures, the

Trading Desk provided only the level of nonborrowed reserves estimated
to be consistent with targeted Ml.

If depositories needed additional

reserves to meet their requirements because Ml was above target, they
would have to borrow them at the discount window.

In practice, the

system was structured so that there was some borrowing even when Ml was
on target.

Only when Ml was far below target for a while in 1980 was

46. Economists have debated the importance of the discount rate as
a mechanism for changing policy. Sometimes Federal Reserve
announcements indicated that the rate was changed to catch up with
market rates. Other times they cited monetary policy concerns. At
issue is whether these announcements had an impact beyond that of open
market operations. See Cook and Hahn, "The Information Content of
Discount Rate Announcements and Their Effect on Market Interest Rates,"
Journal
of Money, Credit,
and Banking,
vol. 20, no. 2 (May 1988),
pp. 168-80; Lombra and Torto, "Discount Rate Changes and Announcement
Effects," Quarterly
Journal
of Economics,
February 1977, pp. 171-76; and
Daniel L. Thornton, "The Market's Reaction To Discount Changes: What's
Behind The Announcement Effect? Federal Reserve Bank of St. Louis,
Working Paper Series, November 1991, pp. 2-23.
47. In November 1978, reserve requirements, the discount rate, and
the funds rate target were all raised simultaneously as a dramatic
gesture to attack the rising rate of inflation and weakening exchange
value of the dollar.




-29-

Meulendyke
borrowing allowed to drop to frictional levels, leading the federal
funds rate to fall below the discount rate.
The adjustment mechanism depended heavily on the enforced
reluctance to borrow.

When banks borrowed to satisfy their reserve

requirement, they reduced their future access to the discount window.
Consequently, when the banking system as a whole had to borrow a higher
volume of reserves to meet requirements, individual banks would bid up
the federal funds rate as they tried to avoid being one of the banks
that turned to the window.

The process gave banks the message to cut

back on deposit-expanding activities.

Chart 6 gives key borrowing and

interest rate relationships during these years.
The move to the new procedures inspired discussion of the
appropriate guidelines for setting and changing the discount rate. Some
Board members initially had expected that the discount rate would be
changed more frequently than before to keep it more closely aligned with
market rates.

In practice, the basic discount rate was changed fairly

frequently--sixteen times between October 1979 and October 1982--but it
still moved much less than the funds rate. At times, unprecedented
weekly average spreads developed between the funds rate and the discount
rate.
During two periods of exceptionally restrictive provision of
nonborrowed reserves, in 1980 and again in 1981, the volume of borrowing
ran very high.

The Board introduced a surcharge on frequent borrowing

by large banks as part of the Administration's credit restraint program
in March 1980.A8

The frequency limits for access at the basic rate

were similar to those that had been proposed a decade earlier for the
basic borrowing privilege.

In addition, banks did not have unlimited

access to the discount window even when they paid the surcharge. The

48. A more detailed discussion of the rationale underlying the
program of credit restraint is given in a statement by Frederick H.
Schultz, Vice Chairman, Board of Governors of the Federal Reserve
System, before the Subcommittee on Access to Equity Capital and Business
Opportunities of the House Committee on Small Business, April 2, 1980.
It is reprinted in the Federal Reserve Bulletin,
April 1980.




-30-

Meulendyke
funds rate often exceeded even the combined basic rate and surcharge-which reached a high of 18 percent in 1981.A9
Borrowed Reserve Targeting in the 1980s and Early 1990s
Borrowed reserve targeting replaced nonborrowed reserve
targeting in 1983 as the primary guide for choosing desired reserve
levels.

The shift in emphasis removed the automatic linkage between

reserves and money targets. Borrowed reserve targeting made more formal
use of the relationship between the amount of borrowing and the spread
between the federal funds rate and the discount rate that arises from
the restrictions on heavy use of the discount window. As was the case
under the previous procedures, forcing increased borrowing tended to
lead the banks to bid up the federal funds rate relative to the discount
rate as they sought to avoid having to borrow.

Reduced borrowing

encouraged less aggressive bidding for Federal funds and the rate would
fall.

The FOMC raised borrowed reserve objectives when it wanted to

tighten policy and lowered them when it wanted to ease policy.50
Chart 7 shows key borrowing and rate relationships during these years.
A change in the discount rate was viewed as a substitute for a
change in the borrowing assumption. Whenever the discount rate was
raised or lowered, the FOMC made an explicit decision whether that
action by itself accomplished the desired policy adjustment.

On some

occasions, the amount of assumed borrowing was left unchanged so that
the average federal funds rate would be expected to rise or fall by the
same amount as the discount rate move. At other times, the borrowing
allowance was changed in a direction that lessened the impact of the
discount rate change.

For example, the FOMC would raise the borrowing

49. The surcharge was initially imposed in March 1980. It was then
removed in May of that year, only to be reimposed in September. In
1981, the surcharge underwent further changes. It was increased in May,
reduced in September, reduced again in October, and finally eliminated
in November.
50. Marvin Goodfriend, "Discount Window Borrowing, Monetary Policy,
and the Post-October 6, 1979 Federal Reserve Operating Procedure,"
Journal of Monetary Economics, September 1983, pp. 343-56, offers a
critique of that relationship and suggests that it will inevitably be
unreliable.




-31-

Meulendyke
assumption when the discount rate was lowered so that the average funds
rate would not fall by as much as the discount rate.
Increased Reluctance to Borrow in the 1980s and Early 1990s. A series
of banking crises and failures beginning in 1982 reintroduced a source
of reluctance to borrow that had largely disappeared after the 1930s.
Once again, banks became concerned that borrowing at the discount window
might be interpreted as a sign that they were so weakened financially
that they could not borrow funds from normal sources.

The concern was

especially high in 1984, when Continental Illinois National Bank
suffered a crisis of confidence, experienced runs by its large
depositors, and was forced to borrow massive amounts from the Federal
Reserve to keep operating.

Continental's experience made many other

banks more hesitant to borrow, and wider spreads of the funds rate over
the discount rate emerged for a given amount of borrowing fostered by
the Federal Reserve. As more banking crises developed and then were
resolved, the reluctance to borrow became alternately more and less
severe, but it never returned to its pre-1984 pattern.
By the fall of 1987, the borrowing relationship became
sufficiently uncertain that the Federal Reserve felt compelled to reduce
its reliance on it as a guide to policy.

Since that time, the Fed has

given greater weight to indicators of money market conditions such as
the federal funds rate. Nonetheless, the extreme reluctance to borrow
and the resulting uncertainty about how banks will respond to changing
levels of reserve availability have also introduced some volatility of
the funds rate. When banks have not wanted to borrow, they have reacted
to a reserve shortage by bidding up the funds rate to very high levels
before they finally turn to the discount window.

Indeed, on one

occasion in 1990, the funds rate reached 100 percent, a level not seen
even when interest rates and borrowing levels were routinely much higher
a decade earlier. While efforts have been made to explain to the banks
and the public that occasional borrowing is an appropriate action to
relieve temporary shortages of reserves, the message has so far had
limited impact.
The reluctance to borrow has compounded the reserve management
difficulties associated with low reserve requirements, described in the




-32-

Meulendyke

previous section.

The low requirements reduced depositories' ability to

handle normal day-to-day variation in reserve flows because the range of
reserve levels that fell between excess reserves and overdrafts
narrowed.

The extreme reluctance to borrow weakened one means for banks

to recover from an unexpected reserve shortage.
The problems that arise when borrowing and required reserves do
not behave as desired underscore the importance of these tools.

The

policy process benefits when both reserve requirements and the discount
window can play their assigned supporting roles in the monetary policy
process.

Open market operations can be hard pressed to achieve policy

goals without their help.




1. Required Reserves and Applied Vault Cash 1951 -1992*
70,000

60,000 [-

50,000
(0

c
o
I 40,000
m




30,000

20,000

10.000

0
* All figures are quarterly averages.
Note: Before December 1959, the Federal Reserve did not allow
vault cash to count towards the fulfillment of reserve requirements.

2. Required Balances and Excess Reserves
45.000
40,000
35,000
30,000
*

25,000

«/»

20.000
15,000
Required Reserve Balances
Required Reserve Balances plus Required Clearing Balances
Required Reserve Balances plus Required Clearing Balances plus Excess Reserves

10,000
5,000
0

J

i

i

i

i

1980

i

i

i

1981

j _

1982

1983

L-

I

1984

1985

1986

1987

1988

1989

1990

1984

1985

1986

1987

1988

1989

1990

_J

I

I

I

I

I

I

i

1

I

1

J

I—J

I

1 ,.J. I

I

1

1991 1992

2.000
Excess Reserves
1,500
CO

C

o

1,000
«/>




500

0
1980
1981
1982
1983
Note: All figures are quarterly averages.

1991 1992

*

o

<

'%

CD

T3
0)
U_

E

CD
0)

03

CO

O
O
CO

o
o
o
o

CO

o
o

to

suojino $

o
o
o
o

CO




4. Borrowed Reserves and Selected Interest Rates 1950-1965
Borrowed Reserves as a
Percentage of Total Reserves

1,600
1,400
1,200

Borrowed Reserves

1,000
800
600
400
200
0

N

nrflFlh

HI^I^IJWI^

4

A

3
2
1
0

\

Discount Rate

NN
NNN

y

^- ^\

" ^ X *

LUI

Effective Federal Funds Rate
(starting in 1960)
3-month new bill rate

1950 1951 1952 1953 1954 1955 1956 1957 1958 1959 1960 1961 1962 1963 1964 1965

Note: Quarterly averages except for discount rate. Discount rate
is the rate in effect on the last day of the quarter.



5. Borrowed Reserves and Selected Interest Rates 1966-1979
6

Borrowed Reserves as a
Percentage of Total Reserves

'5
c
a>

4

a.

3
2
1
0
2,500
Borrowed Reserves

2,000
CO

I 1,500
N

KJ

N

N
N

N
SrTlnNf^lrpnrTirnKTiF]

^

12
c
a>
<D
Q.

10

Effective Federal Funds Rate

8
6
4

1966
1967
1968
1969
1970
1971
1972
Note: Quarterly averages except for the discount rate. Discount rate
is the rate in effect on the last day of the quarter.



1973

1974

1975

1976

1977

1978 1979

6. Borrowed Reserves and Selected Interest Rates 1979-1982
Borrowed Reserves as a
Percentage of Total Reserves

1

o
Borrowed Reserves

2,000

51

1,500
1,000

51

500
0

I KWM ISSSH R

KWN K

NVsN K
Effective Federal Funds Rate

Discount Rate

8

79Q4 80Q1
81Q1
Note: Quarterly averages except for the discount rate. Discount rate
is the rate in effect on the last day of the quarter.



82Q1

82Q4

7. Borrowed Reserves and Selected Interest Rates 1983-1992

c
<D
O
k-

<D
Q.

(0

C

o

I
4fy

12

10
c

Effective Federal Funds Rate

8

2
<D
Q.

6
4
1983

1984

1985

1986

1987

Note: Quarterly averages except for the discount rate. Discount rate
is the rate in effect on the last day of the quarter.



1988

1989

1990

1991

1992

Meulendyke
REFERENCES
Black, Robert P.

Report

of

the Ad Hoc Subcommittee

on Reserve

Proposals,

May 13, 1966.
Board of Governors of the Federal Reserve System.
years.

Annual Report,

Board of Governors of the Federal Reserve System.
various years.

Annual Statistical

Board of Governors of the Federal Reserve System.
various months.

Federal

Board of Governors of the Federal Reserve System.

Reappraisal

Reserve

Discount

Mechanism,

various

Reserve
of the

Digest,
Bulletin,
Federal

vols. 1-3, August 1971.

Carson, Deane. "Is the Federal Reserve System Really Necessary?" Journal
Finance, vol. 19, no. 4 (December 1964), pp, 652-61.

of

Cook, Timothy, and Thomas Hahn. "The Information Content of Discount Rate
Announcements and Their Effect on Market Interest Rates," Journal of
Money, Credit,
and Banking, vol. 20, no. 2 (May 1988), pp. 167-80.
Feinman, Joshua. "Estimating the Open Market Desk's Daily Reaction
Function." Federal Reserve Board, Division of Monetary Affairs,
Washington, D.C., August 1991.
Friedman, Milton. A Program for Monetary
University Press, 1959.

Stability.

New York: Fordham

Froyen, Richard T. "The Determinants of Federal Reserve Discount Rate
Policy," Southern Economic Journal, vol. 42, no. 2 (October 1975),
pp. 193-200.
Goodfriend, Marvin. "Discount Window Borrowing, Monetary Policy, and the
Post-October 6, 1979 Federal Reserve Operating Procedure," Journal of
Monetary Economics, September 1983, pp. 343-56.
Goodfriend, Marvin, and Monica Hargraves. "A Historical Assessment of the
Rationales and Functions of Reserve Requirements," Federal Reserve Bank
of Richmond, Economic Review, vol. 69, no. 2 (March/April 1983), pp.3-21.
Gramley, Lyle E., and Samuel B. Chase Jr. "Time Deposits in Monetary
Analysis," Federal Reserve Bulletin,
vol. 51, no. 10 (October 1965),
pp. 1380-1404.
Hall, Robert E.

"A Free Market Policy to Stabilize the Purchasing Power of

the Dollar," in Barry Seigel, ed., Money in Crisis:
The Federal
Reserve,
the Economy, and Monetary
Reform,
Pacific Studies in Public Policy.

Cambridge Massachusetts: Ballinger,




1984, pp. 303-21.

Meulendyke
Heller, Walter W. Report of the Committee on Financial
President
of the United States,
April 1963.

Institutions

Jones, David. "Excess Reserves under MCA," internal memorandum.
Governors of the Federal Reserve System, November 10, 1983.

to

the

Board of

Lombra, Raymond E. , and Raymond G. Torto. "Discount Rate Changes and
Announcement Effects," Quarterly Journal of Economics, vol. 91, no. 1
(February 1977), pp. 171-76.
Mayer, Thomas.
1968.

Monetary Policy

in the United

States.

New York: Random House,

Mayer, Thomas, James S. Duesenberry, and Robert Z. Aliber. Money, Banking,
and The Economy, 3d ed. New York: W.W. Norton and Company, 1987.
Meigs, A. James. Free Reserves
Chicago Press, 1962.

and the Money Supply.

Chicago: University of

Meulendyke, Ann-Marie. "A Review of Federal Reserve Policy Targets and
Operating Guides in Recent Decades," in Intermediate
Targets and
Indicators
for Monetary Policy:
A Critical
Survey.
Federal Reserve Bank
of New York, July 1990. Reprinted from Federal Reserve Bank of New York,
Quarterly Review, vol. 13, no. 3 (Autumn 1988), pp. 6-17.
Meulendyke, Ann-Marie. "Monetary Policy Implementation and Reserve
Requirements," internal working paper. Federal Reserve Bank of New York,
July 1992.
Meulendyke, Ann-Marie. U.S. Monetary Policy
Reserve Bank of New York, 1990.

and Financial

Markets.

Federal

Miller, William G. "Proposals on Financial Institution Reserve Requirements
and Related Issues," testimony before the House Committee on Banking,
Finance and Urban Affairs, July 27, 1978.
"Monetary Policy and Open Market Operations during 1990," Federal Reserve
Bank of New York, Quarterly Review, vol. 16, no. 1 (Spring 1991),
pp. 66-74.
"Monetary Policy and Open Market Operations during 1991," Federal Reserve
Bank of New York, Quarterly Review, vol. 17, no. 1 (Spring 1992),
pp. 72-95.
Partee, J. Charles. "Statement before the Subcommittee on Financial
Institutions Supervision, Regulation and Insurance of the House Committee
on Banking, Finance and Urban Affairs," Federal Reserve
Bulletin,
November 1983, pp. 850-51.




-2-

Meulendyke
Porter, Richard D., and Kenneth J. Kopecky. "The Role of Reserve Requirements
as a Public Policy Tool," Conference on Reserve Requirements and the
Role of the Federal Reserve System, Washington, D.C., January 18-19,
1979, pp. 863-80.
Schultz, Frederick H. "Statement before the Subcommittee on Access to Equity
Capital and Business Opportunities of the House Committee on Small
Business," Federal Reserve Bulletin,
vol. 66, no. 4 (April 1980),
pp. 309-11.
Sellon, Gorden H. Jr. "The Instruments of Monetary Policy," Federal Reserve
Bank of Kansas City, Economic Review, vol. 69, no. 5 (May 1984),
pp. 3-20.
Small, David, and Brian Madigan. "An Analysis of Excess Reserves," internal
memorandum. Board of Governors of the Federal Reserve System, July 1,
1986.
Smith, Warren L. "The Discount Rate as a Credit-Control Weapon," Journal
Political
Economy, vol. 66, no. 2 (April 1958), pp. 171-77.
System Committee on the Discount and Discount Rate Mechanism.
Discount Mechanism," March 12, 1954.

of

"Report on the

Thornton, Daniel L. "The Market's Reaction To Discount Changes: What's Behind
the Announcement Effect?"
Federal Reserve Bank of St. Louis, Working
Paper Series, November 1991.
Tolley, George S. "Providing for Growth of the Money Supply," Journal
Political
Economy, vol. 65, no. 6 (December 1957), pp. 465-85.

of

Waud, Roger N. "Public Interpretation of Federal Reserve Discount Rate
Changes: Evidence on the 'Announcement Effect,'" Econometricaf
vol. 38,
no. 2 (March 1970), pp. 231-50.
Weiner, Stuart E. "Payment of Interest on Reserves," Federal Reserve Bank of
Kansas City, Economic Review, vol. 70, no. 1 (January 1985), pp. 20-21.




-3-

A Note on Theories of Money Stock Determination
Robert L. Hetzel1
The purpose of this brief review of theories of money stock
determination is to encourage economists to once again work on such models.
The current lack of interest in them probably explains the irrelevance of
textbook discussions to the actual monetary arrangements that determine the
money stock.
A discussion of money stock determination needs to make explicit whether
the central bank is using an interest rate or the quantity of reserves as its
policy instrument.

This choice possesses different implications for the way

in which the central bank gives the money stock and the price level welldefined equilibrium values.

It also yields different implications for which

of the behavioral relationships of the public are key for determining the
money stock.

For example, textbook discussions, which do not clearly identify

the policy instrument, confuse the roles of the "three tools" of monetary
policy: open market operations, reserve requirements, and the discount window.
For example, textbooks do not mention that in the 1970s when the Fed targeted
the funds rate directly, changes in required reserves ratios and in the
discount rate had no first-order effects on the money stock.

EARLY BANK-RATE THEORIES
Henry Thornton in his book Paper Credit (1802) and in speeches before
Parliament (1811) formulated the first theory of money stock determination
with rate targeting by the central bank.

(See Hetzel 1987.)

Thornton's model

is in the quantity theory tradition, which explains the determination of the

The author is Vice President and Economist at the Federal Reserve Bank
of Richmond.




Hetzel
price level through the interaction of a money supply and money demand
function.

The key element of his theory is that the public cares only about

real variables, that is, real quantities and relative prices.

In particular,

the real rate of interest is only transitorily affected by changes in fiat
money creation.
hypothesis.)

(This idea is now referred to as the natural rate

Money creation allows the central bank to set the market rate at

a different value than the real rate adjusted for expected inflation, but only
transitorily.

The flip side of the natural rate hypothesis is that only the

central bank can give nominal variables like the money stock and the price
level well-defined values.

Thornton argued that the Bank of England, despite

its real-bills rhetoric, gave nominal variables determinate values by
targeting the exchange rate.
Thomas Joplin (1823) gave Thornton's idea of a natural rate of interest
its modern meaning as the real rate of interest that equates saving and
investment.

With the Resumption Act of 1819 that returned Britain to the gold

standard, the idea of a central bank that creates fiat money virtually
disappeared.

Joplin thought that the banking system, through variation in its

reserves-deposits ratio, created changes in money that caused transitory
divergences in the market and the natural rate.

This idea reappeared later in

the work of Knut Wicksell (1898) and Irving Fisher (1918).
With the supremacy of the gold standard in the nineteenth century, the
idea of central bank money creation practically disappeared and, along with
it, the idea of a natural rate of interest.

David Hume's price-specie flow

mechanism became the basic model of money stock determination in the
nineteenth century.




Wicksell was unique in reinventing the idea of a natural
2

Hetzel
rate of interest, but he had no central bank in his model.

He also made no

distinction between real and nominal variables.
Abandonment of the gold standard in World War I, as in the Napoleonic
Wars, led to the reemergence of theories of fiat money creation.

Gustav

Cassel (1928) developed the market rate-natural rate theory of Thornton and
Wicksell.

He pointed out that achieving price level determinacy required more

than equality of the market rate and the natural rate.

An infinite number of

price levels are consistent with equality of these two rates.

Cassel argued

that the central bank should vary its discount rate in order to keep the price
level at a targeted value.

Interest in theories of money stock determination

in which the central bank uses an interest rate as its policy variable
disappeared in the Depression with the prevalence of elasticity pessimism.

THE RESERVES-MONEY MULTIPLIER THEORY
The reserves-money multiplier theory emerged at the end of World War I.
It had been advanced occasionally in the nineteen hundreds, but had never
caught on (Humphrey 1987).
United Kingdom.

Pigou (1917) and Keynes (1923) advanced it in the

In the United States, Phillips (1921) built up the reserves-

money multiplier formula from a deposit expansion process whereby a reserve
injection creates deposits as it passes from bank to bank.

This deposit

creation process continues until the newly-created reserves are absorbed into
required reserves.
The Phillips' description of the deposits creation process was flawed
from the beginning.

Even if changes in aggregate reserves are exogenous, bank

deposit creation is constrained by the interest rate on reserves in the




3

Hetzel
interbank market for reserves, not the quantity of reserves a bank holds.

By

the 1920s, there was a Fed funds and a call money market that allowed banks to
buy and sell reserves.

Quantity theorists, however, liked the idea of the

reserves-deposits expansion process because it provided an easy refutation to
real bills proponents who argued that banks cannot create deposits.

Quantity

theorists could use the Phillips7 story to argue that real bills proponents
failed to understand that what is true for the individual bank is not
necessarily true for the banking system.
The revival of models of money stock determination in the 1950s centered
on reserves-money multiplier models.

Why did quantity theorists turn to these

models given that the Fed was targeting free reserves, which is an indirect
procedure for targeting the interest rate?

In the 1950s, economists had not

yet rediscovered the natural rate hypothesis and the idea of a natural rate of
interest.

Without these concepts, quantity theorists could not model money

stock determination with central bank rate targeting in a way that made the
money supply function differ from the money demand function by depending
crucially on central bank behavior.

The reserves-money multiplier theory

offered an easy explanation of how central banks control the price level by
controlling the supply of money.
In the 1970s, Fed economists working on a monthly model for use at FOMC
meetings rejected the reserves-money multiplier framework.
Pierce and Parry 1975 and Davis 1974.)

(See Thomson,

Given the Fed's target for the funds

rate, they viewed the money stock as being demand determined by the public's
demand for money function.

The problem with this view was that the price

level was taken as determined outside the model.




4

The model then did not

Hetzel
distinguish between the determination of nominal and real money.

With the

price level exogenously determined, the Fed can vary the real quantity of
reserves to control the (market and real) rate of interest and the nominal and
real quantity of money demanded by the public.

If the price level is

endogenously determined, however, these models say nothing about the nominal
quantity of money.

While there is a determinate relationship between the

interest rate and the real quantity of money, there is no determinate
relationship between the interest rate and the nominal quantity of money.

RATIONAL EXPECTATIONS MODELS WITH RATE TARGETING
The key conceptual issue in models of money stock determination with
rate targeting by the central bank is how nominal variables are rendered
determinate, that is, given a well-defined equilibrium value?

Patinkin (1965)

pointed out that the central bank must "concern" itself with some nominal
variable.

In his model, that nominal variable is bank reserves.

The price

level is then rendered determinate through a real balance effect.

Because an

arbitrary rise in the price level reduces the real value of a variable the
public cares about, bank reserves and money, it causes the public to reduce
its real expenditure, and the price level is returned to its equilibrium
value.

When the central bank targets an interest rate, however, all nominal

variables including bank reserves are determined endogenously.

There is no

real balance effect.
In the early 1980s, a number of economists figured out how to explain
nominal determinacy with interest rate targeting by the central bank.

The key

papers were by Dotsey and King (1983); Canzoneri, Henderson and Rogoff (1983);




5

Hetzel
and McCallum (1981, 1986).

Their models embodied the natural rate hypothesis

so that only the central bank, not the public, could give nominal variables a
well-defined equilibrium value.

The initial models achieved nominal

determinacy by causing the central bank to behave in such a way that the
public's expectation of the future price level remained fixed.

Where Patinkin

fixed reserves, these models fixed the expected future value of the price
level.

They did so by not allowing base drift in money.
Goodfriend (1986) brought these models closer to actual central bank

behavior by allowing the public's expectation of the future price level to
vary in response to macroeconomic shocks.

He did so by allowing base drift in

the money stock where the amount of such drift depends upon the extent to
which the central bank desires to smooth interest rates.

Goodfriend gave the

central bank a cost function that made it averse to large "jumps" in the price
level relative to expectations.

By making the central bank care both about

the difference between the contemporaneous price level and the prior period's
expectation of the contemporaneous price level and about the difference
between the contemporaneous price level and the expected future price level,
he imposed a level and a change constraint that made the public's expectation
of the future price level well defined, while still allowing that expectation
to change in response macroeconomic shocks.
Instead of a real balance effect, these models make use of a relative
price effect.

Given the public's expectation of the future price level, an

arbitrary change in the contemporaneous price level changes the interest rate
by changing expected inflation.

Changes in the interest rate then affect the

demand for money and the reserves-supplying behavior of the central bank in a




6

Hetzel
way that returns the price level to an equilibrium value.

The simplifying

assumption that makes these models tractable analytically is rational
expectations.

This assumption allowed expectations to be determined in a

simple enough way that the models could highlight how the central bank makes
nominal values well defined by the control it exercises over the way the
public forms its expectation of the future values of nominal variables.

AN EXAMPLE AND SOME BASIC PRINCIPLES
This section draws on the models briefly described above to highlight
some of the basic concepts in a theory of money stock determination.

Consider

an example where the central bank implements monetary policy by setting an
interest rate.

The natural rate hypothesis implies that the central bank

cannot set its interest rate target arbitrarily.

It must have procedures that

allow it to set its rate target in a way that tracks on average the economy's
equilibrium interest rate.

The central bank, however, is assumed to smooth

changes in the economy's equilibrium interest rate by supplying reserves when
market rates rise and withdrawing reserves when market rates fall.

Changes in

reserves and the money stock then emerge in response to the macroeconomic
shocks that impinge upon the economy.
It is also necessary to make some assumption about the central bank's
subsequent behavior toward the random changes in money introduced by rate
smoothing.

The central bank can either offset these changes subsequently, in

part or in full, or incorporate them permanently into the level of the money
stock.

In practice, central banks follow the latter "let bygones-be-bygones"

policy of base drift in money and prices.




7

Hetzel
Assume that a persistent real shock raises the equilibrium real rate of
interest.

For example, a new technology leads to increased investment.

When

the market rate rises initially, the central bank buys government securities.
The monetary base and the money stock increase.

Because the real rate of

interest is ultimately determined solely by real factors like investment
opportunities and the public's thrift, the real rate of interest must
eventually rise to a higher equilibrium value that is independent of the
increase in money.

Similarly, the real quantity of money desired by the

public will ultimately be unaffected by the actions of the central bank.

The

rise in market rates will make currency and bank reserves more costly to hold,
but the return on bank deposits that pay interest will rise.

The equilibrium

real quantity of money may then either decrease or increase.

In either event,

the supply of money changes in a way that is largely unrelated to any change
in the public's demand for real money.

For this reason, it will be convenient

to assume that the rise in market rates leaves the real quantity of money
demanded by the public unchanged.
Ultimately, the change in the nominal quantity of money will not affect
any of the new equilibrium values of the real variables, the real rate of
interest and the real quantity of money.

At the original price level,

however, the public is now holding a larger quantity of real money balances.
The price level must rise to return real money balances to their lower,
equilibrium value.

This example can be used to elucidate a number of key

concepts in a theory of money stock determination. 2
2

For a somewhat different treatment, see any of the expositions by
Milton Friedman that feature a helicopter drop of money, for example, Friedman
(1992, Chapter 2 ) .




8

Hetzel
First, the example illustrates the distinction between nominal and real
variables.

In the short run, macroeconomic shocks originating in the private

sector produce changes in the nominal money stock.

In the long run, however,

the central bank exercises complete control over the nominal money stock.

The

central bank determines the amount of base money to create in response to such
shocks.

It also determines the extent to which the money created in response

to shocks will affect permanently the level of the money stock.
In contrast, the public largely determines the real quantity of money.
The qualification "largely determines" is added in recognition that the
central bank can indirectly influence the real quantity of money in that a
higher rate of inflation increases the cost of holding real money.
of the example, however, is unaffected by this qualification.

The result

The increase in

the stock of money does not ultimately increase the real quantity of money.
Second, it is important to keep separate the different popular meanings
of the word "money."

A theory of money stock determination concerns the

quantity of money, defined as some monetary aggregate like the monetary base,
Ml or M2.

Money is often also used to mean credit.

In this example, the

increase in investment demand and higher real rate of interest will increase
real saving and credit.

Money is also often used to mean income.

In the

example, real income is probably largely unchanged, although the composition
of output generating income will change to include more investment and less
consumption.

In the example, the real quantity of money, real credit, and

real income can all behave differently.
Third, the example illustrates the quantity theory approach to analyzing
the determination of the price level through the interaction of a money demand




9

Hetzel
and a money supply function.

In the example, the money supply function shifts

as the central bank buys government securities in response to a rise in
interest rates.

Because no corresponding shift in the money demand function

occurs, the price level rises.
The nominal money demand function is the product of the price level and
the real quantity of money desired by the public.

The nominal money supply

function depends upon the reserve-supplying behavior of the central bank.

In

the short run, shifts in this function depend upon the extent to which the
central bank smooths the interest rate, that is, the extent to which it
supplies reserves when the interest rate changes.

In the long run, shifts in

the money supply function depend upon the extent of base drift, that is, the
extent to which, if at all, the central bank subsequently offsets the changes
in money produced by changes in the interest rate.

Finally, shifts in this

function depend upon the trend rate of growth of money and inflation the
central bank accepts and the public expects.

A consequence of the natural

rate hypothesis is that only the central bank can determine the trend rate of
growth of money and prices.
Milton Friedman gave the quantity theory a particular empirical
expression by arguing that shifts in the money supply function have
historically been large relative to shifts in the money demand function.

For

this reason, over long periods of time, the price level and the nominal
quantity of money move together.

Friedman summarizes this view by saying that

inflation is everywhere and always a monetary phenomenon.

Note, however, that

the analytical usefulness of the quantity theory only requires that
unpredictable changes in money demand are small relative to shifts in the




10

Hetzel
money supply function and to predictable shifts in money demand.
Equivalently, unpredictable changes in money demand should be small relative
to changes in nominal expenditure or output.
Note that with rate targeting the key behavioral relationships of the
money supply function are not the reserves-currency and reserves-deposits
ratios discussed in textbooks.

Fluctuations in these ratios are automatically

offset at the prevailing funds rate target.

For example, if currency flows

out of banks or if banks increase the desired level of excess reserves, the
funds rate rises.

In order to maintain its funds rate target, the central

bank supplies reserves, thereby accommodating changes in these ratios and
avoiding a change in bank deposits.
Fourth, what appears true for the individual is not necessarily true for
individuals collectively.

In the example, after the increase in the money

stock, individuals believe they can reduce their money holdings to a desired
lower level.

The public, however, cannot reduce its nominal money holdings.

The individuals who sold government securities to the central bank did so
because they were offered a good price, not because they wanted to reduce
their holdings of assets.

After selling securities to the central bank,

individuals allocate their increased cash among different assets to replace
the securities they sold.

Temporarily, the increased demand for assets keeps

the interest rate below its new, higher equilibrium rate.

As a consequence,

real expenditure rises until the price level increases sufficiently to return
real money balances to their original level.

The interest rate then rises to

its new, higher equilibrium value.
More generally, economic fallacies often arise out of inappropriate




11

Hetzel
generalization of individual experience.

To the individual, it appears that

the cause of inflation is the rise in prices of individual commodities.

The

cause of inflation then is sought for in the determinants of the relative
prices of individual commodities, rather than in the behavior of money.
Fifth, the central bank must ensure that the price level and money stock
possess equilibrium values.

The central bank, however, must do more than

simply set an interest rate target that is consistent with the economy's
natural rate of interest (augmented by expected inflation).

At the central

bank's prevailing rate target, an arbitrary perturbation in the price level
will produce a corresponding change in the demand for bank credit and in bank
deposits and money.
together.

All nominal magnitudes can then wander off aimlessly

The central bank must keep some nominal value steady.

How do

central banks impart this nominal steadiness to equilibrium values?
Central banks dislike "large, unusual jumps" in nominal prices. 3

This

concern imparts an "inertia" to the public's expectation of the future price
level.

Arbitrary changes in the contemporaneous price level, therefore,

produce changes in the contemporaneous price level relative to the expected
future price level.

These changes create a relative price effect analogous to

the real balance effect as the mechanism for eliminating arbitrary changes in
the price level.

Specifically, given an arbitrary change in the price level,

some real or relative variable must change to produce an inverse change in the
public's real expenditure.

That change is the change in the price level

relative to the expected future price level, which produces an inverse

3

This statement is given content by defining jumps relative to expected
values. See the discussion of Goodfriend (1987) above.




12

Hetzel
movement in the interest rate.

The movement in the interest rate affects the

reserve supplying behavior of the Fed and the public's demand for money in a
way that returns the price level to its equilibrium value.
This relative price effect can be explained by analogy.

Consider how

nominal determinacy is achieved when the central bank of a country targets its
exchange rate with another country.

For the sake of argument, assume that the

Fed targets the Deutsche Mark price of a dollar.

As shown in equation (1),

the DM/$ exchange rate equals the product of the ratio of the German price
level (DM/German good) to the U.S. price level (S/US good) and the real terms
of trade (German good/US good).
German price level.

The nominal benchmark for the dollar is the

If the U.S. price level rises arbitrarily, the foreign

exchange value of the dollar falls, and the Fed buys dollars with Deutsche
marks.

The monetary base and the money stock fall and the price level returns

to its equilibrium level.
DM
...

DM

(l)

—

German good

=

$

German good

•
$

US good

US good

In the case of an interest rate target, the Fed targets the price of
today's dollars ($t) in terms of tomorrow's dollars ($ t + 1 ), or one plus the
interest rate.

As shown in equation (2), this price equals the product of the

ratio of the expected future price level to the contemporaneous price level
and the real terms of trade with the future.

With a rate target, the nominal

benchmark is the expected future price level.

Now, an arbitrary rise in the

contemporaneous price level produces a fall in the ratio of the expected
future price level to the contemporaneous price level, the first factor on the




13

Hetzel
right side of (2). The fall in this ratio produces a decline in the market
rate of interest by reducing the inflation premium.

A decline in the market

rate of interest produces an increase in the demand for money.
prompts the central bank to sell securities.

It also

The demand for money increases,

while the monetary base and the money stock fall, and the price level returns
to its equilibrium level.

S^

(2)

$f

tK
=

good)f"

(-L.)

9

(good),.,

(good),

good '
Sixth, the public is forward looking.

It must form an expectation of

the future price level in order to determine a market rate of interest.

In

making saving and investment decisions, individuals care about the price of
today's goods in terms of tomorrow's goods.

They contract, however, in terms

of the price of today's dollars in terms of tomorrow's dollars.

Individuals

must, therefore, form an expectation of the future purchasing power of the
dollar.

The central bank determines how the public forms that expectation.

Consider again the example of the real shock with rate smoothing by the
central bank that causes money and prices to increase.

If the public expects

that the central bank will reverse the increase in money and prices in the
future, then the public will expect a subsequent fall in prices, or at least a
temporary reduction in inflation relative to trend.

A temporary reduction in

the inflation premium will for a while moderate the rise in the market rate
produced by the rise in the equilibrium real rate.

(This situation probably

obtained in World War II.)
Alternatively, if the public expects the central bank to incorporate




14

Hetzel
each period's random change in money and prices permanently into their future
levels, the interest rate will rise immediately by the amount of increase in
the equilibrium real rate (apart from a temporary liquidity effect).

The

public might even expect that the central bank will allow the rate of
inflation to rise permanently, in which case the market rate will rise by more
than the increase in the real rate.

In this case, the price level will also

rise by more than the increase in money.
As this discussion illustrates, the response of the public to today's
action of the central bank depends upon what the public expects the central
bank to do tomorrow.

(The public may not actually watch the actions of the

central bank, but it will respond to contemporaneous changes in the price
level in a way that is consistent with the behavior of the price level that
the central bank has produced over time.)

For this reason, at least since the

1970s, economists have generally formulated their recommendations for the
central bank as strategies to be maintained over time, rather than as
particular policy actions.

The idea is that the policymaker can predict the

consequences of a policy action taken as part of a known strategy because he
has some basis for predicting what the public anticipates in the way of
subsequent policy actions (Lucas 1975).
Seventh, the example involves both monetary policy and fiscal policy.
The monetary policy action undertaken by the central bank is the increase in
the monetary base.

Monetary policy, that is, the systematic behavior of the

central bank, is described by the extent to which the central bank changes the
monetary base when interest rates change and the extent to which such changes
become permanently incorporated into the level of the monetary base.




15

The

Hetzel
fiscal policy side of the central bank's action is the reduction of the
government debt held by the public due to the purchase of the government
security.

The revenue the government must collect in the future to pay off

its debt falls.

Taxpayers can increase their consumption.

Who pays for this windfall to taxpayers?

When the price level rises,

whoever holds money must add to his money holdings in order to maintain their
real purchasing power.

The money holder must refrain from consumption while

restoring the real value of his cash balances to their former level.

(The

additional dollars held are like receipts showing payment of the tax.)

There

is a wealth transfer from holders of cash balances to taxpayers in general.
Inflation is a tax levied on whoever holds money.
Pressure for inflation comes from confusion between money and wealth
creation.

This confusion turns on ignorance of all the principles listed

above: the difference between money and credit or income; the difference
between nominal and real money; the fallacy of generalizing on the basis of
particular examples; failure to understand the central responsibility of the
central bank for the behavior of the price level; and failure to realize that
the public is forward looking in the way it forms its expectations of central
bank behavior.

Pressure for inflation, however, also comes from the fact that

while money creation does not augment wealth, it can redistribute it.

A

central fact of the political economy of money creation, and of its eternal
appeal, is that the tax money creation imposes does not have to be explicitly
legislated.




16

Hetzel
REFERENCES
Canzoneri, Matthew B., Dale W. Henderson and Kenneth S. Rogoff, "The
Information Content of the Interest Rate and Optimal Monetary Policy,"
The Quarterly Journal of Economics, November 1983, 545-65.
Cassel, Gustav, "The Rate of Interest, The Bank Rate, and the Stabilization
of Prices," Jhe Quarterly Journal of Economics, August 1928, 511-29.
Davis, Richard G., "Implementing Open Market Policy with Monetary Aggregate
Objectives," in Federal Reserve Bank of New York Monetary Aggregates and
Monetary Policy, October 1974.
Dotsey, Michael and Robert G. King, "Monetary Instruments and Policy Rules in
a Rational Expectations Environment," Journal of Monetary Economics,
September 1983, 357-82.
Fisher, Irving, Elementary Principles of Economics, New York: The Macmillan
Co., 1918.
Friedman, Milton, Money Mischief, New York: Harcourt Brace Jovanovich, 1992.
Goodfriend, Marvin, "Interest Rate Smoothing and Price Level TrendStationarity," Journal of Monetary Economics, May 1987, 335-48.
Hetzel, Robert L., "Henry Thornton: Seminal Monetary Theorist and Father of
the Modern Central Bank," Federal Reserve Bank of Richmond Economic
Review, July/August 1987, 3-16.
Humphrey, Thomas M. "The Theory of Multiple Expansion of Deposits: What It
Is and Whence It Came," Federal Reserve Bank of Richmond Economic
Review, March/April 1987, 3-10.
Joplin, Thomas. Outlines of a System of Political Economy (1823), New York:
Augustus M. Kelley, 1970.
Keynes, John Maynard, A Tract on Monetary Reform (1923), in lire Collected
Writings of John Maynard Keynes, vol. 4, London: The Macmillan Press,
1971.
Lucas, Robert E., "Econometric Policy Evaluation: A Critique," in The
Phillips Curve and Labor Markets, edited by Karl Brunner and Allan
Meltzer, pp. 19-46. Carnegie-Rochester Conference Series no. 1. New
York: North Holland, 1975.
McCallum, Bennett T., "Price Level Determinacy with an Interest Rate Policy
Rule and Rational Expectations," Journal of Monetary Economics, November
1981, 319-29.
, "Some Issues Concerning Interest Rate Pegging, Price Level Determinacy,




17

Hetzel
and the Real Bills Doctrine," Journal of Monetary Economics, January
1986, 135-60.
Pigou. A. C. "The Value of Money," Quarterly Journal of Economics, November
1917, 38-65.
Phillips, Chester A. Bank Credit, New York: The Macmillan Co., 1921.
Thomson, Thomas D., James L. Pierce, and Robert T. Parry, "A Monthly Money
Market Model," Journal of Money Credit and Banking, November 1975, 41131.
Thornton, Henry, An Enquiry into the Nature and Effects of the Paper Credit
of Great Britain (1802) and Iwo Speeches (1811), ed., F. A. v. Hayek,
New York: Rinehart and Co., 1939.
Wicksel1, Knut, Interest and Prices (1898), New York: Augustus M. Kelley,
1965.




18

INTEREST RATE POLICY AND THE
INFLATION SCARE PROBLEM: 1979-1992
Marvin Goodfriend1
U.S. monetary policy since the late 1970s is unique in the post-war era
in that rising inflation has been reversed and stabilized at a lower
rate for almost a decade. The inflation rate of 3 to 4% per year,
representing a reduction of 6% or so from its 1981 peak, is the result
of a disinflationary effort that has been long and difficult.
This paper analyzes the disinflation by reviewing the interaction
between Federal Reserve policy actions and economic variables such as
the long-term bond rate, real GDP growth, and inflation.

The period

breaks naturally into a number of phases, with the broad contour of
events as follows. A period of rising inflation was followed by
disinflation which, strictly speaking, was largely completed in 1983
when inflation stabilized at around 4% per year.

But there were two

more "inflation scares" later in the decade when rising long-term rates
reflected expectations that the Fed might once more allow inflation to
rise.

Confidence in the Fed was still relatively low in 1983, but the

central bank has acquired more credibility since then by successfully
resisting the inflation scares.

1. The author is Vice President and Associate Director of Research
at the Federal Reserve Bank of Richmond. The paper has benefitted greatly
from discussions with Timothy Cook and Robert King, and from presentations
at the 1992 NBER Summer Institute and the Board of Governors of the Federal
Reserve System. Comments by John Boschen and George Moore were also very
helpful.




Goodfriend

I analyze the conduct of monetary policy using a narrative
approach that pays close attention to monthly movements of long and
short-term interest rates. My approach is intended to complement
existing studies such as the VAR-based analyses by Bernanke and Blinder
(1992) and Sims (1991), and the more conventional studies of the period
by Friedman (1988) and Poole (1988).

The goal is to distill

observations to guide future empirical and theoretical analysis of
monetary policy with the ultimate objective of improving macroeconomic
performance.

Based on a familiarity with the Fed over this period and

the work of Fed economists, I interpret policy actions in terms of the
Federal funds rate rather than a measure of money.

I view the paper as

a case study of the Federal Reserve's interest rate policy.
The Fed's primary policy problem during the period under study
was the acquisition and maintenance of credibility for its commitment to
low inflation.2

I measure credibility by movements of inflation

expectations reflected in the long term interest rate.

For much of the

period the Fed's policy actions were directed at resisting inflation
scares signalled by large sustained increases in the long rate.

A scare

could take well over a year of high real short-term interest rates to
contain.

Moreover, just the threat of a scare appears to have made the

Fed tighten aggressively in one instance and probably made it more
cautious when pushing the funds rate down to encourage real growth on a
number of occasions.

2. See Rogoff (1987) for a theoretical survey of credibility,
reputation, and monetary policy.




- 2 -

Goodfriend

Inflation scares are costly because resisting them requires the
Fed to raise real short rates with potentially depressing effects on
business conditions.

Hesitating to react is also costly, however,

because by revealing its indifference to higher inflation the Fed
actually encourages workers and firms to ask for wage and price
increases to protect themselves from higher expected costs.

The Fed is

then inclined to accommodate the higher inflation with faster money
growth.
Inflation scares present the Fed with a fundamental dilemma whose
resolution has decided the course of monetary policy in the post-war
Prior to the 1980s, the Fed generated an upward trend in the

period.

inflation rate by reacting to inflation scares with a delay.

The more

prompt and even preemptive reactions since the late 1970s have been a
hallmark of the recent disinflationary era.
The plan of the paper is as follows.

First, I introduce and

discuss the premises that underlie my interpretation of monetary policy
in the body of the paper.
presented next.

The chronological analysis of policy is

Finally, I summarize the main empirical findings in a

series of observations chosen to sharpen further our interpretation and
evaluation of the conduct of monetary policy.

A brief conclusion

follows.

PREMISES UNDERLYING THE INTERPRETATION OF POLICY
The first step in any study of monetary history is to choose an
indicator of the stance of policy.




For example, in their study of U.S.

- 3 -

Goodfriend

monetary history Friedman and Schwartz (1963) focus on the monetary base
because it summarizes monetary conditions whether or not a country is on
the gold standard and whether or not it has a central bank.

Focusing on

the base allowed them to tie together a long period marked by many
institutional changes, making possible their famous empirical findings
about money, prices, and business conditions*
For my purposes, however, the base is not a good choice of
indicator.

Although the Fed could have used the base as its instrument

by controlling it closely in the short-run, it has not chosen to do so.
Instead, the Fed has chosen to use the Federal funds rate as its policy
instrument.

Hence this study, which seeks to investigate the short-run

interactions between Fed policy and other economic variables, interprets
policy actions as changes in the Federal funds rate. The remainder of
this section discusses the premises underlying my interpretation of
policy.

Interest Rate Targeting
Throughout its history the Fed's policy instrument has been the Federal
funds rate or its equivalent.

At times, notably from the mid to late

1970s, it has targeted the funds rate in a narrow band commonly 25 basis
points wide (Cook and Hahn 1989).

More often, it has targeted the funds

rate indirectly, using the discount rate and borrowed reserve targets.
Although the funds rate appears noisier under borrowed reserye

targeting

than under direct funds rate targeting, it is nevertheless tied
relatively closely to a chosen Federal funds rate target (Goodfriend




- 4 -

Goodfriend

1983).

Since a borrowing target tends to be associated with a

particular spread between the funds rate and the discount rate,
targeting borrowed reserves lets a discount rate adjustment feed through
one-for-one to the funds rate.

Forcing banks to borrow more reserves at

a given discount rate also raises the funds rate (Goodfriend and
Whelpley 1986).

The Fed has used the borrowed reserve procedure to help

manage the funds rate since it ended its experiment with nonborrowed
reserve targeting in October 1982 (Wallich 1984, Thornton 1988).
Significant Federal funds rate movements since then should be viewed as
deliberate target changes.
It is less obvious that Federal funds rate changes in the period
of the New Operating Procedures from October 1979 to October 1982 should
be interpreted as deliberate.

Under those procedures, the Fed was to

fix the path of nonborrowed reserves available to depository
institutions so that increases in the money stock would force banks to
borrow more reserves at the discount window and thereby automatically
drive up the funds rate and other short term interest rates.
Despite the widespread emphasis on automatic adjustment in the
description of the post-October 1979 procedures, however, it was wellrecognized at the time that movements in the funds rate would also
result from purely judgmental actions of the Federal Reserve (Levin and
Meek 1981, Annual Reports of Open Market Operations 1981-83).

These

actions included: (1) judgmental adjustments to the nonborrowed reserve
path taken at FOMC meetings that changed the initially expected reserves
banks would be forced to borrow at the discount window (in effect, a




- 5 -

Goodfriend

funds rate target change by the FOMC), (2) judgmental adjustments to the
nonborrowed reserve

path between FOMC meetings, (3) changes in the

discount rate, and (4) changes in the surcharge that at times during the
period was added to the basic discount rate charged to large banks.
Cook (1989) presents a detailed breakdown of policy actions
affecting the funds rate during this period showing that two-thirds of
the funds rate movement was due to judgmental actions of the Fed and
only one-third resulted from automatic adjustment. Moreover, as we
shall see below, the large Federal funds rate movements in the
nonborrowed reserve targeting period are overwhelmingly attributable to
deliberate discretionary actions taken by the Fed to manage short-term
interest rates.

In light of this, it is more accurate to refer to the

period from October 1979 to October 1982 as one of aggressive Federal
funds rate targeting than one of nonborrowed reserve targeting.

The Role of Money
The Federal Reserve was established with a mandate to cushion short-term
interest rates from liquidity disturbances. Between the Civil War and
the creation of the Fed, such disturbances caused short rates to rise
suddenly and sharply from time to time. While generally trading in a
range between 4 and 7 percent, the monthly average call loan rate
reported by Macaulay (1938) rose roughly 5 percentage points in one
month on 26 occasions between 1865 and 1914. Moreover, as a result of
banking crises, sudden changes of over 10 percentage points occurred 8
times during the same period.




These episodes were distinctly temporary,

- 6 -

Goodfriend

ranging from one to four months, with many lasting for no more than one
month.

Such extreme temporary spikes are absent from interest rates

since the founding of the Fed (Miron 1986, Mankiw, Miron, and Weil
1987).
In line with its original mandate, the Fed has routinely
accommodated liquidity disturbances at a given targeted level of shortterm interest rates.

Furthermore, by giving banks access to the

discount window the Fed has been careful not to exert excessively
disruptive liquidity disturbances when changing its interest rate
target.3

It follows that easing or tightening has mainly been

accomplished by changing the level of short rates to set in motion
forces slowing the growth of money demand in order to allow a future
reduction in money growth and inflation.
To view the Federal Reserve's policy instrument as the Federal
funds rate is thus to set money to the side, since at any point in time
money demand is accommodated at the going interest rate.

This does not

say, however, that money can be left out of account altogether.

The

Fed, the markets, and economists alike recognize that trend inflation is
closely connected to trend money growth, and that achieving and

3. Total reserve
demand is not very interest elastic in the short
run. So whenever the Fed cuts nonborrowed reserves to support a higher
Federal funds rate target, it allows banks to satisfy a roughly unchanged
reserve demand by borrowing the difference at the discount window. The
negative relation between nonborrowed reserves and the funds rate in part
reflects the administration of the discount window, which creates a
positive relation between bank borrowing and the spread between the funds
rate and the discount rate. Christiano and Eichenbaum (1991) emphasize the
importance of this mechanism in understanding the liquidity effect.




- 7 -

Goodfriend

maintaining price stability requires controlling money.

During the

period under study, money growth was often viewed as an important
indicator of future inflation or disinflation by both the Fed and the
markets.
Furthermore, we know from the work of McCallum (1981) and others
that an interest rate policy just describes how changes in interest
rates correspond to changes in the money stock. At a deeper level,
then, there is an equivalence between talking in terms of interest rates
or money.

The important difference is that simple interest rate rules

descriptive of policy have implications for how money and prices
actually evolve over time (Goodfriend 1987, Barro 1989).

We should keep

this in mind when reviewing the current period for clues about how
policy influences the inflation rate.

Ultimately we seek to understand

what it is about interest rate policy that turns one-time macroeconomic
shocks into highly persistent changes in the growth of money and prices.

Interpreting Comovements Between Short and Long Rates
The Fed targets the funds rate in order to stabilize inflation and real
economic growth as best it can. Output and prices, however, do not
respond directly to weekly Federal funds rate movements but only to
longer-term rates of perhaps six months or more.

Hence, the Fed targets

the funds rate with the aim of managing longer-term money market rates.
It exercises its leverage as follows. The market determines longer-term
rates (abstracting from a time varying term premium and default risk) as
the average expected level of the funds rate over the relevant horizon.




- 8 -

Goodfriend

To see why, consider the pricing of a three-month bank loan.

A bank

could fund the loan with a three-month CD, or it could plan to borrow
Federal funds overnight for the next three months.. Cost minimization
and competition among banks keep the CD rate in line with the average
expected future funds rate; competition in the loan market links loan
rates to the CD rate and expected future funds rates.

Finally,

arbitrage among holders of money market securities links Treasury bill
and commercial paper rates to CD rates of similar maturity.
Since simplicity is crucial in communicating policy intentions,
the Fed manages its funds rate target to maintain an expected constancy
over the near-term future.

Target changes are highly persistent and

seldom quickly reversed, so that a target change carries the expected
level of the funds rate with it and thus longer-term money market rates
too.4

In this way, interest rate policy as practiced by the Fed

anchors the short end of the term structure of interest rates to the
current Federal funds rate.
By the above argument, the interest rate on long bonds too must
be determined as an average of expected future short rates.

At best,

the Fed affects short-term real interest rates temporarily, so average
future short rates over the horizon of a 30-year bond should sum to a
real interest rate that varies in a range perhaps 1 or 2 percentage

4. Goodfriend (1991) contains a discussion of evidence consistent
with this view reported in Fama (1984), Fama and Bliss (1987), Mankiw,
Miron, and Weil (1987), Hardouvelis (1988), and Cook and Hahn (1989).




- 9 -

Goodfriend

points around 3% per year--pius the expected trend rate of inflation.5
From this perspective, we can view fluctuations in the long-term rate as
driven by: (1) a component connected with the current Federal funds rate
target that anchors short maturity rates, and (2) a component driven by
expectations of inflation.

Because the present discounted value of

coupon payments far out in the future is smaller at higher interest
rates, we should expect a given funds rate target change to exert a
greater effect on the long bond at higher rates of interest.6

It is

5.
Consider a bond paying nominal interest (i) taxable at rate (r),
when the expected inflation rate is (*e). The real after-tax ex ante
return on such a bond is then r » (l-r)i-ire, so the expected inflation rate
over the life of the bond may be expressed as ?re * [i - r/(l-r)](l-r).
Woodward (1990) reports market expectations of the after-tax real rate
of interest on long-term bonds using quarterly data on British index-linked
gilt-edged securities from 1982:2 to 1989:1. The ex ante post-tax real
rate ranged from 1.5% to 3.2% per annum with a mean of 2.6%.
Assuming investors keep after-tax ex ante rates on long-term government
bonds in the U.S. and U.K roughly equal, we can set r =.026 in the above
expression to infer long-term expected inflation in the U.S. A tax rate in
the U.S. of .20, for example, yields 7re = [i-3.2](.8). If we take i as the
yield to maturity on a 30-year U.S. government bond, then 7re is the average
per annum inflation rate expected over the 30-year horizon.
The tax rate in the above expression is the marginal rate that applies
to the relevant marginal investor, e.g., individual, corporation, or
foreigner. The rate is difficult to determine. Its exact value, however,
is not important for the analysis in the text. The analysis relies on the
view that significant changes in the long-term nominal rate primarily
reflect proportional movements in inflation expectations, a view supported
by the narrow range of ex ante post-tax real rates reported by Woodward.
6.
A given Federal funds rate target change will exert a greater
effect on the long-term bond rate the shorter the average life of the
security as measured by its duration. The duration of a coupon bond may be
thought of as the term to maturity of an equivalent zero coupon bond that
makes the same total payments and has the same yield. The duration of a 30
year coupon bond selling at par is approximately 1/r, where r is the yield
to maturity. See Moore (1989). Thus, the duration of the 30 year
government (coupon) bond discussed in the text is only about 12.5 years at
an interest rate of 8% and 7.1 years at a 14% interest rate.




- 10 -

Goodfriend

useful to distinguish three sources of interaction between the Federal
funds rate and the long-term rate:

Pure Cyclical Funds Rate Policy Actions. The Fed routinely lowers the
funds rate in response to cyclical downturns and raises it in cyclical
expansions.

I call such policy actions purely cyclical if they maintain

the going trend rate of inflation.

Even purely cyclical policy actions

exert a pull on longer rates, however, so they are a source of positive
comovement between the funds rate and the long rate. But because
cyclical actions strongly influence only the first few years of expected
future short-term interest rates, only a relatively small fraction of
purely cyclical funds rate changes are transmitted to the long rate.

Long-Run Inflation.

Changes in the trend rate of inflation are a second

source of positive comovement between the funds rate and the long rate.
While the long rate moves automatically with inflation expectations, the
funds rate does not unless the Fed makes it do so. Nevertheless, the
Fed can hold short-term real rates relatively steady in the presence of
rising or falling inflation by moving the funds rate up or down to allow
for a rising or falling inflation premium.

In so doing, it causes short

and long rates to move relatively closely together.

Aggressive Funds Rate Policy Actions. The Fed occasionally takes
particularly aggressive funds rate policy actions to encourage real
growth or to stop and reverse




a rising rate of inflation. Aggressive

- 11 -

Goodfriend

actions combine a purely cyclical effect with a potential change in the
long-run rate of inflation.

The cyclical effect moves the long rate in

the same direction as the funds rate, while the inflation effect moves
the long rate in the opposite direction.

Thus the net effect of

aggressive actions on the long rate is somewhat complex.
Consider an aggressive reduction in the funds rate to encourage
real growth.

Initially, funds rate actions taken to fight recession

pull the long rate down too.

However, excessive easing that raises

inflation can cause the long rate to reverse direction and begin to
rise, even as the Fed continues to push short rates down.

Thus we might

expect to see the long rate move in the opposite direction from the
funds rate near cyclical troughs. A sharp funds rate increase during
the ensuing recovery exerts two conflicting forces.

It tends to raise

the long rate by reversing the cyclical funds rate decline, but it also
reverses somewhat the expected rise in inflation, tending to lower the
long rate.

For a relatively brief recession with little excessive

easing, the cyclical funds rate effect would dominate the inflation
effect, so the long rate would tend to rise with the funds rate during
the recovery.

Thus, the long rate would move opposite from the funds

rate for only a few months near a recession trough.
Now consider an aggressive increase in the funds rate intended to
bring down the trend rate of inflation.

Such a tightening potentially

shifts both components of the long rate since short rates rise and
expected long-run inflation may fall.

One expects the first effect to

dominate initially, however, because a large aggressive increase in




- 12 -

Goodfriend

short rates exerts an immediate significant upward pull on the long
rate, while the public may not yet have confidence in the disinflation.
If the Fed persists with sufficiently high short-term real rates,
however, inflation and real growth eventually slow and the Fed can
tentatively bring rates down somewhat.

A declining long rate, at this

point, would suggest that the Fed's disinflation has acquired some
credibility.

Inflation Scares
I call a significant long rate rise in the absence of an aggressive
funds rate tightening an "inflation scare", since it reflects rising
expected long-run inflation.7

Inflation scares are costly because the

higher inflation that they signal reduces the efficiency of the payments
system, with negative consequences for employment, productivity, and
economic growth.

Moreover, scares are costly because they present the

Fed with a difficult dilemma.

Resisting them requires the Fed to raise

real short rates with potentially depressing effects on business
conditions.

But failing to respond promptly creates a crisis of

confidence that encourages the higher inflation to materialize: workers
and firms ask for wage and price increases to protect themselves from
higher expected costs.

In short, by hesitating, the Fed sets in motion

higher inflation that it is then inclined to accommodate with faster
7. Since short maturity rates are anchored to the Federal funds rate
target, they cannot convey as clear a signal of inflation expectations as
the long rate. See Dotsey and King (1986) for more on the informational
implications of interest rate rules.




- 13 -

Goodfriend

money growth. The record of rising inflation and disinflation reviewed
below contains examples of scares that resulted in higher money growth
and inflation, as well as those that were successfully resisted by the
Fed.8
A REVIEW OF INTEREST RATE POLICY
This study focuses on the period of disinflation beginning in October
1979.

Nevertheless, I begin my review by briefly describing conditions

in the immediately preceding years. For the most part, the data
discussed throughout come from charts and tables included at the back of
the paper.

Rising Inflation: the Late 1970s
Inflation was rising gradually in the late 1970s, with rates of 6.9%,
7.9%, and 8.6% in 1977, 1978, and 1979 as measured by fourth quarter
over fourth quarter changes in the GDP deflator. The corresponding real
GDP growth rates were 4.5%, 4.8%, and 2.5%. The persistent inflation
scare throughout the late 1970s carried the 30-year government bond rate
from 7.8% in early 1977 to 9.2% by September 1979. Over the same
period, the Fed steadily increased the Federal funds rate from around
4.7% to 11.2%, raising short-term real rates from a range between 0 to
-2% to between 0 and +2%. The negative short-term real rates at the

8. An inflation scare may be consistent with either a positive or a
negative association between money or prices, on one hand, and unemployment
or real growth on the other, depending on the nature of the underlying
macroshock that sets it off.




. H .

Goodfriend

beginning of the period suggest that initially the Fed was actively
encouraging inflation in order to stimulate real growth, though the
steady increase in real short rates indicates a modest effort to resist
inflation.

Aborted Inflation Fighting: October 1979 to July 1980
By the time Paul Volcker became Fed Chairman in August 1979, oil price
increases following the Iranian revolution in November 1978 greatly
worsened the inflation outlook. Oil prices were to double by early 1980
and triple by early 1981 from November 1978 levels, and by the fall of
1979 the Fed felt that more drastic action was needed to fight
inflation. The announcement on October 6, 1979 of the switch to
nonborrowed reserve targeting officially opened the period of
disinflation policy.
The first aggressive policy actions in this period took the
monthly average funds rate from 11.4% in August 1979 to 17.6% in April
1980.

Cook (1989) reports that only 1 percentage point of this 6 point

rise can be attributed to automatic adjustment. Virtually all of it
represented deliberate policy actions taken by the Fed to increase
short-term interest rates.

It was the most aggressive series of actions

the Fed had taken in the post-war period over so short a time, although
the 5 percentage point increase from January to September of 1973 was
almost as large.
For its part, the 30-year rate rose sharply from 9.2% in August
to a temporary peak of 12.3% in March after which it fell back to 11.4%




- 15 -

Goodfriend

in April.
rise.

A closer look reveals the sources of this sharp long rate

The sharp 2 percentage point monthly average funds rate rise from

September to October pulled the long rate up about 0.6 percentage
points.

The monthly average funds rate then held in a range between

13.2% and 14.1% through February.

January 1980 later turned out to be

an NBER business cycle peak, and evidence of a weakening economy caused
the Fed to pause in its aggressive tightening.

But with the funds rate

relatively steady, the long rate jumped sharply by around 2 percentage
points between December and February, indicating a very serious
inflation scare.
The scare was probably caused in part by the ongoing oil price
rises, but the Fed's hesitation to proceed with its tightening may have
contributed to the collapse of confidence.

In any case, the Fed reacted

with an enormous 3 percentage point increase of the monthly average
funds rate in March, 1 percentage point of which was due to the
automatic adjustment.

The long rate hardly moved in response,

suggesting that the positive effect of the aggressive rise was offset by
a decline in expected inflation.

Moreover, the long rate actually came

down by 0.9 percentage points in April even as the Fed pushed the funds
rate up another 0.4 percentage points, suggesting that the Fed had
already begun to win credibility for its disinflation policy.
When one considers that business peaked in January, there is
reason to believe that inflation would have come down as the recession
ran its course in 1980 if the Fed had sustained its high interest rate
policy.




The imposition of credit controls in March, however, forced the

- 16 -

Goodfriend

Fed to abort that policy.

Schreft (1990) argues persuasively that by

encouraging a decline in consumer spending the credit control program
was largely responsible for the extremely sharp -9.9% annualized decline
in real GDP in the second quarter of 1980. Supporting her view is the
fact that personal consumption expenditures accounted for about 80% of
the decline in real output, more than twice its average 35% contribution
in post-war U.S. recessions.
Accompanying the downturn in economic activity was a sharp fall
in the demand for money and bank reserves that, according to Cook
(1989), caused a 4.2 percentage point automatic decline of the funds
rate from April to July. The Fed enhanced the automatic easing with
judgmental actions, e.g., reducing the discount surcharge, that reduced
the funds rate by an additional 4.3 percentage points over this period.
The sharp interest rate decline coupled with the lifting of credit
controls in July led to strong 8.4% annualized real GDP growth in the
fourth quarter of 1980.

In spite of the credit controls, or more

accurately, because the credit controls caused the Fed to interrupt its
inflation-fighting effort, inflation rose through the year from an
annual rate of 9.8% in the first quarter to 10.9% in the fourth quarter
as measured by the GDP deflator.

Aggressive Disinflation Policy: August 1980 to October 1982
It was clear in late summer and early fall of 1980 that inflationary
pressures were as strong as ever. After being pulled down roughly 2
percentage points by the aggressive funds rate easing from April to




- 17 -

Goodfriend

June, the 30-year rate rose by about 50 basis points between June and
July as the Fed continued to push the funds rate down another 50 basis
points.

The reversal signalled an inflation scare induced by the

excessively aggressive easing, and the Fed began an unprecedented
aggressive tightening.

Of the roughly 10 percentage point rise in the

monthly average funds rate from July to December 1980, Cook (1989)
attributes only about 3 percentage points to the automatic adjustment.
Thus, the runup of the funds rate to its 19% peak in January 1981 marked
a deliberate return to the high interest rate policy.

As measured by

the GDP deflator, which was rising at nearly a 12% annual rate in the
first quarter of 1981, real short-term rates were a high 7% at that
point.
As soon as the funds rate peak had been established, however,
very slow growth in Ml and bank reserves

automatically put downward

pressure on the funds rate. According to Cook (1989), about 3.4
percentage points of the 4 percentage point drop in the funds rate
between January and March was attributable to the automatic adjustment.
Since the automatic adjustment had correctly signalled weakness in the
economy in the second quarter of 1980, the Fed was initially inclined to
let rates fall in early 1981. However, real GDP actually grew at a 5.6%
annual rate in the first quarter, and when the strength of the economy
became clear, the Fed took deliberate actions to override what it took
to be a false signal that disinflation had taken hold.

Reversing field,

it ran the funds rate back up to 19% by June, using a series of
deliberate tightening actions to supplement what Cook (1989) reports




- 18 -

Goodfriend

would only have been a 0.8 percentage point automatic funds rate rise.
It was not long before the aggressive disinflationary policy began
to take hold.

Annualized real GDP growth was -1.7% in the second

quarter of 1981. The third quarter posted 2.1% real growth, but an NBER
business peak was reached in July and real growth fell to -6.2% in the
fourth quarter of 1981 and -4.9% in the first quarter of 1982.
Meanwhile, the quarterly inflation rate as measured by the GDP deflator
fell from 11.8% in the first quarter of 1981 to the 4.5% range by early
1982.
The Fed brought the funds rate down from 19% at the business peak
in July to 13.3% in November and held the funds rate in the 13 to 15
percent range until summer 1982 when it brought short rates down another
4 percentage points to around 10%. The funds rate reduction through
November 1981 was large in nominal terms, but when one considers that
inflation had declined to the 4.5% range by early 1982, the funds rate
decline actually represented a 1 or 2 percentage point rise in shortterm real rates. Thus, one should still view policy as aggressively
disinflationary in early 1982. As calculated by Cook (1989), automatic
adjustments accounted for only 1 percentage point of the final 9
percentage point funds rate decline in the nonborrowed reserve targeting
period, which ended formally in October of 1982. This last great
decline should be seen as a deliberate funds rate easing calculated to
achieve a sustained reduction in inflation without excessive harm to
real growth.




The long rate provides a picture of the Fed's progress over the

- 19 -

Goodfriend

nonborrowed reserve
inflation.

targeting period in reducing the trend rate of

The 30-year rate rose about 5 percentage points from a

trough in June of 1980 to its 14.7% peak in October 1981. About 2
percentage points of that rise appears to be connected with the rundown
and runup of the funds rate in 1980, the remaining 3 point gain through
October 1981 reflected a continuing serious inflation scare.

In fact,

the sharp rise in the long rate after the funds rate had reached its
peak in early 1981 may have contributed to the Fed's inclination to
persist with its 19% funds rate until August 1981. Moreover, the
discernable declining trend in the long rate from October 1981 to August
1982 indicates that the policy was still exerting disinflationary
pressure.

When the Fed finally decided to relax its disinflation policy

by dropping the funds rate by over 4 percentage points in the summer of
1982, the long rate fell by around 3.5 percentage points along with it.
We can decompose this last decline in the long rate into a purely
cyclical component and an inflation expectations component using
evidence from earlier in the aggressive funds rate targeting period.
The sharp 2 percentage point funds rate rise from September to October
1979 pulled the long rate up 0.6 percentage points; and the sharp 8.5
percentage point funds rate reduction between April and July 1980 pulled
the long rate down 2 percentage points.

Taking 25% as the fraction of

cyclical funds rate policy actions transmitted to the long rate, about
2.5 percentage points of the 3.5 percentage point fall in the long rate
in the summer of 1982 reflected a reduction of inflation expectations.




- 20 -

Goodfriend

Establishing Credibility: November 1982 to Spring 1986
Real GDP growth was still poor in the second half of 1982, running -1.8%
and 0.6% in the third and fourth quarters, respectively.

Consequently,

the Fed continued to ease after relaxing its disinflationary policy,
pushing the monthly average funds rate down to 8.5% by February 1983.
November 1982 turned out to be an NBER business cycle trough, and real
GDP growth was 2.6% in the first quarter of 1983. But the Fed kept the
funds rate around 8.5% through May while the long rate remained steady
at around 10.5%.

It gradually became clear, however, that a strong

recovery had begun.

Real GDP grew at a spectacular 11.3% annual rate in

the second quarter of 1983 and at rates of 6.1%, 7.0%, 7.9%, and 5.4% in
the following four quarters.
The Fed reacted to the recovery by raising the funds rate from
8.6% in May to 9.6% in August 1983. But the long rate rose
simultaneously from 10.5% to 11.8%, initiating a serious inflation scare
only a year after the Fed had relaxed its disinflation policy.
Annualized quarterly inflation as measured by the GDP deflator was 4.8%
or below throughout 1983 and 1984 with the exception of the first
quarter of 1984, when it was 6%.

Nevertheless, the long rate embarked

on a spectacular rise to a 13.4% peak in June 1984. Amazingly, this was
only about a percentage point short of its October 1981 peak, even
though by 1984 inflation was 4 or 5 percentage points lower than in
1981.
The Fed tightened in an effort to resist the inflation scare,
raising the funds rate to an 11.6% peak in August of 1984. The long




- 21 -

Goodfriend

rate began to decline in June 1984, indicating that the scare had been
contained:

The 7% real short rates needed to contain the scare

ultimately brought quarterly real GDP growth down to the more normal 2
to 3 percent range in the second half of 1984. The Fed then lowered the
funds rate rapidly by 3.2 percentage points from August to December and
held it around 8% through 1985.
Meanwhile, the long rate fell about 6 percentage points from its
June 1984 peak to the mid-7% range by the spring of 1986.

By then, the

long rate was 3 percentage points below where it had been at the start
of the 1983 scare. The Fed's containment of the scare apparently made
the public confident of another 3 percentage point reduction in the
trend rate of inflation.

Maintaining Credibility: Spring 1986 to Summer 1990
Real GDP growth weakened considerably in the second quarter of 1986 to
-0.3% from the strong 5.4% rate in the first quarter.

With inflation

appearing to have settled down in the 4% range, the Fed moved to
encourage real growth by dropping the funds rate to the mid-6% range.
Strong real growth in 1987 was accompanied by still another inflation
scare in which the long rate rose about 2 percentage points from around
7.5% in March to 9.6% in October.
Although real GDP growth was very strong throughout the year,
this time the Fed responded to the scare with only a relatively modest
increase in the funds rate.

As it happened, the October stock market

crash contained the scare somewhat, but the long rate remained above 8%.




- 22 -

Goodfriend

With real growth still reasonably strong in 1988, the Fed proceeded to
raise the funds rate sharply from the 6 to 7% range in early 1988 to a
peak of 9.8% in March 1989.
Though there was some evidence of a modest rise in inflation in
1988, the sustained funds rate tightening during the year is unique in
that it was undertaken without a rise in the long rate. A preemptive
tightening may have been needed to rewerse

the perception that policy

had eased permanently following the stock market crash. At any rate,
the result was an increase in credibility reflected in a further decline
in the long rate in 1989. Though that fall was partially reversed in
early 1990, a gently declining trend in the long rate was discernable by
then, indicating growing confidence on the part of the public in the
Fed's commitment to low inflation.

The 1990-91 Recession
The period of weak real growth in 1989 ending in an NBER business cycle
peak in July 1990 may have been partly due to the high real short rates.
Temporary oil price increases following the invasion of Kuwait, however,
also helped account for the near zero real growth in the third quarter
of 1990, -3.9% real growth in the fourth quarter, and -2.5% in the first
quarter of 1991.
The Fed responded to the recession by bringing the funds rate
down from slightly above 8% in the fall of 1990 to around 3% today.
is remarkable that this sustained easing has not yet caused the long
rate to rise, even though real short rates are now around zero. Real




- 23 -

It

Goodfriend

short rates were also about zero when excessive easing sparked the
inflation scare in July 1980, but they were around 4% when excessive
easing triggered the June 1983 scare, and around 3% at the time of the
scare in April 1987.9 The real short rate floor at which easy policy
becomes excessive no doubt varies to an extent with underlying real
economic conditions such as government tax and spending policy,
productivity shocks, or shifts in investment and consumer demand.10
But long rates may also be more tolerant of aggressive funds rate easing
when the public is more confident of the Fed's commitment to maintain a
low trend rate of inflation.
OBSERVATIONS
The record of interest rate policy reviewed above contains a number of
empirical findings that are important for interpreting and evaluating
monetary policy. This section summarizes the main findings in a series
of observations.
1) Inflation scares appear to be central to understanding the
Fed's management of short-term interest rates. The gradual funds rate
rise from 1977 to October 1979 was undertaken in an environment of
slowly rising long rates. The sharp long rate rise in early 1980,
during a 4 month pause in the funds rate rise, was probably an important

9. The effect of the credit control program on consumer spending may
account for the real rate getting as low as it did in 1980 before
triggering a scare.
10. See, for example, the discussions in Campbell and Clarida (1987)
and Poole (1988).




- 24 -

Goodfriend

factor inducing the Fed to undertake its enormous 3 percentage point
tightening in March.

Sharply rising long rates in the first nine months

of 1981 indicated that the Fed had yet to win credibility for its
disinflationary policy, and probably contributed to the Fed's
maintaining very high real short rates for as long as it did.

On the

other hand, the declining long rate from October 1981 to October 1982
encouraged the Fed to ease policy by indicating the public's growing
confidence in the disinflation.
The serious inflation scare set off in the summer of 1983
largely accounts for the runup of the funds rate to August 1984.

The

credibility acquired by the Fed in containing that scare yielded a 3
percentage point reduction in the long rate that allowed the funds rate
to come down further too.

There was no inflation scare per se when the

Fed raised the funds rate in 1988. Nevertheless, that series of actions
may be understood as preemptive, taken to reverse a public perception
that policy had permanently eased following the stock market crash.

The

current funds rate easing has yet to trigger a sustained rise in the
long rate, but the possibility of an inflation scare has probably
limited the funds rate decline somewhat.
2) One might reasonably have expected the aggressive disinflation
policy beginning in late 1979 to reduce long-term interest rate
volatility by quickly stabilizing long-term inflation expectations at a
low rate.

Yet the reverse was true initially.

Long rates turned out to

be surprisingly volatile due to a combination of particularly aggressive
funds rate movements and inflation scares. Amazingly, it took until




- 25 -

Goodfriend

1988 for the unusual long rate volatility to disappear.
3) One might also have expected the aggressive funds rate actions
beginning in 1979 to be accompanied by opposite movements in the long
rate.

Again, the result was just the reverse.

The aggressive actions

moved the long rate in the same direction, apparently influencing the
long rate primarily through their effect on short maturity rates. Only
at funds rate peaks and troughs did the long rate move in the opposite
direction.

The long rate appeared to be influenced by a change in

expected inflation only after sustained aggressive funds rate actions.
4) The long rate reached its peak in October 1981, indicating
that it took two years for policy to reverse the rise in the trend rate
of inflation.

It would be a mistake, however, to conclude that

acquiring credibility necessarily takes so long.

On the contrary, a

close look reveals that the long rate had already turned down in April
1980 while the funds rate was still rising, indicating that some
credibility had been won by then.

Credibility might even have been

achieved sooner if the Fed had not hesitated temporarily between
December 1979 and February 1980 to continue the aggressive funds rate
tightening begun in October.

In any case, the credit control program

interrupted the disinflation policy in May 1980 and high interest rates
were restored fully only in early 1981. The automatic adjustment
feature of the nonborrowed reserve

operating procedure then caused a

sharp decline in the funds rate between January and March of 1981 that
was only fully reversed by June. Thus, three interruptions account for




- 26 -

Goodfriend

the delay in the Fed's acquisition of credibility for its disinflation
policy.
5) Interestingly enough, the long rate was roughly in the same 8%
range in the early 1990s as it was in the late 1970s, in spite of the 4
or 5 percentage point reduction in the inflation rate.

Apparently,

investors then perceived the 7 to 9% inflation rate as temporarily high,
while, if anything, they perceive the current 3 to 4% rate as a bit
below trend.

The slowly declining long rate in the current period is

indicative of the steady acquisition of credibility, but the high long
rate indicates a lingering lack of confidence in the Fed.
6) The Fed appears to have remarkable latitude to push the
Federal funds rate down in the recent recession without triggering a
rise in the long rate. On three occasions when trying to encourage real
growth in the 1980s (July 1980, June 1983, and April 1987) it could not
push the funds rate more than 1 or 2 percentage points below the long
rate before triggering an inflation scare; yet it pushed the funds rate
4 percentage points below the long rate in 1992.
The greater flexibility to reduce short rates evident in the
current recession is reminiscent of that in early post-war recessions
when the Fed presumably had more credibility.

Chart 2 shows that the

funds rate was pushed almost 3 percentage points below the long rate
during the August 1957 - April 1958 recession before the long rate began
to rise.

The funds rate came down over 2 percentage points below the

long rate in the April 1960 - February 1961 recession without much of a




- 27 -

Goodfriend

rise in the long rate.11
7) The preceding observation suggests a powerful argument in
favor of a Congressional mandate for price stability.

By reducing the

risk of inflation scares, such a mandate would free the funds rate to
react more aggressively to unemployment in the short run.

Thus, a

mandate for price stability would not only help eliminate inefficiencies
associated with long-run inflation, but the added flexibility conferred
on the funds rate might improve countercyclical stabilization policy as
well. 12

CONCLUSION
The paper used institutional knowledge of Fed policy procedures, simple
economic theory, and the inflation scare concept to analyze and
interpret interest rate policy as practiced by the Fed since 1979.

It

focused on the primary policy problem during the period: the acquisition
and maintenance of credibility for the commitment to low inflation.

We

saw that the Fed might have acquired credibility for its disinflation
relatively quickly in early 1980 had it been able to sustain a high
interest rate policy then.

After all, long term rates were roughly

equal to the inflation rate in 1979, indicating that the public believed

11. Kessel (1965) contains a good description and analysis of the
cyclical relation between long and short rates.
12. See Black (1990) for a discussion of the benefits of price
stability. Hetzel (1990 and 1992) discusses a proposal that the U.S.
Treasury issue indexed bonds to provide a better indicator of long-run
inflation expectations.




- 28 -

Goodfriend

inflation was only temporarily high at the time.

Unfortunately, a

series of interruptions delayed the actual disinflation for two years,
probably raising the cost in terms of lost output of acquiring
credibility.
Soon after relaxing its disinflation policy in 1982, the Fed's
credibility was again challenged with a serious inflation scare that
carried the long rate up from 10.5% to 13.4%.

It took 11 months and 7%

real short rates to contain the scare, indicating how fragile the Fed's
credibility was in 1983 and 1984. The long rate decline to the 7.5%
range by the spring of 1986 reflected a big gain in credibility.

Yet

the Fed was tested by another scare in 1987 that ended with the stock
market crash.

The crash itself, however, then set in motion

expectations of excessive easing that the Fed resisted with a 3
percentage point funds rate rise in 1988 and 1989, a tightening that
probably weakened real growth somewhat in 1989 and 1990.
Reviewing the policy record makes one understand how fragile the
Fed's credibility is and how potentially costly it is to maintain.

Even

after inflation had stabilized at around 4% in 1983, inflation scares
and the Fed's reaction to them were associated with significant
fluctuations in real growth.

With that in mind, one cannot help but

appreciate the potential value of a Congressional mandate for price
stability that would help the Fed establish a credible commitment to low
inflation.

In fact, there is evidence that an interest rate policy

assisted by such a mandate would work well.

Both the Bundesbank and the

Bank of Japan follow interest rate policies resembling the Fed's and




- 29 -

Goodfriend

yet, for the most part, they have achieved better macroeconomic
performance.

Perhaps it is because they each enjoy a stronger mandate

for price stability than does the Fed.




- 30 -

Goodfriend

REFERENCES
Annual Reports of Open Market Operations, Federal Reserve Bank of .
New York Economic Review. 1981-83.
Barro, Robert J., "Interest Rate Targeting." Journal of Monetary
Economics. January 1989, 23, 3-30.
Bernanke, Ben and Alan Blinder, "The Federal Funds Rate and the
Channels of Monetary Transmission," American Economic Review.
September 1992, 82, 901-21.
Black, Robert P., "In Support of Price Stability," Federal Reserve
Bank of Richmond Economic Review. January/February 1990, 3-6.
Campbell, John Y. and Richard H. Clarida, "The Dollar and Real
Interest Rates," Carnegie-Rochester Conference Series on Public
Policy. 1987, 27, 103-40.
Christiano, Lawrence and Martin Eichenbaum, "Liquidity Effects,
Monetary Policy, and the Business Cycle," Federal Reserve Bank of
Minneapolis, June 1991.
Cook, Timothy, "Determinants of the Federal Funds Rate:
1979-1982," Federal Reserve Bank of Richmond Economic Review.
January/February 1989, 3-19.
Cook, Timothy and Thomas Hahn, "The Effect of Changes in the
Federal Funds Rate Target on Market Interest Rates in the 1970s,"
Journal of Monetary Economics. November 1989, 24, 331-51.
Dotsey, Michael and Robert G. King, "Informational Implications of
Interest Rate Rules," American Economic Review. March 1986, 76,
33-42.
Fama, Eugene F., "The Information in the Term Structure," Journal
of Financial Economics. 1984, 13, 509-28.
Fama, Eugene and R. Bliss, "The Information in Long Maturity
Forward Rates," American Economic Review. 1987, 77, 680-92.
Friedman, Benjamin M., "Lessons on Monetary Policy from the
1980s," The Journal of Economic Perspectives. Summer 1988, 2, 5172.
Friedman, Milton and Anna J. Schwartz, A Monetary History of the
United States Princeton: Princeton University Press, 1963.




- 31 -

Goodfriend

Goodfriend, Marvin, "Discount Window Borrowing, Monetary Policy,
and the Post-October 6, 1979 Federal Reserve Operating Procedure,"
Journal of Monetary Economics. September 1983, 12, 343-56.
, "Interest Rate Smoothing and Price Level
Trend-Stationarity," Journal of Monetary Economics. May 1987, 19,
335-48.
, "Interest Rates and the Conduct of Monetary
Policy," Carnegie-Rochester Conference Series on Public Policy.
Spring 1991, 34, 7-30.
Goodfriend, Marvin and William Whelpley, "Federal Funds:
Instrument of Federal Reserve Policy," Federal Reserve Bank of
Richmond Economic Review. September/October 1986, 3-11.
Hardouvelis, G., "The Predictive Power of the Term Structure
During Recent Monetary Regimes," Journal of Finance. 1988, 43,
339-56.
Hetzel, Robert L., "Indexed Bonds as an Aid to Monetary Policy,"
Federal Reserve Bank of Richmond Economic Review. January/February
1992, 13-23.
, "Maintaining Price Stability: A Proposal,"
Federal Reserve Bank of Richmond Economic Review. March/April
1990, 53-55.
Kessel, Reuben A. The Cyclical Behavior of the Term Structure of
Interest Rates. New York: National Bureau of Economic Research,
1965.
Levin, Fred J. and Paul Meek, "Implementing the New Operating
Procedures: The View from the Trading Desk," in New Monetary
Control Procedures, edited by Stephen H. Axil rod, Washington:
Board of Governors of the Federal Reserve System, 1981.
Macau!ay, Frederick R., Bond Yields. Interest Rates, and Stock
Prices. Cambridge, MA: National Bureau of Economic Research,
1938.
Mankiw, N. Gregory, Jeffrey Miron, and David N. Weil, "The
Adjustment of Expectations to a Change in Regime: A Study of the
Founding of the Federal Reserve," American Economic Review. June
1987, 77, 358-74.
McCallum, Bennett T., "Price Level Determinacy with an Interest
Rate Policy Rule and Rational Expectations," Journal of Monetary
Economics. November 1981, 8, 319-29.




- 32 -

Goodfriend

Miron, Jeffrey, "Financial Panics, the Seasonality of the Nominal
Interest Rate, and the Founding of the Fed," American Economic
Review, March 1986, 76, 125-40.
Moore, George, "Forward-Looking Bond Yield Approximations," Board of
Governors of the Federal Reserve System, September, 1989.
Poole, William, "Monetary Policy Lessons of Recent Inflation and
Disinflation," The Journal of Economic Perspectives, Summer 1988,
2, 73-100.
Rogoff, Kenneth, "Reputation, Coordination, and Monetary Policy,"
Carnegie- Rochester Conference Series on Public Policy, Spring
1987; reprinted in Robert J. Barro, editor, Modern Business Cycle
Theory, Cambridge,MA: Harvard University Press, 1989, 236-64.
Schreft, Stacey L., "Credit Controls: 1980," Federal Reserve Bank
of Richmond Economic Review, November/December 1990, 25-55.
Sims, Christopher, "Interpreting the Macroeconomic Time Series
Facts: The Effects of Monetary Policy," Yale University, August
1991.
Thornton, Daniel, "The Borrowed-Reserves Operating Procedure:
Theory and Evidence," Federal Reserve Bank of St. Louis Economic
Review, January/February 1988, 30-54.
Wallich, Henry C , "Recent Techniques of Monetary Policy," Federal
Reserve Bank of Kansas City Economic Review, May 1984,
21-30.
Woodward, G. Thomas, "The Real Thing: A Dynamic Profile of the Term
Structure of Real Interest Rates and Inflation Expectations in the
United Kingdom, 1982-89," Journal of Business, July 1990, 63,
373-98.




- 33 -

Chart 1

FEDERAL FUNDS RATE AND 30-YEAR BOND RATE
January 1977 - April 1992

20

30-Year Bond Rate

3

'|Mlllllttll|IIIUttltMIIIIIIIIIIII|HMttlMII|IIHMIMM|llMltlllll|IIMIIIIIIt|IIIHIIIIII|MIIIIIIIM|lltlHttMIIIIIIMIIIM|IIIIMIMIt|IMn

77
P«rc«nt Chang*. 40 to 40:
Rttl OOP
4.4
I l l i c i t Prlct
Doflitor
7.3




78

79

80

6.0

0.0

-0.2

6.4

0.7

10.1

81

82

83

84

85

86

87

88

89

90

91

-0.1

-i.l

6.7

4.9

3.3

2.2

4.5

3.3

1.7

-0.1

0.3

0.4

4.4

4.0

4.3

3.6

2.6

3.3

4.2

4.2

4.2

3.0

92

Chart 2

FEDERAL FUNDS RATE AND 20-YEAR BOND RATE
August 1954 - December 1964
4.5

4.0 +
20-Year Bond Rate
3.5 +

E

| 3.0
c
<
c
a 2.5
-M
C
0)

u
t-

*

^

cu 2 . 0
CL

1.5 +

1.0 +

5 ' I » III |I »I U M M M \ t1 M 111 M 1111 M 111 1 III11 I I 111 111 M 11 M t M I i M M |»It < M M N I ( M M M I M M | M I M II M H 11 M M I H 1 11II




55

56

57

58

59

60

61

62

63

i

0

C

O

o i N i n i n s N < ) n c D o o i n N N O N N N < f < r H O ) a ) a ) M n o o o ) a )
OCOOKIO)(OOOCO(DOOOOCOCOC08)

000)0)0)(BC0(D(DC0C0NNOC0C0

O O H O O N O H H

) < W 0 2 O '

,

) h Z < I 1 1 < e i 0 Z O '

,

) & . r < r n ' l < W 0 Z Q 1 b . I <

(DcoeotooacoNNiotfxo

H < n e o e o c o m N O ) o o o i n « ( M N N N H N H H N H c o n o ) N H a ) ^ 0 ) o o i o ^ N c o « o o o )

h

o>a>CBO>CDa>cncoa>coeoeococoi

1 l i 4 l < S ^

r\rsNNrNr»cooo(orsr<>(OiOiO(Oiotntn(Oioiou)(OiOiO(OiO(ONr«(OtoiOiO(l)(ONNrsoooo(D(Ooo

N N N i f l N o m m o o i r i N O H n n N

m r*. i-t r»* M n m o u > « H O t f > ^ o ( 0 ( o a > N r ^ r ^ ( n c M o c ^ ^ a > < « > t n i n G o r s . ^ f ^ . a > t H t n c 4 c n r ) c n
sr «e co <»
. . . . o « < n i n i o i n o i n « a a ) n t n i n N n ( O N t n n n « n i n N N i n ( D a > i n a ) 0 ) H c o < r < o
( 0 0 ( 0 0 ( 0 0 ) 0 ) 9
>(00(0<DO)0)OIIO)0(00)0)
OOOOOOCBCDflDi

CO OO

moeoNNnfflONO) mN<(oeooiiflNiOKO>ifl*HC)0(OMn(n
m m o N O) t n eo o> CD o> O N H ( o * o i ( O O i i n H ( o e o o o ) * H H r ) O N

5 sss:

§

51

,3t

9

I a?.

•-)U43:<2:,-5,n<woa5Q'-ju«s<z,^,n<wozQ'->ui3:<2:«^n<wozQ»^u.3:<i:»->t-5<wozQ

n o n n o N n ( o m c o m i n ( O i n n « H « 0 ) ( O i o N
(0^ma><a-cotDmr«.aorva>ro(04'^c4iocMcointn
N N N n n N n 4 « « n n « « n n n n n N N H o o o o o o o o H H H H H H H H c s N i n n n N N H H H

)

rs» m o ra o
co n H to
0 ) O O O N N H O 0 I O H H H N N

Z < I ^ n < M O 2 Q n h Z < Z ^ n < « O Z O

i N N ( 0 ( o < f m i n o ) N H H o i n t o H N o n c o N i o i n ( D 4 N i o o ) H O ) N i o n ^ o o ) n
en oj rv.(© Q i ^ H i n H i n N N a u j i n N s i o c n n i n ^ ^ n ^ i n i n o t N n o N t t n a ^
0 ) i n « n o a o ) N < n n n N n ^ 4 < 7 ^ < f N o o o ) o > o « o c o e o e o ( o c o o ) a ) a a > o ) 0 ) 0 ) C D O ) o o H H H H O ) 0 )

1 h I < l n i < I Q O 2 D n t u I < I ' m < t 0 O S Q ^ U

COCOCOaOa0CDCDC0eOC00DCOCOO)CDO>O)COCOCO

•-•<oo>fn«no>c>JO^,r^»Hto
(OiO(DN(on40)H^inini
^^^^inininintf>(DU>(C^tO(0(0^^rs.(DeococoooooooooOf^<r>r)cnr>^-r^r^ocococoor4w^a>

•^b«z<z,^,-)<woaBQ,->ui2:<zn«-)<coozQr)u*2:<s:^)r)<wo*Q^u*2:<s:^,^<wozQ




Table 2
FEDERAL FUNDS HATE AHD 20-YEAR GOVERIMEBT BOND SATE
August 1954 - December 1964

Federal
20-Year
Funds
Govt. Band
Rate
Rate
(Percent per Annum)
1954:

1955:

1956:

1957:

1958:

1959:




A
S
0
N
D
J
F
M
A
M
J
J
A
S
0
N
D
J
F
M
A
M
J
J
A
S
0
N
D
J
F
M
A
M
J
J
A
S
0
N
D
J
F
M
A
M
J
J
A
S
0
N
D
J
F
M
A
M
J
J
A
S
0
N
D

1.21
1.07
0.90
0.91
1.26
1.37
1.29
1.35
1.43
1.43
1.62
1.66
1.90
2.18
2.24
2.35
2.48
2.44
2.50
2.50
2.62
2.75
2.71
2.74
2.74
2.95
2.96
2.88
2.94
2.93
3.00
2.96
3.00
3.00
3.00
2.99
3.24
3.50
3.50
3.22
2.98
2.72
1.67
1.20
1.26
0.63
0.93
0.68
1.53
1.76
1.80
2.27
2.42
2.48
2.40
2.80
2.96
2.90
3.39
3.44
3.50
3.76
3.98
4.00
3.99

2.58
2.60
2.61
2.65
2.67
2.75
2.83
2.84
2.85
2.87
2.86
2.94
3.01
3.00
2.93
2.93
2.98
2.94
2.91
2.99
3.14
3.06
3.00
3.08
3.22
3.28
3.26
3.37
3.45
3.41
3.30
3.32
3.40
3.49
3.65
3.72
3.75
3.73
3.76
3.61
3.38
3.27
3.31
3.29
3.17
3.17
3.23
3.39
3.65
3.80
3.81
3.76
3.86
3.95
3.96
3.99
4.06
4.13
4.14
4.16
4.15
4.29
4.19
4.20
4.33

Federal
20-Year
Funds
Govt. Bond
Rate
Rate
(Percent per Annum)
1960:

1961:

1962:

1963:

1964:

J
F
M
A
M
J
J
A
S
0
N
D
J
F
M
A
M
J
J
A
S
0
N
D
J
F
M
A
M
J
J
A
S
0
N
D
J
F
M
A
M
J
J
A
S
0
N
D
J
F
M
A
M
J
J
A
S
0
N
D

3.99
3.97
3.84
3.92
3.85
3.32
3.23
2.98
2.60
2.47
2.44
1.98
1.45
2.54
2.02
1.50
1.98
1.73
1.16
2.00
1.88
2.26
2.62
2.33
2.14
2.37
2.70
2.69
2.29
2.68
2.71
2.93
2.90
2.90
2.94
2.93
2.91
3.00
2.98
2.90
3.00
2.99
3.02
3.49
3.48
3.50
3.48
3.38
3.48
3.48
3.43
3.47
3.50
3.50
3.42
3.50
3.45
3.36
3.52
3.85

4.42
4.28
4.14
4.23
4.20
4.04
3.91
3.84
3.86
3.92
3.96
3.91
3.90
3.84
3.81
3.81
3.74
3.89
3.93
4.04
4.04
4.01
4.00
4.07
4.10
4.12
4.04
3.93
3.92
3.96
4.05
4.01
4.00
3.94
3.93
3.92
3.94
3.97
3.98
4.03
4.02
4.02
4.06
4.03
4.09
4.12
4.16
4.19
4.19
4.17
4.22
4.24
4.20
4.17
4.16
4.18
4.20
4.20
4.17
4.18




Table 3
QUARTERLY CHANGES IH REAL GDP AHD GDP IMPLICIT PRICE DEFLATOR
(Seasonally Adjusted Compound Annual Rates)
1Q 1977 - 1Q 1992

Real
GDP
(Pareant)
1977:

1978:

1979:

1980:

1981:

1982:

1983:

1984:

1985:

1986:

1987:

1988:

1989:

1990:

1991:

1992:

1
2
3
4
1
2
3
4
1
2
3
4
1
2
3
4
1
2
3
4
1
2
3
4
1
2
3
4
1
2
3
4
1
2
3
4
1
2
3
4
1
2
3
4
1
2
3
4
1
2
3
4
1
2
3
4
1
2
3
4
1

6.0
6.9
5.7
-0.8
2.8
13.5
3.1
4.8
0.1
0.4
2.5
0.7
1.7
-9.9
0.1
8.3
5.6
-1.7
2.1
-6.2
-4.9
1.6
-1.8
0.6
2.6
11.3
6.1
7.0
7.9
5.4
2.2
2.7
2.7
3.2
5.2
2.3
5.4
-0.3
2.3
1.3
3.0
5.1
4.0
5.9
2.6
4.3
2.5
3.9
2.5
1.9
1.1
1.2
1.7
1.6
0.2
-3.9
-2.5
1.4
1.8
0.4
2.4

IaaiUcit
Price
Deflator
(Percent)
6.1
8.4
7.4
7.3
5.7
10.7
8.3
8.8
8.6
8.4
9.6
8.1
9.8
9.6
10.0
10.9
11.8
7.5
9.6
8.8
4.5
5.5
4.4
3.4
4.8
2.8
4.2
4.2
6.0
4.1
4.5
2.6
4.9
3.0
2.6
3.9
2.1
2.1
2.9
3.3
3.3
2.9
3.3
3.6
3.6
4.4
5.1
3.9
5.4
4.2
3.4
3.7
4.4
4.4
4.7
3.2
5.0
3.1
2.1
1.7
3.1




COMMENTS ON "INTEREST RATE POLICY AND THE INFLATION SCARE PROBLEM:
1979-1992"
R. Alton Gilbert
Marvin Goodfriend states his goal in writing this paper as
follows:

M

The goal is to distill observations to guide future

empirical and theoretical analysis of monetary policy with the
ultimate objective of improving macroeconomic performance."
expect this goal to be realized.

I

I expect references to this

paper as justification for using the federal funds rate as the
measure of monetary policy actions and references by people who do
theoretical and empirical work on credibility of central bank
commitments to the control of inflation.
One reason the paper will be attractive to others is that it
provides simple indicators of monetary policy actions and the
credibility of monetary policy.

The federal funds rate is the

indicator of monetary policy actions; a rise (decline) in the
federal funds rate indicates a tightening (easing) of policy.

The

long-term government bond rate is a measure of the credibility of
Federal Reserve policy to control inflation.

Changes in the

long-term rate are interpreted as changes in long-term inflation
expectations. A rise in the long-term government bond rate
indicates that the commitment of the Federal Reserve to contain
inflation in less credible to investors, and a fall in the rate
indicates greater credibility.

The paper uses these indicators of

policy actions and credibility of policy to examine the conduct of
monetary policy from October 1979 to 1992.
The purpose of my comments is to illustrate some problems in
applying these simple indicators of monetary policy in
interpreting specific events.

Problems with using the federal

funds rate as a measure of monetary policy actions are well known.
Monetary policy actions may be inflationary even if the federal
funds rate is rising, and policy actions may be deflationary even
if the federal funds rate is falling.

1

In many situations changes




Gilbert

in interest rates must be supplemented with data on monetary
aggregates to avoid errors in interpreting monetary policy
actions.
INDICATORS OF MONETARY POLICY IN 1983-84
The problem of judging whether a rise in the federal funds
rate indicates a tightening of monetary policy can be illustrated
by referring to events in 1983-84. Goodfriend refers to the rise
in the federal funds rate from 8.6 percent in May 1983 to 9.6
percent in August 1983 as a tightening of monetary policy.

This

rise in the federal funds rate was accompanied by a rise in the
long-term interest rate. Thus, while the rise in the federal
funds rate from May to August of 1983 is characterized as a
tightening of monetary policy, it was not effective in ending the
inflation scare.

It took an additional rise of 2 percentage

points in the federal funds rate in the following year to begin
reversing the rise in the long-term rate.
Table 1 supplements the data on interest rates from
Goodfriend's paper with growth rates of Ml. Growth rates of Ml
reflect the policy actions of the Federal Reserve, through reserve
requirements and the effects of policy actions on reserves.
During the spring and summer of 1983, Ml was rising rapidly.
was not a period of restrictive monetary policy.

This

The rise in the

federal funds rate from May to August of 1983 reflects the effects
of an economic expansion on interest rates, rather than the"
effects of restrictive policy actions of the Federal Reserve. The
federal funds rate peaked in the summer of 1984, when the Federal
Reserve brought money growth to a halt.
Reserve begin tightening monetary policy?

So when did the Federal
Adding information on

Ml growth indicates that June 1983 is too early.

2




Gilbert

INDICATORS OF MONETARY POLICY IN 1979-81
Table 2 presents the same data for the years 1979-81.

I am

going backwards in time in examining 1979-81 because
interpretation of movements in interest rates in this period is
more complex than in the 1983-84 period.
Goodfriend's Analysis
First I present my understanding of Goodfriend's analysis.
The Federal Reserve began tightening monetary policy in August
1979, but the large increase in the federal funds rate in March
1980 was necessary to gain credibility for the anti-inflation
policy of the Federal Reserve.

Declines in long-term rates after

March 1980 reflect greater credibility of anti-inflation policy.
Declines in the federal funds rate in May, June and July of 1980,
however, undermined the credibility of the Federal Reserve's
anti-inflation policy, causing the long-term rate to begin rising
again in July 1980. The Federal Reserve began tightening policy
again in August 1980, but long-term rates did not peak until
October 1981, two years after the Federal Reserve began its
anti-inflation policy.

Because the Federal Reserve temporarily

abandoned its tightening of policy in the spring and summer of
1980, it took longer than it might have to reverse the inflation
scare in 1981.
An Alternative View
My alternative explanation for movements of interest rates
in 1980 that focuses on the effects of the credit control policy,
which was imposed in March 1980 and removed in July 1980.

This

alternative explanation does not require assumptions about the
Federal Reserve gaining and losing credibility for its
1. For information on the credit control program and
its implications for the conduct of monetary policy in
1980, see Gilbert and Trebing (1981).

3




Gilbert

anti-inflation policy over periods of a few months.

This

alternative explanation does not require labeling monetary policy
as easing when the money stock was falling sharply and tightening
and the money stock was rising rapidly.
Imposition of credit controls caused a sharp drop in the
demand for credit, which caused the declines in short-term and
long-term interest rates. This decline in the demand for credit
was accompanied by a sharp decline in the money stock, especially
in April 1980, because the operating procedure used at the time
2
tended to be procyclical.
These declines in interest rates were
reversed in the summer of 1980, when the credit controls were
removed.

Again, with a procyclical operating procedure, the money

stock rose rapidly after credit controls were removed.
Using Ml as the indicator of monetary policy actions, there
is a much shorter lag between the tightening of monetary policy
and the peak of long-term interest rates in 1981.

During much of

the period from August 1980 through October 1981, Ml growth was
rapid.

The Federal Reserve did not consistently slow money growth

until May 1981, and long-term interest rates peaked in October
1981.
It is difficult to determine the degree to which the decline
in long-term interest rates that began in the fall of 1981
reflected lower expectations of long-run inflation.

The economy

was in a severe recession by the fall of 1981, and the decline in
demand for credit must have depressed long-term rates to some
extent.
CONCLUSIONS
I conclude my comments by considering their implication for
interest rates as indicators of monetary policy.

I find the the

2. See Gilbert (1985) for a general description of the
nonborrowed reserves operating procedure used from the
fall of 1979 to the fall of 1982.

4




Gilbert

federal funds rate to be an unreliable indicator of monetary
policy actions.

In some cases the federal funds rate rose (fell)

while the money stock rose rapidly (declined sharply).

When money

growth is included as an indicator of monetary policy, there is a
shorter lag between the tightening of monetary policy and the
following peaks of long-term interest rates.

Finally, changes in

long-term interest reflect forces in addition to changes in the
credibility of Federal Reserve anti-inflationary monetary policy,
including credit controls and recessions.

5




Gilbert

REFERENCES
Gilbert, R. Alton. "Operating Procedures for Conducting Monetary
Policy," Federal Reserve Bank of St. Louis Review (February
1985), pp. 13-21.

Gilbert, R. Alton and Michael E. Trebing. "The FOMC in 1980: A
Year of Reserve Targeting," Federal Reserve Bank of St.
Louis Review (August/September 1981), pp. 2-22.

6




Gilbert

Indicators of Monetary Policy, 1983-84

Month
1983

January
February
March
April
May
June
July
August
September
October
November
December

1984

January
February
March
April
May
June
July
August
September
October
November
December

Federal
funds rate
8.68%
8.51
8.77
8.80
8.63
8.98
9.37
9.56
9.45
9.48
9.34
9.47
9.56
9.59
9.91
10.29
10.32
11.06
11.23
11.64
11.30
9.99
9.43
8.38

30-year
government
bond rate

Annual growth
rate of Ml

10.63%
10.88
10.63
10.48
10.53
10.93
11.40
11.82
11.63
11.58
11.75
11.88

8.94%
14.15
16.53
11.34
14.21
9.78
12.59
6.31
6.28
12.59
3.29
3.28

11.75
11.95
12.38
12.65
13.43
13.44
13.21
12.54
12.29
11.98
11.56
11.52

9.35
3.72
8.76
8.45
5.29
9.08
1.57
0.44
8.04
-1.10
8.00
10.80




M

** tno

I ro M
oo <r» A oo oo

0 > M O O t O P » 0 0 0 0 * > 0 > C h t f l

V J O N J S I O O O O M C O U I H P C O

O vo CJ

O O l ^ M v l C O ^ C O v J C D H f O
C h O C D t n K J t o e o O N O U J t J O

Ul H M H

| k U U 1 U O K ) O O O J ^ U H O t

M K) M P» I
fO CO *> CO ** *> -J

K J I O H M M O ^ D O M I O I O O

• ^ U } O > ^ H U 1 V O O \ t O O \ 0 0 M
U i U i O O « O < * 4 v 0 O i O O t 0 O 4 *

N J H O ) N 1 K ) ^ O K ) K ) O U Q )

U U ^ ^ ^ U ( O U U K ) K } K )

O t A > t 4 U X > X * t < - 4
o o ) c c c p » ^ d p » a ) p )
c t » d » Q H - 0 K M H C r 3
O f t C <
»
P- O h C
I T ( D CO
H ? C H
ID p rt
pi h
CD
><

O O U I K J O ^ V O V O O S I N J ^ U

(D CD

&&

(D O
o <
<D CD

O 5S

10ODOOOOO>O**IOC»MP»OO
O U l H v J H l J v J 0 0 P \ O U I O

(D 0>
c p» »d P*
CD
HO h C
h- 3* C Pi
pi h

> c c x >x

O W
O CD
f t H3 c c
O f t C *<
t r i D CD
(0 3
H
&
«D

U U O C O O O O H U l s J v J ^ O O

Z
O
<
(D
g
&
CD
h

00
O

K ) t J U 1 ( f l v J v O U ) 0 0 U 1 ^ U l i O

D
CD
O
CD
g
&
ID
h

00

J

C

A

I

C

J

M

rt C K
ID «
3 ct
&
CD
*1

O

O

CD

O

O

O

O

O

O

p- O h C
p ? d
»
pi h
H *<
*<

n ( D C c c p i » d p i c D p i
O
tr
ID
*

v O L J t / i o o r o i o o o o o t t J O o t
O H U l N j ^ ^ U O K J O O ^ i O
dp

H O I O S J N J K I O ^ O O V O ^ M

M U J O O M t O t O v O M O O O t O
t O O U 1 < J O O C j r o \ 0 O O U l O 4 *
dP

OOlOV0OOOOOOVOv0\0lOOO

dp

« O M - J * > \ 0 ^ K > r O O O O O

C

D 55
CD O
O
<
CD
w CD
w
a 3
& &
CD CD
•1 h

M

O

ft 3
CD C
pi
O M

0* io
O O l*i
a <: o
a (D i
h »<
n 3
•
Pi 3 P»
f t CD * 1
CD 3
ft

hi pi
PI H»
ft
CD

3 CD
a
a
n CD

-J
vO
I
00

SO

o
rPo

o
3
CD
ft
P»
n
•<

X

M
3
a
Po
pi
ft
o
h
a

ro

CD
n
ft

tr

o
p-

INTEREST RATE OPERATING PROCEDURES OF FOREIGN CENTRAL BANKS
John Morton and Paul Wood

The interest rate operating procedures employed by central banks in the
major industrial countries in implementing monetary policy over the past
decade have varied considerably.

This variation has not only been among

countries at any moment in time, but also often within countries over
time.

The main aim of this paper is to investigate the experience of

these countries with various interest rate operating procedures, both
descriptively and in terms of statistical measures, in light of their
stated intentions and goals in terms of monetary policy implementation.
The first section of the paper lays out some possible criteria for
selection of an interest rate operating procedure, and some of the
implications of choosing different monetary policy targets. The second
section describes in more detail the recent experiences of six industrial
countries (Japan, Germany, France, the United Kingdom, Switzerland, and
Canada) with respect to interest rate operating procedures. The third
section presents and discusses some statistical measures of this
experience.
INTEREST RATE OPERATING PROCEDURES
Questions regarding the desirable properties, characteristics, and
criteria of selection of an interest rate operating procedure can be
usefully divided into two broad categories. The first category involves
the choice of target variable, or variables, at which interest rate
policy is aimed. These targets could be either ultimate macroeconomic
targets, such as output, unemployment, or inflation, or intermediate
targets, such as growth of one or more monetary aggregates or exchange
rates. The second general category of factors involves the choice of a

1. Division of International Finance, Board of Governors of the Federal
Reserve System. We would like to thank Karen Johnson and Ralph Smith for
useful comments, and James Heil for research assistance.




Morton and Wood

particular interest rate operating procedure, given the selection of a
particular target.

In practice, of course, both categories are

interrelated.
The choice of a policy target, which then becomes the focus of the
interest rate operating procedure, can depend on a variety of factors.
On a theoretical level, these would include prominently the likely source
of economic disturbances, focussing on whether disturbances originate
mainly from real or monetary factors, or from domestic or foreign
sources.

The possible choice of a monetary target would additionally

depend importantly on the stability of monetary relations, in particular
the demand for money.

Political or institutional constraints can also be

important in deciding on an exchange rate target, the most prominent
recent example being possible membership in the European Monetary System
(EMS).
The individual country experiences described in the next section,
while exhibiting substantial variations, do appear to show some general
trends with regard to choice of policy targets.

Over the past decade

there appears to have been a general, though not universal, movement away
from monetary targets and, for some countries, a movement towards
exchange rate targets.

This trend is most evident, of course, among EMS

members, other than Germany.

There also appears to have been a general

tendency to adopt a more flexible, ad hoc approach to targets, with a
variety of targets having shifting relative weights under different
circumstances.
The likely relationship between choice of a particular policy
target and the variability of interest rates used to achieve that target
is unclear.

The outcome would depend on such factors as the main source

of economic disturbances and the strictness with which policy targets are
adhered to.

In general, an intermediate target, such as monetary

aggregate growth, might be more strictly followed, in the sense of a
change in the target variable triggering a prompt and at times large
change in interest rates.

In this case, adoption of such a procedure

might be expected to result in a more variable interest rate path.
However, such a procedure might, over time, bring greater stability,




- 2 -

Morton and Wood

eliminating possible discretionary swings in policy, and reducing
interest rate variability.
Given the choice of a particular target, the question arises as to
the interest rate procedures used to achieve that target.

Assuming that

at any moment a particular interest rate level could best achieve some
desired level of the target, an interest rate implementation procedure
would seem to be desirable if it could achieve that interest rate level,
and undesirable if it could not.

Thus, it would appear desirable to have

an interest rate procedure that was "flexible," in the sense of allowing
prompt and, if needed, large changes in interest rates.

Conversely, an

"inflexible" system, which somehow hindered interest rate changes, would
appear undesirable.
As demonstrated in more detail in the next section, several
countries have, over the past decade, moved to more flexible--in the
sense defined above--interest rate operating procedures.

These changes

have sometimes been confined to the interest rate operating procedures
themselves (Germany and the United Kingdom) or have taken place as part
of a wider change to a more market-oriented monetary policy framework
(Japan and France) . Movement to a more flexible interest rate operating
procedure may involve moving to less of a reliance on the discount rate,
since discount rate changes may be hindered by concerns over announcement
effects.

Despite the seeming general desirability of a "flexible"

interest rate operating procedure, monetary authorities may at times be
reluctant to adopt procedures which are seen to lead to "unstable," or
overly volatile interest rates, showing a preference for a more stable
interest rate path.

INDIVIDUAL COUNTRY EXPERIENCES
Japan
The Japanese financial system has traditionally been characterized
by a high degree of government control and restrictions.

In terms of

monetary policy, authorities have relied heavily on discount rate changes
and quantitative controls.




Over the past decade, this system has

- 3 -

Morton and Wood
undergone substantial liberalization, featuring new financial
instruments, more international openness, interest rate decontrol, and
less reliance on the discount rate and more reliance on open market
operations.
A variety of pressures have encouraged the move to financial
liberalization.

An increase in government deficits starting in the mid-

1970s eventually led to the breakup of the system whereby banks were
forced to accept government debt at below-market rates. The first real
open market to emerge in the 1970s was the gensaki market, a repurchase
market for government bonds. Market pressures to break down restrictions
in domestic financial markets also came from abroad, particularly the
United States. Various foreign exchange restrictions were reduced
starting in 1980, and the Euroyen market grew rapidly in subsequent
years.
The process of financial liberalization, which started later in
Japan than in any other of the major industrial countries, with the
possible exception of France, gained real momentum after the mid-1980s.
Banks had been permitted to issue negotiable certificates of deposit in
1979.

In 1985, money market certificates, yen-denominated bankers

acceptances, and large denomination time deposits were introduced.
Treasury bills first appeared in 1986, and commercial paper in 1987.
At the beginning of the 1980s, the Bank of Japan relied heavily on
the discount mechanism to regulate credit conditions. Most interest
rates were tied, formally or informally, to the discount rate.

The Bank

supplied credit to the market almost exclusively through changes in
discount window lending.

The discount rate was kept well below market

interest rates, meaning that there was always an excess demand for
discount borrowing.

The Bank of Japan decided each day how much discount

lending to make, both in total and to individual banks, effectively
rationing credit.
lending.

The Bank also imposed ceilings on the growth in bank

However, as more and more financial transactions took place at

market-determined interest rates, and non-bank sources of credit grew in




- 4-

Morton and Wood

importance, this system became increasingly less efficient.

In

response,, starting about 1988, Japanese authorities adopted a policy of
relying increasingly on open market operations and less on discount
window lending as a way of supplying reserves to the banking system.
They also looked to day-to-day control of the overnight rate as the main
interest rate control mechanism, giving less importance to the discount
rate.

These shifts in operating procedure are still incomplete and
3
ongoing.
The Bank of Japan appears to have maintained an interest rate
target rather than a monetary aggregate target over the past decade,
i.e., monetary authorities appear to have varied interest rates in
response to changes in macroeconomic targets, such as output and
inflation, rather than money supply growth.

The Bank of Japan announces

H

forecasts'* for M2+CD growth, but does not appear to treat these as

targets.

For one thing, the forecast is for four-quarter growth rates

but is only announced at the beginning of the end-point quarter, meaning
that much of the forecast is already history.

Also, in recent years, as

money growth has fluctuated sharply, M2+CD forecasts have been varied in
line with actual data, rather than changing more slowly, as would more
likely be the case were they treated as targets.
Germany
Entering the 1980s, the Bundesbank relied primarily on the Lombard
window to extend credit to banks.

Borrowing at the Lombard window is

done at the initiative of banks.

The rate on those loans, the Lombard

rate, is a rate set by the Bundesbank and adjusted infrequently, often in
conjunction with an equal movement in the discount rate.

During the

first half of the decade, the call money rate tended to be near the
Lombard rate.

When bank reliance on Lombard borrowing became too heavy,

the Bundesbank would attempt to gain control over Lombard borrowing by

2. By the end of 1991, over 60 percent of city bank deposits carried
market rates of interest.
3. The Bank of Japan still relies mainly on varying the amount of
discount window lending, rather than open market operations, for daily
adjustments of credit availability.




- 5 -

Morton and Wood
putting quantitative limits on what banks could borrow at the Lombard
window or by suspending the Lombard rate and substituting a special
Lombard rate that could be changed daily.

The Bundesbank began to use

repurchase agreements (RPs) in the early 1980s and increasingly met bank
liquidity needs with RPs beginning in 1983. The rates on RPs generally
exceeded the Lombard rate during this time.
In 1985, in response to excessive use of the Lombard window, the
Bundesbank raised the Lombard rate above the RP rate to make the Lombard
window the borrowing source of last resort for banks.

Since 1985, the

call money rate has tended to track the RP rate, somewhere between the
discount rate and the Lombard rate. The RP rate has been moved around
more than the Lombard rate, which is only adjusted a few times a year.
The Bundesbank effectively targets the call money rate.

Funds are

injected into and withdrawn from the market through weekly auctions at
which RPs are tendered with a maturity of approximately one month, with
two-month RPs also offered on occasion.

Sometimes RP funds are offered

at a fixed rate announced in advance. More commonly, the funds are
auctioned at a rate sufficient to clear the market.

The Bundesbank sets

the quantity tendered after it observes the bids, so it has some control
over the repurchase rate. The Bundesbank can also inject funds through
emergency short-term RP tenders and by moving Treasury funds held at the
Bundesbank into commercial banks.
Germany has had a monetary target continuously since 1975. For
most of this period, the target was stated in terms of central bank money
(CBM), a weighted sum of the components of M3.
itself became the targeted aggregate.

Starting in 1988, M3

In terms both of success in

achieving targeted aggregate growth and the apparent importance attached
to this goal, the Bundesbank seems to have been relatively committed to
monetary aggregate growth as a policy goal. A complication to monetary
targeting was introduced by the monetary union of eastern and western
Germany in 1990.

Since then, interpretation of monetary aggregate

changes has become considerably more difficult and ambiguous.




- 6-

Morton and Wood

France
More than any other major industrial country, with the possible
exception of Japan, France had tightly regulated financial markets moving
into the 1980s.

There were extensive foreign exchange controls, a number

of financial instruments were either officially or effectively
prohibited, many interest rates (including bank deposit rates) were
regulated, and monetary control was exerted largely through ceilings on
bank credit growth.

Starting in the early 1980s, this highly regulated

system began.to be liberalized and deregulated.

The liberalization

process was prompted partly by market pressures, arising from financial
innovations and foreign competition, and partly from a deliberate
government policy aimed at increasing market efficiency.

Major events in

this liberalization process included the 1982 introduction of short-term
bond mutual funds (SICAVs), which provided strong competition to
regulated-rate bank deposits, the introduction of negotiable certificates
of deposit and commercial paper in 1985, and opening of the short-term
treasury bill market to non-financial corporations and individuals in
1986.
Although the process of financial liberalization has continued at
various rates throughout the past decade, a key change in the procedures
for implementing monetary policy took place in January 1987.

The

previous system of quantitative controls on bank asset growth was
abolished, to be replaced by reserve requirements on liabilities, and the
daily setting of the call-money market rate was also ended.
The interest rate operating procedure established in 1987, which
still in substance prevails, involves two key official interest rates.
These are the intervention rate and the 5- to 10-day repurchase rate.
Under normal circumstances, the interbank rate is between these two

4. The "encardrement du credit" credit ceiling system is described in
Marc Quintyn, "From Direct to Indirect Monetary Policy Instruments: The
French Experience Reconsidered," IMF Working Paper. March 1991, pp. 5-7.
5. Technically, credit growth ceilings were ended in 1985. However,
they were replaced by a marginal reserve requirement system that in large
part served as a functional equivalent. This system was abolished in
January 1987.




- 7 -

Morton and Wood
rates, with the intervention rate acting as a lower bound and the
repurchase rate an upper bound.

The main instrument used to influence

the interbank interest rate is the intervention rate (the rate at which
repurchase funds are offered approximately once a week by the Bank of
France).

The Bank of France also controls the quantity of reserves

allocated at the 5-10 day repurchase rate, though borrowing at that
facility is done at the initiative of individual private banks.
French monetary authorities have maintained monetary targets
continuously since 1977.

The strength of the French commitment to

these targets appears to have varied over time, but in general has not
been as strong as in some other countries, such as the United Kingdom.
More important, especially in recent years, has been the French
commitment to an exchange rate target, formally an EMS parity, in
practice the mark.

France joined the EMS in 1979, but devalued the franc

within the EMS three times in the 1981-1983 period.

A key event in

France's commitment to an exchange rate target was the 1983 decision of
the Mitterrand government, after much internal debate, to remain within
the EMS.
1987.

The last realignment of the franc within the EMS was in January

Since then, maintaining a stable franc-mark exchange rate has

clearly been the paramount goal of French monetary policy.
United Kingdom
Over the past decade, monetary policy operating procedures have
varied in the United Kingdom.

At times, authorities appear to have

operated mainly with an interest rate target, varying interest rates in
response to macroeconomic goals, such as output or inflation.

At other

times, monetary aggregate targeting has been given priority, and, more
recently, an exchange rate target has been most important.
The Bank of England's interest rate operating procedures are
conducted mainly through its money market dealing rates. These are the
rates at which the Bank supplies liquidity daily to the market, primarily
through open market transactions in commercial bills with the discount

6. The targeted aggregate has alternated between M2 to M3, with the
aggregate undergoing a major redefinition in 1987.




- 8-

Morton and Wood

houses.

The Bank operates in four maturity bonds, ranging from 1-14 days

to 64-91 days.

In practice, the money market dealing rates operate much

like discount rates in that they are set by the Bank and changed only
infrequently.
It should be noted that in 1981, when the Bank instituted the
system described above, "it was hoped that this mechanism would provide
scope for a reasonable degree of flexibility in short-term interest
rates."

Under the previous system, a Minimum Lending Rate, at which

the Bank provided funds to the market, was officially set and remained
unchanged for long periods of time.

It was hoped that the new money

market dealing rates would be partially set by market forces, and vary
from day to day.

Thus, the Bank's interest rate intentions would be

revealed more indirectly, and the "announcement effect" of official
lending rate changes would be lessened.

As actual events unfolded, this

o

goal could not be achieved.
The Bank remained the dominant force in
the bill market, supplying significant funds each day, with its lending
rates achieving a de facto discount rate status.
British monetary authorities first adopted a monetary target in
1976.

The importance attached to this target increased sharply with the

start of the Thatcher government in 1979.

However, over time, various

problems with monetary targeting--in particular financial innovations
which seemed to distort monetary aggregate growth--led to a de-emphasis
of monetary targets, although they still officially remain in place.
Important stages in the process of moving away from monetary targets
included the use of more than one monetary aggregate as a target
(starting in February 1982), use of a MO as a target, MO consisting
almost entirely of currency, and thus not under the active control of
monetary authorities (starting in February 1984), and significant and
persistent overshooting of the original M3 target (starting in March
1985).

7. A. L. Coleby, "Change in Money-Market Instruments and Procedures in
the United Kingdom," in Changes in Monev-Market Instruments and
Procedures: Objectives and Implications. Bank for International
Settlements, March 1986, p. 200.
8. For a description of this episode, see Coleby, pp. 201-204.




- 9 -

Morton and Wood
U.K. monetary authorities have also at various .times used interest
rate changes in order to achieve an exchange rate objective.

In the late

1970s and early 1980s, the exchange rate of greatest interest to
authorities was that of the pound against the U.S. dollar. However,
increasingly, the exchange rate of the pound in terms of EMS currencies,
and particularly the German mark, became the main objective.

For a

period in 1987-1988, British authorities maintained an unofficial but
strong effort to keep the pound-mark exchange rate in a narrow range.

In

retrospect, this experiment was judged to have had an unfavorable
outcome. Upward exchange market pressure on the pound against the mark
led to an easing of monetary policy and reduction in interest rates that
provided an undesirable inflationary stimulus to the domestic economy.
Since October 1990, the pound has been an official member of the EMS's
exchange rate mechanism.

Experience in this later period has generally

been more successful.
Switzerland
The Swiss National Bank's policy has been geared to two main
objectives over the past decade, monetary targets and the exchange rate
of the franc (mainly against the dollar early in the period, and against
the mark in recent years).

The relative importance of these two

objectives appears to have varied over time. The Swiss have maintained
monetary targets since 1975. The target has been stated in terms of
adjusted central bank money (essentially the monetary base) since 1980.
Between 1982 and 1987, Swiss monetary authorities came quite close to
achieving their targeted monetary growth rate. However, both before and
after this interval, substantial deviations from targets occurred.
In the 1978-1979 period, there developed a clear conflict between
exchange rate and monetary target objectives. There was strong upward
pressure of the franc (mainly against the dollar), which would have
required a significant easing of monetary policy and lowering of interest
rates to counter.

However, such an easing was likely to lead to a

significant overshooting of the monetary target.

Swiss authorities chose

to put greater weight on the exchange rate objective in this instance.




- 10 -

Morton and Wood

This resulted in a surge of money growth, a temporary abandonment of the
monetary target in 1979, and a subsequent increase in inflation to what
was, by Swiss standards, alarming levels.
complicated by several factors.

Analysis of this episode is

In particular, the oil price shock of

1979, which contributed to inflationary pressures, and an upward shift in
money demand, much of it coming from abroad.

Nonetheless, Swiss

authorities appear to have concluded that it was a mistake to allow an
overshooting of the monetary target to the extent that took place.
The more recent period of substantial deviation from monetary
targets was triggered by two changes introduced in 1988 which
substantially shifted the demand for base money.

First, a new electronic

interbank payments system was established which sharply lowered
commercial banks' need for clearing balances at the Swiss National Bank.
Secondly, the authorities modified cash liquidity requirements, moving
from an end-of-month to monthly average measure and lowering overall
requirements.

The net result of these changes was a substantial

undershooting of the central bank money target over the 1988-1990 period,
difficulty in interpreting the actual degree of ease or tightness of
monetary policy, and an increase in inflationary pressures in 1991-1992.
Despite this recent difficulty with monetary targeting, there remains a
reluctance by some in Switzerland to move fully to an exchange rate
9
target.
This decision is increasingly taking the form of possibly
joining the EMS, effectively pegging the Swiss franc to the mark.

This

possibility appears strengthened by the recent referendum vote in favor
of IMF membership, and the subsequent announcement by the Swiss
government that it would apply for EC membership.
The Swiss National Bank has two means of affecting market
liquidity.

The first, and most important, involves foreign exchange

swaps, usually of 1- to 3-month maturity.

The second involves moving

government balances into and out of the commercial banking system.

The

Swiss National Bank maintains both a discount rate, set at the Bank's

9. See, for example, G. Rich. "The Orientation of Monetary Policy and
the Monetary Policy Decision-Making Process in Switzerland," Swiss
National Bank, 1991.




- 11 -

Morton and Wood
discretion and changed only infrequently, and a Lombard rate which, since
May 1989, has been computed daily by a formula which sets the rate 200
basis points (rounded to the nearest 1/8 percent) above a reference call
money rate.

This latter change appears to have been motivated by a

desire to make the Lombard rate a penalty rate, and Lombard lending truly
an exceptional source of bank liquidity, as well as avoiding announcement
effects from Lombard rate changes.
Canada
Canadian monetary policy has been dominated over the past decade
mainly by the important influence of, and need to respond to, conditions
in the United States. Given the close integration of U.S. and Canadian
financial markets, this has usually meant that Canadian short-term
interest rate changes mirror those in the United States fairly closely.
Unlike the situation of EMS members, however, there has never been a
formal or official commitment to keep the Canadian dollar-U.S. dollar
exchange within some specified narrow range. It has been a general
policy of the Bank of Canada to pursue a "leaning against the wind"
intervention policy, moderating but not totally resisting exchange rate
movements. In addition, monetary authorities have explicitly recognized
the trade-off between exchange rate and interest rate changes, with, for
example, a potentially inflationary depreciation of the Canadian dollar's
foreign exchange value leading to some compensating tightening of
monetary policy through higher Canadian interest rates.
Canada first adopted an official monetary target in November 1975.
However, various problems--including financial innovations which
distorted monetary aggregates growth, and conflicts with other important
targets, particularly the exchange rate--led to the abandonment of
monetary targeting in November 1982. More recently, since 1990 targeting
of another type has been introduced. Canadian financial authorities have
announced a multi-year series of declining inflation targets. Although
money aggregate growth (of M2) is to be used as one indicator guiding
policy, it has been made clear that actual inflation is the main target.




- 12 -

Morton and Wood
In March 1980, the Bank of Canada adopted a formula for
determining its discount rate. More specifically, the discount rate is
determined by a pre-announced rule based on the outcome of the weekly 3month Treasury bill auction.

Since this procedure eliminates the

"announcement effect" of discount rate changes, it provides the maximum
degree of flexibility in implementing interest rate policy.
STATISTICAL MEASURES OF VOLATILITY
The preceding two sections suggest that, both in theory and in the views
of central bank officials, a desirable property of an interest rate
operating procedure is to be "flexible.H Although it is difficult to
arrive at a simple, unambiguous definition of this term, its core meaning
appears to involve the ability to change interest rates promptly and
fully when needed. This could involve both day-to-day changes, and
cumulative adjustments over time. On the other hand, central bankers
also may wish to have a relatively smooth path of key interest rates,
either for political reasons or in order for policy to be more easily
predicted by market participants. Therefore, it is ambiguous whether
interest rate volatility is good or bad.
The discussion in the previous sections suggests several
hypotheses as to the relative variability of interest rates under
differing operating procedures and different policy regimes, although in
some cases expected results are ambiguous. The switch to an interest
rate operating procedure that is more market-oriented and less tied to
official rates, such as that undertaken by Germany in 1985, might be
expected to lead to a somewhat more variable interest rate path.
Similarly, a general liberalization of financial market structure and
monetary policy operating procedures, such as that which took place* in
Japan and France around the mid-1980s, might also be expected to increase
interest rate variability. Adoption by Canada of a formula discount
rate, where official interest rate changes are unhampered by announcement
effects, might be expected to result, other things being equal, in
relatively greater interest rate variability than in other countries.




- 13 -

Morton and Wood
The implications for interest rate variability of differing
monetary policy target variables is a priori more uncertain. Thus, the
expected relationship between interest rate variability in countries with
a relatively strong commitment to a monetary aggregate target, such as
Germany and Switzerland, and countries without monetary aggregate
targets, such as Japan and Canada, is unclear, although there might be a
weak presumption of somewhat greater interest rate variability in
countries following a monetary aggregate target.

There is a similar

uncertainty about the role of exchange rate targets. Here, the main
division would be between EMS members (France and, since 1990, the United
Kingdom) and countries with no formal exchange rate commitment (Japan,
Canada, and Switzerland).

The situation here is further complicated by

the fact that, even without a formal exchange rate arrangement, some
countries still tie their monetary policies strongly at times to exchange
rate targets (Canada to the U.S. dollar and Switzerland to the mark).
In tables 1, 2, and 3, we show measures of daily interest rate
volatility for the six countries in our study.

For five of the

countries, we divided the sample where there was a significant change in
the operation of monetary policy.
for Canada).

(We did not find any break in policy

In Japan, the break we chose is at the beginning of 1985,

which marked the approximate beginning of rapid financial liberalization.
For Germany, the break is in February 1985, when the Bundesbank shifted
to relying on RP agreements rather than the Lombard window as the primary
source of liquidity for the banking system.

In France, the break is at

the beginning of 1987, when French financial markets were liberalized and
the Bank of France switched from direct credit allocation to open market
operations.

In the United Kingdom, the break is in October 1990, when

sterling entered the exchange rate mechanism of the EMS. For
Switzerland, the break is in 1988, when Swiss cash liquidity requirements
were changed to a monthly-average basis from a month-end basis.
The measure of daily volatility shown in Table 1 is the standard
deviation of interest rates on a daily basis.

For each of these

countries, the interest rate path over time has either a substantial
trend or prominent cycles, so that standard deviation measures over long




- 14 -

Morton and Wood

periods are dominated by the large differences from mean, rather than
day-to-day changes, and are thus relatively invariant for different
frequencies.

Because of this, we computed the standard deviation of

daily data around the monthly mean for each month, then averaged the
monthly standard deviations for each sub-period, and that is shown in
Table 1.

In Table 2, we show the standard deviation of daily changes in

interest rates for each sub-period.

The measure shown in Table 3 is the

average absolute daily change in interest rates.
The first comparison we can make is between volatilities of
overnight and three-month interest rates.

For each of our measures and

for every country and every time period, overnight interest rates are
more volatile than three-month interest rates.

That would be expected if

overnight rates reflect temporary liquidity pressures in addition to
changes in monetary policy.

We also find that countries with the least

volatility in overnight rates also tend to have the least volatility in
three-month interest rates.
Comparing across countries, we find that Japanese interest rates
have been less volatile at both the overnight and three-month maturities.
German interest rates have generally been the next least volatile,
followed closely by those of France, which has set its monetary policy to
stabilize the franc-mark exchange rate especially in the more recent
period.

Interest rate volatility in the United Kingdom tends to be

somewhere in the middle of this group of countries, while volatility has
been the highest in Switzerland and Canada.

We can thus find no clear

division in interest rate volatility along the lines of countries that
have monetary aggregate targets versus those that do not, because Japan
and Canada (two countries without monetary targets) are on opposite ends
of the volatility spectrum.

Likewise, there is no clear division between

EMS and non-EMS countries.
Comparing across time periods, we see that standard deviations
around monthly means have declined for all the countries and all
maturities except for the Japanese three-month interest rate.

However,

the three-month rate used here (the CD rate) is only available starting
June 1984, so the first sub-period has only seven months of data for that




- 15 -

Morton and Wood

rate.

The standard deviation for the German overnight -rate was almost

unchanged.

The most striking decline in interest rate volatility is also

the most predictable.

Swiss overnight interest rates became much less

volatile after the switch to monthly-average liquidity requirements which
reduced the sharp increases in overnight rates that tended to occur at
the end of each month under the previous regime of month-end reserve
requirements.
The measure of average absolute change shown in Table 3 shows
declines in volatility in the latter sub-periods except for Germany and
the United Kingdom.

It is somewhat surprising that interest rate

volatility has decreased in the more recent sufc-periods for Japan and
France, which moved to more flexible interest rate operating procedures.
However, it is possible that the deepening of financial markets in the
latter period of financial liberalization has contributed to lower
interest rate volatility.

In Germany, the move to a more flexible

operating procedure in 1985 has been accompanied by slightly more
volatility by the measure in Table 3.




- 16 -

Morton and Wood

REFERENCES

Bank for International Settlements. Changes in Money-Market Instruments
and Procedures: Objectives and Implications. BIS, March 1986.
Batten, Dallas S. and Michael Blackwell et. al.
Policy in the Major Industrial Countries,"
10, July 1990.

"The Conduct of Monetary
IMF Occasional papers. No.

Bernanke, Ben and Frederic Miskin. "Central Bank Behavior and the Strategy
of Monetary Policy: Observations from Six Industrialized Countries,"
Mimeo, 1992.
Crow, John W. "The Bank of Canada and the Money Market,"
April 1989.

Bank of Canada,

Deutsche Bundesbank. "The Deutsche Bundesbank: It's Monetary Policy
Instruments and Functions, 3rd Edition," Deutsche Bundesbank Special
Series. No. 7, July 1983.
Frowen, Stephen F. and Dietman Kuth ed. Monetary Policy and Financial
Innovations in Five Industrial Countries: The U.K.. the USA. West
Germany. France and Japan. St. Martin's Press, 1992.
Kasman, Bruce and Anthony Rodrigues. "Financial Reform and Monetary
Control in Japan." FRBNY Quarterly Review. Autumn 1991.
Kneeshaw, J. T. and P. Van den Bergh. "Changes in Central Bank Money
Market Operating Procedures in the 1980s," BIS Economic Papers. No.
23, January 1989.
Osugi, K. "Japan's Experience of Financial Deregulation Since 1984 in an
International Perspective," BIS Economic Papers. No. 26, January 1990.
Quintyn, Marc. "From Direct to Indirect Monetary Policy Instruments:
French Experience Reconsidered," IMF Working Papers. March 1991.

The

Rich, G. "The Orientation of Monetary Policy and the Monetary Policy
Decision-Making Process in Switzerland," Swiss National Bank, 1991.
Suzaki, Yoshio, Akio Kuroda and Hiromichi Shirakawa. "Monetary Control
Mechanism in Japan," Bank of Japan Monetary and Economic Studies. Vol.
6 No. 2, November 1988.
Temperton, Paul. U.K. Monetary Policy:
Martin's Press, 1991.




- 17 -

The Challenge for the 1990s. St.

- 18 -

Data Series
Overnight
Germany:
Japan:

Frankfurt Interbank Call Money Rate

Tokyo Unconditional Lender Rate

France:

Paris Day to Day Money Rate

United Kingdom:
Canada:

U.K. Call Money Rate

Canadian Day to Day Money Rate

Switzerland:

Zurich Call Money Rate

Three-Month
Germany:
Japan:
France:

Frankfurt Interbank Loan Rate

Rate on Certificates of Deposit (Secondary Market)
Paris Interbank Rate

United Kingdom:
Canada:

Canadian Finance Company Paper

Switzerland:




Interbank Sterling Interest Rate

Swiss Interbank Rate

- 18 -

Morton and Wood
Table 1
Average of Standard Deviations of Daily Data Around Monthly Means
Overnight rate

Three-month rate

1980-85

0.25

0.16

1985-92

0.24

0.08

1980-84

0.18

0.01

1985-92

0.13

0.07

1980-86

0.30

0.22

1987-92

0.23

0.14

1980-90

0.43

0.25

1990-92

0.36

0.18

1980-92

0.58

0.23

1980-87

3.56

0.24

1988-92

0.20

0.18

Germany

Japan

France

United Kingdom

Canada

Switzerland




19 -

Morton and Wood

Table_2
Standard Deviations of Daily Changes in Interest Rates
Overnight rate

Three-month rate

1980-85

0.38

0.13

1985-91

0.32

0.06

1980-84

0.13

0.00

1985-91

0.10

0.04

1980-86

0.23

0.14

1987-91

0.20

0.09

United Kingdom
1980-90

0.40

0.18

1990-91

0.35

0.11

1980-91

0.58

0.16

1980-87

6.98

0.20

1988-91

0.18

0.11

Germany

Japan

France

Canada

Switzerland




- 20

Morton and Wood
Table 3
Average Absolute Change of Interest Rates (Daily Data)
Overnight rate

Three-month rate

1980-85

0.110

0.063

1985-92

0.118

0.035

1980-84

0.073

0.047

1985-92

0,063

0.018

1980-87

0.121

0.063

1987-92

0.114

0.051

1980-90

0.226

0.093

1990-92

0.233

0.063

1980-92

0.536

0.087

1980-87

0.884

0.105

1988-92

0.375

0.059

Germany

Japan

France

United Kingdom

Canada

Switzerland




- 21 -

Overnight and Official Discount Rates for Japan
(Monthly Average)
Percent

- = OVERNIGHT
- = OFFICIAL DISCOUNT
8

H6
to

-i 4

H 3

1
1982



1

I
1983

1984

1
1985

1

1
1986

1987

1
1988

1

I
1989

1990

1991

1992

Overnight, Discount, Lombard, and RP Rates for Germany
(Monthly Average)
Percent

= OVERNIGHT
= DISCOUNT
= LOMBARD
= RP

12
11
10
9
8

rt

o o
rt 3

7
6
5

3
v

1
1982



1
1983

1

X
1984

1985

J

1
1986

i

1987

1
1988

1
1989

1
1990

1
1991

1992

2

*:
o
o

Overnight and Money Market Intervention Rates for France
(Monthly Average)
Percent

rr
O O
ST 0

rt 3

ho

o
o

,1982



1983

1984

1985

1986

1987

1988

1989

1990

1991

1992

Overnight and Money Market Dealing Rates for the United Kingdom
(Monthly Average)
Percent

o
03

0

o
o

1982



1983

1984

1985

1986

1987

1988

1989

1990

1991

1992

Overnight and Lombard Rates for Switzerland
(Monthly Average, Month End, Respectively)
Percent

-OVERNIGHT
= LOMBARD

15
14
13
12

t

11
M

H 10
&

»i
rt
O O
ST 0
0)
rt 3
a
Ln

ON

O
O

a

1982



1983

1984

1985

1986

1987

1988

1989

1990

1991

1992

Overnight and Discount Rates for Canada
(Monthly Average)
Percent

»i
rt
O O
rt 3

<^

1982



1983

1984

1985

1986

1987

1988

1989

1990

1991

1992

s:
o
o
a-




A COMPARISON OF MONETARY POLICY OPERATING
PROCEDURES IN SIX INDUSTRIAL COUNTRIES
Bruce Kasman1

The institutional environments in which the central banks of the
industrial world operate have changed substantially since the mid1970s. Financial market liberalization, along with regulatory and
technological change, has altered the relationships between
central bank policy tools and objectives. Authorities have
responded to these changes by revising the techniques and
procedures they use to implement monetary policy. In Japan and
France, where far-reaching reforms of the financial system have
taken place, central bank operating procedures have been
substantially transformed. In countries where well-developed
capital markets existed earlier, the revisions in monetary policy
operating procedures have been considerably less dramatic.
As financial liberalization and innovation proceed, the
institutional settings of the central banks have become more
uniform. Although arrangements still vary across countries, this
convergence suggests that a comparison of central bank operating
procedures is now likely to be of greater relevance to policy
makers than at any time in the past.
An assessment of foreign practices may provide a
particularly useful perspective on the changing conditions
affecting the operations of the Federal Reserve's Open Market
Desk. A noticeable increase in banks' reluctance to borrow at the
Federal Reserve's discount window in recent years has at times
contributed to large daily fluctuations in the Federal funds rate.
Moreover, reductions in reserve requirements in 1990 and April of
this year have led to occasional conflicts between the Desk's
reserve management strategy and more volatile day-to-day
conditions in the funds market. With other central banks offering
a wide variety of alternative techniques for implementing policy

1.

Federal Reserve Bank of New York.
The author is grateful to Andre Bartholomae, Kevin Clinton,
Spencer Dale, David Longworth, Ann-Marie Meulendyke, Michel
Peytrignet and George Rich for providing useful comments and
information. Valuable assistance in preparing this paper was
provided by Matthew Maring.

Kasman

and a number currently operating in an environment of low,
nonbinding reserve requirements, an examination of operating
procedures followed by foreign central banks seems timely.2
This article describes monetary policy operating
procedures in six industrial countries — the United States,
Germany, Japan, the United Kingdom, Canada, and Switzerland. The
object is to shed light on central bank strategies elsewhere in
the industrial world and to compare them with the practices of the
Federal Reserve. As part of this review, particular attention is
given to the institutional environments in which central banks
operate. The intermediate and ultimate objectives of a central
bank, while important in an overall survey of monetary policy
transmission, are not discussed in any detail.
Our review suggests that basic central bank intervention
strategies are currently quite similar across the industrial
world. Nearly all the central banks analyzed use interest rate
operating objectives to guide their daily activities. In
addition, although the central banks employ different instruments,
they all implement policy principally through daily operations
supplying or absorbing reserves at market-determined prices.
The Federal Reserve and several foreign central banks are
also alike in having chosen to lower their reserve requirements in
recent years. In most cases, the foreign monetary authorities
have adjusted their operating procedures to accommodate this
change. Specifically, they have provided a more elastic intraday
supply of central bank reserves, largely through their credit
facilities. In this way, they limit any tendency for reduced
reserve margins to lead to higher day-to-day interest rate
volatility.
Our analysis suggests that some of the practices observed
abroad might be helpful in limiting the short-run volatility of
the federal funds rate in the United States. However, our
analysis also indicates that the volatility of the federal funds
rate, although higher since the 1990 cut in reserve requirements,




2. A good discussion of Federal Reserve operating procedures
following the reduction in reserve requirements can be found in
"Monetary Policy and Open Market Operations during 1991." This
Quarterly Review, Spring 1992, pp. 72-95.
-2-




Kasman

remains low relative to that of comparable rates in most other
countries. Moreover, we find no evidence that federal funds rate
variability, within its current range, is transmitted to other
money markets. Thus, the rise in interest variability that has
accompanied the reduction in reserve requirements in the United
States has probably not materially affected the monetary policy
transmission mechanism.
COMPARING OPERATING PROCEDURES IN SIX INDUSTRIAL COUNTRIES
Key Features of Central Bank Operating Procedures
A central bank must choose implementation procedures that enable
it to achieve its macroeconomic goals. Although the six central
banks considered in this article have different objectives and
operate under varied institutional environments, the key features
of their implementation strategies are currently quite similar.
All six central banks implement policy by controlling the
aggregate level of reserves available to the banking system.
Although they are not in a position to control movements in all
components of their balance sheets, particularly those related to
their function as banker to the government and their holdings of
foreign currency reserves, these banks currently have sufficient
information and operational leeway to neutralize the effects of
other activities and regulate the aggregate supply of reserves
with a high degree of control.
In managing the reserve position of the banking system,
central banks generally pursue short-run operating objectives.
Operating objectives link reserve management activities to the
intermediate and ultimate goals of policy and, in most countries,
are also used to signal central bank policy intentions to market
participants. Ideally, the authorities exert close control over
operating objectives.
Bank reserves have served as an operating objective but the
relationship between reserves and economic activity generally has
been viewed as too volatile for reserves to function as an
effective short-run guide to policy. Most of these central banks

-3-




Kasman

have instead geared their reserve management activities toward
short-term interest rate objectives.3 A wide variety of money
market interest rates are employed as operating objectives.
Nonetheless, influence over overnight interest rates is a goal
common to the daily activities of all six of these central banks.
Each of these countries has a well-functioning interbank money
market where individual banks trade reserves on deposit at the
central bank.4 If the aggregate supply of banking system reserves
does not correspond to demand, the cost of overnight funds in this
market is immediately affected.
Although central banks' reserve management activities give
them considerable control over short-term interbank rates, their
influence on interest rates must extend to maturities well beyond
overnight rates to affect economic activity. Central bank
influence over longer term rates is indirect and principally
determined by market forces. Through arbitrage, longer term rates
reflect market expectations of future short-term rates. A central
bank's leverage over longer term rates is obtained largely through
its influence on these expectations. By taking steps to
communicate credible intentions about the range in which overnight
and other short-term interest rates should trade in the future,
central banks can transmit their interest rate policies throughout
the money market term structure and beyond.
To this end, most of these central banks limit themselves
to infrequent adjustments in their operating objectives. Targeted
interest rates are generally changed in small steps and only after

3. The notable exception is the Swiss National Bank, which has
maintained bank reserve operating targets for most of this period.
In addition, the Federal Reserve experimented briefly with
nonborrowed reserve objectives from 1979 to 1982. The choice of
monetary policy operating targets has been the subject of
considerable debate. William Poole provides the seminal
discussion of these issues ("Optimal Choice of Monetary Policy
Instruments in a Simple Stochastic Macro Model," Quarterly Journal
of Economics, vol. 84 (1970), pp. 197-216. For a recent
discussion of interest rate operating objectives in the United
States, see Marvin Goodfriend, "Interest Rates and the Conduct of
Monetary Policy" and the accompanying comments by William Poole in
Carneaie-Rochester Conference Series on Public Policy, no. 34
(1991), pp. 7-39.
4. In Japan and the United Kingdom, nonbank financial
intermediaries participate in the interbank market. In Canada, an
important overnight market in call loans, used by both banks and
investment dealers, exists alongside the interbank market.
-4-

Kasman

a sufficient amount of new information has accumulated to warrant
a change in policy. By encouraging expectations of interest rate
stability over a medium-term horizon, policy makers' gain influence
over rates throughout the term structure.
Although interest rate operating objectives have been
prevalent among these central banks over the past two decades, the
type of implementation strategy employed has, in many countries,
evolved considerably.5 Puring the 1970s, the central bank of
Japan and several European central banks relied heavily on a
system of administered interest rates to implement policy. Banks'
marginal reserve demand in these countries was largely met through
central bank credit facilities, often at below-market rates.6
"Official" or tightly controlled money market rates served as
anchors for regulated deposit and lending rates. Together with
other controls over financial activity, official rate changes were
transmitted largely through their direct effect on bank credit
availability.
This approach came under pressure in the late 1970s. The
delays by some central banks in adjusting interest rates to
counter a buildup of inflation in the late 1970s raised concerns
about the inflexibility of interest rate determination. Many
observers believed that the use of highly visible official rates
constrained banks from adjusting policy in a timely fashion. More
important, however, rising inflation helped spur the
liberalization of financial markets, which in turn substantially
increased the importance of competitive forces in determining
interest rates. Domestic financial markets also became more
closely integrated with foreign markets. As a consequence,
market-determined interest rates and exchange rates played an
increasingly central role in private agents' expenditure
decisions.7

5. An excellent discussion of how monetary policy procedures
have evolved can be found in J.T. Kneeshaw and P. Van den Bergh,
"Changes in Central Bank Money Market Operating Procedures in the
1980s", BIS Economic Papers, no. 23, January 1989.
6. Reliance on subsidized central bank credit sources for bank
reserve needs characterized German, Japanese, and Swiss monetary
policy.




7. A detailed analysis of financial innovation and its effect
on the monetary policy transmission mechanism can be found in
Financial Innovation and Monetary Policy, Bank for International
Settlements, (Basle, Switzerland 1984).
-5-




Kasman

Although procedural changes have been greatest in those
countries where financial change has been most significant, the
central banks under review have in general moved towards marketoriented methods for implementing monetary policy. As noted
earlier, authorities increasingly rely on market-determined
interest rates both as operating objectives and as key elements in
the transmission mechanism.
At the same time, market operations,
in which central banks intervene in financial markets at freely
determined prices, have gradually replaced lending and regulatory
controls as the principal instrument for altering reserve supplies
in most countries.
The shift toward market-oriented interest rate objectives
has helped the central banks to reduce the repercussions arising
from changes in their policy stance. In addition, open market
operations permit central banks to exercise considerable
discretion in the day-to-day management of reserves. While
relying on market forces to determine interest rates, central
banks can intervene at select times to influence the range within
which rates move. Furthermore, the wide variety of available
domestic money market instruments (whose development was greatly
encouraged by monetary authorities in most countries) allows the
banks to construct intervention strategies that span the money
market term structure.
In practice, central banks continue to severely limit the
range in which short-term interest rates fluctuate. By finetuning their market operations, usually on a daily basis, these
central banks alter reserves to accommodate variations in reserve
demand.
This active effort to moderate even transitory interest
rate fluctuations underscores central banks' desire to communicate
their policy intentions clearly to market participants. In nearly
all the countries under review, the stance of monetary policy is
signaled through interest rates. Market interest rates respond to
developments other than policy changes, however, and movements
unrelated to policy must be filtered out before policy inferences
can be drawn. By sharply limiting interest rate variations daily,
central banks ensure that market participants can clearly identify
interest rate targets and quickly ascertain changes in the
monetary policy stance.
To implement an interest-rate-based operating policy
through periodic open market operations, central banks must be
able to predict the demand for bank reserves over some relevant

-6-




Kasman

horizon. Banks need reserves to meet reserve requirements and to
make interbank payments. Central banks have considerable
influence over reserve demand through their role in setting
reserve requirements and interbank clearing rules. Specific rules
(lagged reserve accounting, reserve averaging, and carryover
provisions) and payment systems practices (timing of payments,
overdraft provisions) have been designed, in part, to strengthen
and stabilize the short-term demand for bank reserves. In
general, the stability of reserve demand over a maintenance period
has been a central element underlying central bank implementation
procedures*
In the past, many central banks actively managed reserve
demand by changing reserve requirements and applying other
administrative controls to bank behavior. These practices have
greatly diminished in recent years reflecting, in part, the
general trend towards market-based policy strategies. At the same
time, all six central banks have reduced reserve requirement
ratios over the past decade in an attempt to lighten the burden
they place on banks. In some countries the relaxation of
restrictions on banks' reserve holdings has led to greater
variability in reserve demand, compelling authorities to adjust
their reserve management procedures.
Although this overview of the key features of central bank
implementation strategies suggests broad similarities across
countries, the specific techniques employed by individual central
banks to implement monetary policy vary greatly. Central bank
market operations span a wide spectrum of assets and maturities;
the timing of operations and the frequency with which they are
conducted also differ. Significant differences can be seen as
well in the conditions determining access to central bank credit,
the regulations setting required reserve levels, and the length of
time granted depository institutions to meet their obligations.
In many cases, these differences are institutional in
nature, reflecting the particular environments in which central
banks operate. For example, in conducting open market operations,
central banks must depend on the markets available to them. Where
active secondary security markets are not developed, central banks
may need to make special arrangements for implementing their
reserve management policies.
The remainder of this section compares monetary policy
implementation techniques across the six countries. By examining
the particular institutional environment in which each central

-7-




Kasman

bank operates and by observing the interaction of the specific
instruments central banks employ — open market operations,
central bank lending policy, and reserve requirements -- one can
identify meaningful differences between Federal Reserve and
foreign central bank operating procedures.
Operating Objectives and Procedures
All six central banks gear their short-term reserve management
activities toward influencing interest rates, but specific
interest rate strategies differ from bank to bank. The Federal
Reserve in the United States limits its activities to influencing
overnight interbank rates (the federal funds rate), allowing
market forces to determine the transmission of policy to other
financial markets. The Swiss National Bank also acts to smooth
daily fluctuations in overnight interbank rates, but it is unique
among these central banks in setting no explicit interest rate
operating objective. Although the four other central banks also
actively intervene to smooth fluctuations in overnight rates, they
generally seek to influence money market rates of longer
maturities as well. In Japan, overnight interbank rates remain
the primary operating objective of the central bank, while in
Canada, Germany, and the United Kingdom, rates of longer maturity,
up to three months in some cases, are employed as the primary
operating objective. A summary of the interest rates important to
the banks' policy implementation is presented in Table 1. The
primary interest rate operating objective for each country is
highlighted.
Of the central banks considered, the Bank of England (BOE)
is probably most active in its daily reserve management
activities. Operating in an environment in which reserve
requirements are low and banks each day try to maintain a specific
daily level of operational balances at the BOE, the Bank has
developed a strategy of frequent intraday interventions in money
markets to achieve its interest rate objectives.8
Each morning at 9:45 a.m. the BOE announces its estimate
of the net reserve position of the banking system for the day.
Based largely on expected government transactions and the BOE's

8. To assist the BOE in its daily forecast of the reserve
position of the banking system, each clearing bank is obliged to
specify the size of reserve balances that it will try to maintain
daily.
-8-




Kasman

maturing stock of short-term bills, these estimates signal the
amount of reserves that the BOE anticipates must be supplied to
bring actual balances of clearing banks to the levels the banks
are expected to maintain.9
Because the bulk of the BOE's assets are in short-term
bills (commercial or Treasury) that mature in less than three
months and that do not roll over automatically, the banking system
will usually be projected to have a "cash shortage" at current
interest rates. To meet this shortage, discount houses, which
serve as intermediaries between the BOE and private banks, are
invited to offer bills to the Bank for purchase, indicating the
price at which they are willing to sell.10 The BOE buys bills to
meet the estimated shortage in four maturity bands: zero to
fourteen days, fifteen to thirty-three days, thirty-four to sixtythree days and sixty-four to ninety-one days. It chooses the best
prices offered but holds unchanged the minimum dealing rate (stop
rate) on Band 1 bills maturing in up to fourteen days. As many as
three rounds of these operations may take place in a day, enabling
the BOE to respond to changing intraday market conditions. If
late-day imbalances arise, they are met through credit facilities
available to discount houses.
By purchasing bills across bands (maturities), the BOE
attempts to extend its influence over interest rates throughout
the money market. Variations in the amount of bills purchased in
Band 4 (sixty-nine to ninety-one days), for example, tend to have
a strong influence on three-month Treasury bill rates. The BOE
also has the option of offering repurchase agreements to discount
houses on its own terms if it does not wish to validate the rates
being offered. Mindful of this option, the discount houses will
generally offer prices embodying their expectation of the BOE's
desired rate objectives.
The stop rate changes infrequently. Movements in this rate
signal a shift in BOE policy and are usually reflected immediately

9. The government holds most of its balances with the BOE.
Because its daily transactions with the rest of the economy are
large and fluctuate widely, the BOE's forecast of net government
flows is both the key component of this estimate and the greatest
source of uncertainty.
10. For more detailed information on the role of discount
houses in the U.K. financial system and the BOE's money market
operations more generally, see "Bank of England Operations in the
Sterling Money Market", Bank of England Quarterly, October 1988.
-9-




Kasman

throughout the interbank market and in commercial bank base
lending rates (Chart 1 ) . On occasion, the BOE will send a strong
signal of its intention to shift policy by choosing not to
accommodate a shortage in reserve needs during the day, thereby
obliging discount houses to borrow from the BOE at terms
determined by the Bank. Since the BOE has the flexibility to set
this lending rate either above or below current stop rates, it can
use this procedure to signal a tightening or an easing in policy.
Japanese monetary authorities followed a similar strategy
of tight control over the key intervention rate until the early
1980s. Combining reserve management operations with
administrative control over interbank market participants, the
Bank of Japan (BOJ) was able to stabilize the call-money overnight
interbank interest rate at the level desired for long periods. As
part of a broader reform of financial markets over the past
decade, the BOJ has actively promoted integration of the interbank
with other financial markets and encouraged greater flexibility of
interbank interest rates, particularly on an intraday basis.11
The overnight call rate remains the BOJ's key operating
objective, and although it is subject to greater influence from
market forces than in the past, the BOJ still actively strives to
limit its fluctuations around the targeted level (Chart 2 ) . The
BOJ implements this policy through a variety of market operations,
primarily transactions in commercial bills, and through its daily
management of discount window credit. Control over the "reserve
progress ratio," which measures reserves accumulated by banks
relative to those required within a maintenance period, is a key
element of this policy. Upward pressure on interest rates is
effected by supplying fewer reserves than are necessary for the
reserve progress ratio to rise at an average pace.
Banks have considerable leeway in managing their reserve
positions because the reserve maintenance period is a full month
in Japan. Nevertheless, changes in the reserve progress ratio
clearly convey the BOJ intentions concerning future interest rates
and, as a result, usually lead to a quick response in interbank
interest rates.

11. For a detailed analysis of the evolution of Bank of Japan
policy and references to the literature on financial market
liberalization in Japan, see Bruce Kasman and Anthony P. Rodrigues
"Financial Liberalization and Monetary Control in Japan" this
Quarterly Review (Autumn 1991).
-10-




Kasman

The evolution of BOJ policy over the past decade reflects
a movement towards procedures long practiced by the Federal
Reserve System. Indeed, the two central bank implementation
strategies appear quite similar in their basic characteristics —
an overnight interbank market operating objective, the use of
market operations and discretionary central bank lending
facilities as policy instruments, and a focus on reserve
management over a maintenance period.
Still, important differences remain between the operating
strategies of the Bank of Japan and the Federal Reserve. While
the Federal Reserve conducts most of its daily operations in the
repurchase market for government securities, the BOJ relies on a
variety of private market instruments, including commercial bills,
commercial paper, and certificates of deposit. In part, the BOJ's
reserve management activities reflect the limited development of a
single short-term government securities market in Japan. However,
the BOJ has also employed operations in different instruments to
exert direct influence on money market interest rates. Up until
1988 interbank and other open markets were not fully integrated,
and the BOJ intervened actively in longer term money markets,
primarily to influence the three-month certificate of deposit
rate.
Following a period in 1987 and 1988 in which open market
rates moved well above comparable rates in the interbank market,
the BOJ implemented a series of reforms to facilitate arbitrage
across short-term money markets.12 Since that time the BOJ has
generally limited its efforts to influence direct influence over
interest rates in the interbank market to instruments of seven
days' maturity or less. Market operations' in longer term money
market instruments are now primarily designed to offset seasonal
fluctuations in reserve demand.
The administration of discount window lending also differs
considerably in the two countries. In the United States, banks
initiate the decision to borrow at the Federal Reserve's discount
window, and borrowing is rationed through a set of administrative
guidelines. In Japan, the BOJ decides on the level of bank
borrowing and the length of loans (a factor that determines the

12. For a detailed discussion of money market reforms
implemented since 1988, see Japan's Short-Term Money Market and
Issues, Ministry of Finance and Bank of Japan, Money Market Study
Group, August 1991.
-11-




Kasman

effective cost of a loan). In administering the discount window
lending, the BOJ actively manages loan provision on a daily basis
to respond to intraday fluctuations in reserve positions. The BOJ
is unique among the central banks surveyed in employing lending as
a discretionary instrument of daily reserve management.
The institutional environment in which the Swiss National
Bank (SNB) operates has undergone considerable change in recent
years. From 1980 through 1988 the SNB guided its policy largely
with short-term bank reserve targets. Although interbank interest
rates fluctuated widely on a daily basis, the SNB was reasonably
successful in achieving its primary policy objective of
maintaining low rates of inflation.13
In 1988, the combined effects of implementing an electronic
payment system for settling interbank cash balances (1987) and
introducing new liquidity rules (January 1988) led to a sharp
decline in reserve deposits held at the Bank (Chart 3). 1 4 The
difficulties faced by the SNB in predicting the size of this
decline led to an inopportune expansionary monetary policy in
early 1988. In response, the SNB shifted its operating objectives
away from reserves toward short-term interest rates and exchange
rates.15 Although the SNB has gradually moved back towards an
implementation strategy based on operational targets for bank
reserves, it has continued to emphasi'ze interest rates in its
daily operating procedures.
Each quarter the SNB signals its short-term policy
intentions by announcing a forecast of the level of the monetary

13. See Ben Bernanke and Frederic Mishkin ("Central Bank
Behavior and the Strategy of Monetary Policy: Observations from
Six Industrial Countries," unpublished paper) for a recent
assessment of Swiss monetary policy in relation to other central
bank practices over the past two decades.
14. The new liquidity rules lowered required reserves and
shifted the maintenance period from the end of the month to a
month average.
15. See Organization for Economic Cooperation and Development,
OECD Economic Survey-Switzerland (Paris, 1989) for a discussion of
Swiss monetary policy following these institutional changes.
-12-




Kasman

base in the subsequent quarter.16 Incorporated in this forecast
is an unannounced operational target for the level of bank
reserves held at the SNB. Although this target serves as a guide
to policy operations over each month and each quarter, authorities
have considerable discretion in deciding on their day-to-day
activities. In implementing daily policy, the Bank largely seeks
to smooth fluctuations in overnight interbank rates. Nonetheless,
the interest rate policy of the SNB differs significantly from
that of the other central banks under review. No operational
targets are set for the level of interest rates, and the SNB does
not employ.interest rates to signal its stance to market
participants.
The institutional changes that took place in Switzerland in
the late 1980s have not led to substantial changes in the
implementation procedures employed by the SNB. As before, market
operations are generally conducted once each morning through
foreign currency operations. These transactions, in the form of
U.S. dollar-Swiss franc swaps, are conducted at rates close to
those prevailing in Euromarkets and extend up to one year in
maturity.
Earlier SNB restrictions, which placed limits on end-ofmonth Lombard lending and required banks to give advance
notification of their credit needs, were removed when reserve
requirements were reduced in 1988.17 Nevertheless, in 1989 the
Bank floated the Lombard rate 200 basis points above market rates,
a move that has substantially limited recourse to this facility.
In Germany, interest rates on security repurchase
agreements of one- to two-month maturities are the primary

16. The forecasts are designed to be consistent with mediumrun growth targets for the monetary base. Since 1990, these
medium-run targets have been defined as annual growth rates to be
achieved over a period of three to five years. The targets thus
give the SNB considerable flexibility in determining its quarterly
forecasts.
17. Before January 1988, banks' reserve requirements were
monitored only on the last day of a month. Banks' demand for
reserves consequently soared at this time. With access to Lombard
lending limited by these restrictions, short-term interest rates
often rose very steeply at month's end.
-13-




Kasman

operating objective of the Bundesbank.18 These rates are
determined at periodic tenders typically conducted once a week.
The Bundesbank normally determines the amount of repurchase
agreements offered at a tender by assessing market demand for
reserves, and it chooses the best prices available. On occasion,
it will fix the price (interest rate) at a tender to send a clear
signal of its policy intentions to markets.19
Of the central banks considered, the Bundesbank is
probably the least active in its daily reserve management
activities. Repurchase agreement tenders generally provide the
liquidity needed each day. Occasional "supportive" operations are
undertaken to influence the day-to-day money rate through a number
of reversible fine-tuning measures. Short-term interest rate
smoothing, however, is largely obtained through means other than
market operations, a system that reflects the limited development
of domestic money markets in Germany. Specifically, official rate
facilities on Lombard loans and the Bundesbank's Treasury bill
selling rate bound the range within which money market rates can
fluctuate (Chart 4 ) . In addition, high reserve requirement ratios
and long (one-month) maintenance periods provide banks with
considerable flexibility to arbitrage away transitory shocks to
their reserve positions.
For the Bank of Canada (BOC), the three-month Treasury bill
tender rate is the primary operating objective. The BOC
participates in the weekly auction and buys and sells bills in the
market from time to time, both on an outright and on a buy-back
basis. But the BOC implements policy mainly through daily
transfers of government demand deposits between the BOC and
private banks.20 These transfers are decided late in the day, by
which time the BOC has information on government transactions and
other payment items that might affect bank reserves. Thus, the

18. For a recent discussion of Bundesbank operating
procedures, see Andre Bartholomae, "Some Operational and
Instrumental Aspects of Monetary Targeting in Germany," Deutsche
Bundesbank, unpublished paper, 1991.
19. For example, the Bundesbank employed "volume tenders" in
which it set interest rates for several months following the
October 1987 stock market crash.
20. A detailed description of these operations is found in
Kevin Clinton, "Bank of Canada Cash Management: The Main
Technique for Implementing Monetary Policy", Bank of Canada
Review, January 1991.
-14-

Kasman

BOC is able to determine end-of-day reserve positions with unusual
precision, particularly because these "drawdowns" or "redeposits"
of government balances occur too late for banks to make further
adjustments to their balance sheets. These transfers have a
direct effect on overnight rates in the call and.interbank
markets. Daily reserve management activities are geared, however,
toward maintaining market conditions consistent with the BOC's
weekly Treasury bill rate objective (Chart 5 ) .
Kay Instruments of Rosary* Management
Intervention tools vary widely across the central banks surveyed.
In part, these instruments reflect the differing financial
environments facing authorities in the six countries. The choice
of instruments is, however, also related to specific objectives of
reserve management and the means chosen by the authorities to
signal their policy intentions to financial market participants.
A summary of the market operations employed by the six central
banks is presented in Table 2.
The U.S. Federal Reserve operates mostly in the secondary
market for government securities. The prototypical open market
operation, the outright purchase or sale of government securities
in the secondary market, has long been the major instrument for
providing permanent bank reserves in the United States. The
breadth and depth of this market allow the Federal Reserve to add
or drain large amounts of reserves without significantly
distorting yield structures.
Although outright purchases of securities provide the
primary source of secular reserve creation, the Federal Reserve
typically conducts less than ten outright purchases and sales in
the market each year.21 On a daily basis, policy is implemented
primarily through repurchase agreements (which add reserves) or
matched sale-purchase agreements (which drain reserves). These
reversed security transactions involve lower transactions costs
than outright transactions and provide a much more flexible
instrument for the temporary adjustment of reserve positions.
They are conducted through a large existing private market and may
range up to fifteen days in maturity, although they usually mature
in one or a few days. Although most of these transactions are

21. The Federal Reserve does take advantage of purchase or
sale orders of foreign official accounts when these are consistent
with reserve objectives.




-15-




Kasman

designed to smooth temporary fluctuations in reserve markets, they
are also employed by the Federal Reserve to implement a change in
its policy stance.
In Japan, Canada, and the United Kingdom, as in the United
States, outright purchases of securities are the main asset
counterpart to the expansion in the monetary base over time. In
Japan, the purchase of ten-year government bonds meets the secular
demand for reserves but is not important in short-term reserve
management. The BOJ conducts a variety of other operations to
affect reserve positions on a temporary basis. Outright and
reversed .transactions in commercial bills and other money market
instruments are designed to offset seasonal and other short-term
fluctuations in reserve demand. The discount window lending
activities remain the primary tool to smooth unexpected day-to-day
fluctuations in reserve positions.
Canadian monetary authorities also employ a variety of
instruments to achieve policy objectives. The BOC's weekly
participation in the three-month Treasury bill tender and its
purchases of long-term government bonds at issue are the principal
asset counterparts of money base increases in Canada. On a dayto-day basis, the BOC's drawdown/redeposit mechanism, described
earlier, is its primary instrument of reserve management. The
distribution of drawdowns and redeposits among clearing banks is
determined at twice-monthly auctions where banks bid competitively
for allocation ratios of government demand deposits.
Supplementing this mechanism are other market operations,
including outright purchases of short-term government securities
and repurchase agreements. All open market operations are,
however, routinely neutralized by the BOC as part of its
drawdown/redeposit activities. As a result, open market
operations are geared toward directly influencing particular money
market interest rates.
In the United Kingdom, BOE assets are held primarily in the
form of short-term eligible bills. The BOE routinely purchases
bills to roll over its maturing portfolio and to achieve its
short-term reserve management objectives.22

22. Eligible bills include Treasury bills and commercial bills
carrying two established names, usually those of a British bank
and a discount house. The BOE will buy or sell bills of up to
three months' maturity and does conduct some reversed security
transactions.
-16-




Kasman

As noted earlier, BOE operations are designed to relieve
daily money market shortages through the outright purchase of
bills from discount houses. Although it typically maintains a
fixed stop rate on Band 1 bills, the BOE generally does not
relieve the entire shortage through Band 1 bill purchases. It
conducts bill operations in maturities as long as three months,
designing these operations to exert influence on rates throughout
the money market term structure. In addition, the BOE can refuse
to relieve shortages through bill purchases if it is unhappy with
the rates being offered. In these circumstances, the BOE can
offer repurchase agreements on its own terms or invite discount
houses to use their borrowing facilities at 2:30 p.m. at a rate
set at the BOE's discretion.23
Neither the Bundesbank nor the SNB holds significant
portfolios of securities because well-developed short-term money
markets do not exist outside the interbank market in Germany and
Switzerland. In this environment, the Bundesbank uses central
bank lending (mainly bills rediscounted) and bond repurchase
operations as the major vehicles to augment the monetary base.
The Bundesbank has established special provisions for reversed
security transactions with banks; these transactions serve as the
Bank's primary instrument of short-term reserve management. The
Bundesbank conducts periodic tenders (usually weekly) for one- to
two-month repurchase agreements. These repurchase agreements
consist of a secular component and a component that makes
temporary adjustments to reserve positions. Repurchase agreements
have steadily increased as a share of Bundesbank assets since the
mid-1980s, gradually supplanting discount window lending as the
principal asset counterpart of the money base. Other instruments,
such as foreign exchange swaps and the transfer of government
deposits from the Bundesbank to banks, are employed when daily
adjustments in reserve positions are deemed necessary.24

23. The 2:30 borrowing differs from normal day-to-day late
assistance in that the interest rates on loans are published and
the amounts borrowed do not count against discount houses'
borrowing facilities.
24. Foreign Exchange swaps are usually employed to neutralize
an expansion in reserves resulting from international capital
inflows. Transfers of government deposits between the Bundesbank
and private banks are generally used to offset temporary reserve
shortages associated with tax payments.
-17-




Kasman

In Switzerland, the domestic securities market is extremely
narrow. An active interbank swap market for major foreign
currencies does exist, however, and the SNB employ's currency swaps
as the primary instrument of both permanent and temporary reserve
operations. Conducted daily in the form of U.S. dollar-Swiss
franc swaps with a small number of banks, these operations
currently provide over 90 percent of the reserve creation for
Swiss banks. Since the dollars purchased by the SNB in these
transactions are covered forward, these transactions can be viewed
equivalent to temporary operations in domestic securities.
Because swaps are settled with a two-day lag, the SNB supplements
these activities with same-day shifts of government deposits
between its books and those of private banks.
Central Bank Credit Facilities
The monetary authorities in all six countries considered offer
banks a facility for obtaining credit. The market operations
described above, however, have largely replaced central bank
credit as the major tool for short-term reserve management in
these countries. At present, most central bank lending facilities
are designed to meet unforeseen and temporary end-of-day liquidity
shortages or to provide assistance for institutions in times of
stress. Nonetheless, the role of lending in the six central
banks' implementation strategies varies. A summary of key
characteristics of central bank lending facilities is presented in
Table 3.
In four of the countries considered (Germany, Japan, the
United States, and Switzerland), a collateralized credit facility
is made available to banks at below-market interest rates. In
Germany, Japan, and Switzerland, discount window lending,
determined by quotas, provides an ongoing source of subsidized
funds to meet a portion of secular reserve demand. The
Bundesbank's facility is particularly large, currently accounting
for about one-quarter of total central bank assets (Table 4 ) . The
large volume of subsidized discount window lending in Germany is
designed, in part, to offset the costs to banks of high levels of
required reserves.
Because German and Swiss banks fully use their quotas most
of the time, discount window lending does not accommodate banks'

-18-




Kasman

unanticipated reserve needs in these countries. Both the
Bundesbank and the SNB provide an additional line of credit at a
penal rate to meet unexpected short-term liquidity needs.25
These facilities, called Lombard loans, effectively cap interest
rate increases for short periods. Swiss Lombard rates float daily
at two percentage points above the average of the previous two
days' interbank call money rates. German Lombard rates, in
contrast, are fixed by the Bundesbank and in recent years have
generally remained no more than 100 basis points above the
repurchase agreement rate.
Lombard lending by the Bundesbank has soared for brief
periods on several occasions in recent years. These surges in
lending reflect, in addition to market-related liquidity
developments, a strategy of tightening policy: money market rates
are increased first; once market pressures build, these increases
are validated in official rates.26
In the other countries reviewed, the central bank has
greater freedom to decide the terms on which lending is made
available. In the United States, the Federal Reserve generally
sets the discount rate below short-term market rates and rations
access through administrative guidelines. Lending is designed to
provide for unexpected liquidity needs, particularly at the end of
reserve maintenance periods. For institutions that use the window
frequently, however, future access is reduced, raising the
implicit cost of borrowing. Furthermore, worries about potential
adverse market reactions to discount window borrowing have
developed in recent years as bank failures and earnings stress
have risen. The use of the discount window has, consequently,
been relatively limited.
Of the countries under review, only Japan makes lending an
important instrument in short-term reserve management. Discount
window lending makes up a substantial share of BOJ assets

25. Both central banks impose quotas on access to Lombard
facilities, but the quotas rarely present an effective constraint
on borrowing.
26. The maturity of Lombard loans is determined by the
remaining maturity of securities rediscounted. Generally the
Bundesbank grants such loans with the expectation that borrowing
should be repaid the following day. Nonetheless, there exists
some incentive to borrow heavily through Lombard loans when
repurchase interest rates are expected to increase above Lombard
rates at the subsequent weekly repo tender.
-19-




Kasman

(currently over 10 percent), and the Bank actively manages its
lending policies on a daily basis. The BOJ can either increase or
call discount window loans at its discretion, and typically uses
this instrument to smooth daily fluctuations in bank reserve
positions. In addition, with its "plus-one-day" pricing of loans,
the BOJ's effective lending rate exceeds the discount rate and can
become penal for very short-term loans.27 Discount window
lending thus gives the BOJ a highly flexible instrument for
influencing daily conditions in interbank markets.
England's central bank also has discretion in providing
credit. In its transactions with discount houses the BOE can
decide whether to provide credit and what the price of that credit
will be. Funds are made available for "late assistance" to meet
interbank clearing needs, but the terms of this borrowing are
determined by the BOE and are not disclosed publicly. Generally
funds are lent at or above market rates, in a way that permits the
discount house to predict the cost accurately. As noted earlier,
the BOE occasionally uses its lending policies to signal changes
in its policy stance, allowing discount houses to borrow at a
publicly announced rate after it has refrained from accommodating
reserve demand earlier in the day.
The central bank lending rate of the BOC (the Bank Rate) is
adjusted weekly and set 1/4 percentage point above the previous
Thursday's three-month Treasury bill tender. Until recently,
banks were guaranteed recourse to this facility only once during a
reserve maintenance period. The cost and availability of further
borrowing were subject to the discretion of the BOC. Funds were
provided, but at a rising cost for repeated use.
These restrictions on access to BOC credit- were removed in
November 1991. Banks can now borrow freely at the Bank Rate
either as overnight overdrafts or to meet reserve deficiencies, a

27. The interest charged on discount window
calculated on the period of the loan (using the
rate) plus one day. Thus, the effective rate of
the BOJ reduces the length of time for which it
lend.
-20-

loans is
official discount
interest rises as
is willing to

Kasman

change seen as a necessary prelude to the phased elimination of
reserve requirements that began in June 1992.28
In addition to providing credit to meet short-term
liquidity needs, most countries also offer a facility to absorb
excess reserves so that short-term downward pressures on interest
rates will be limited. In Japan, the BOJ has the option of
withdrawing outstanding loans at will during banking hours. The
Bundesbank's Treasury bill selling rate functions as an effective
floor on call money rates in Germany, and in Canada, matched or
outright sales of Treasury bills serve a similar purpose. In the
United Kingdom, discount houses can offer to purchase securities
from the BOE in the afternoon if surpluses emerge.




Reserve Requirements
Like central bank lending, required reserve ratios have diminished
sharply in recent years. Required reserve ratios in all these
countries stand well below their levels of the early 1980s; in
some countries, requirements no longer effectively constrain bank
behavior. In addition, the once common practice of altering
reserve requirements to adjust the monetary policy stance has
largely been discontinued.
Nonetheless, most central banks still view reserve
requirements as an important part of their implementation
procedures. Requirements are seen as strengthening and
stabilizing the short-run demand for reserves, thus enhancing
central bank control over interest rates.. A summary of important
characteristics of reserve requirement regulations is presented in
Table 5.
Required reserves in all six countries under review are
determined by ratios linked to categories of bank liabilities.29
In the United States and, until recently, in Canada, requirements
have primarily been imposed on transactions deposits, a practice

28. Under the regulations in place since June 1992 a bank with
a cumulative deficiency at the end of a reserve maintenance period
may pay a fee, charged at the Bank rate in lieu of taking an endof-period advance. In practice, banks have adopted the fee option
so that end-of-period advances no longer appear on the BOC balance
sheet.
29. In June 1992, Canada removed required reserve ratios as
part of its phased elimination of reserve requirements.
-21-




Kasman

that reflects earlier attempts to use reserve requirements to
facilitate the targeting of Ml through operating objectives for
bank reserves. Elsewhere, requirements are more broadly based.
In the United Kingdom, Japan, and Switzerland, requirements are
roughly similar across types of eligible liabilities.
In all these countries, the period in which liabilities are
incurred (the accounting period) ends before the period in which
required reserves are held (the maintenance period). These lagged
or semilagged accounting mechanisms are operationally convenient
and, where reserve requirements are binding, provide central banks
with a relatively good estimate of reserve demand within a
maintenance period. For all six central banks except the BOE,
reserve projections at maintenance period horizons are a key
element in determining policy operations.30
Although lagged reserve requirements predetermine the
demand for reserves, they can also severely limit the interest
sensitivity of reserve demand, particularly at the end of
maintenance periods. Unforeseen shifts in either the demand for
or the supply of reserves have often led to large fluctuations in
interbank rates at the end of a maintenance period. To provide
greater flexibility in reserve management, particularly in the
early stages of a maintenance period, nearly all of these central
banks allow required reserves to be met by average reserve
holdings over a maintenance period.31 Reserve averaging gives
value to banks' excess reserve positions by enabling the banks to
maintain offsetting deficiencies during other days within the
period. As a result, banks have an incentive to arbitrage away
the interest rate effects of temporary reserve shocks. Through
this mechanism, required deposits at the central bank can function

30. As noted earlier, clearing banks in the United Kingdom
provide the BOE with an estimate of the operational balances they
wish to hold each day. The BOE uses these estimates as a guide in
determining daily security operations.
31. Reserve averaging extends over one month in Germany,
Japan, and Switzerland, and over two weeks in the United States.
In Canada, reserve averaging extended over two half-month periods
until June 1992, when it was extended to one month.
-22-




Kasman

as an important aid to central banks in promoting interest rate
stability.32
The extent to which bank reserves actually serve as a
buffer stock is related to the level of reserve balances held at
the central bank. Because overnight overdrafts are restricted in
Switzerland, Japan, and Germany, and penalized in the United
States, Canada, and the United Kingdom, the cost of running
reserve deficiencies rises substantially when average reserve
balances are low. In the United States and Canada particularly,
concerns have arisen about the banking system's reduced ability to
absorb reserve imbalances at low reserve levels. Reserve deposits
held at the central banks of both countries have fallen sharply in
recent years as a result of a secular increase in demand for vault
cash to satisfy reserve requirements and, in the United States, a
reduction in reserve requirements (Table 6). 3 3
Reserve management strategy in the United States
traditionally focused on the two-week average reserve levels held
by banks over a maintenance period. Since the cut in reserve
requirements in December 1990, however, the open market desk
encountered increasing conflicts between this strategy and daily
federal funds market conditions. Many banks have become less
tolerant of excess reserve positions early in the maintenance
period, a reaction that has often led to significant late-day
downward pressure in federal funds rates. At the same time, the
funds rate in the morning can be a misleading guide to reserve
market conditions as banks sometimes hold on to reserves early in
the day to guard against inadvertent overdrafts. When faced with
these conflicts in conducting its operations, the Desk has chosen
to pay greater attention to daily trading.conditions in the
federal funds markets to prevent misleading signals from being
sent to markets.34

32. A provision for the carryover of a portion of reserve
surpluses (or shortages) allows for some additional flexibility in
managing reserves across maintenance periods in the United States.
33. In both countries, holdings of vault cash over previous
maintenance periods satisfy current reserve requirements.
Increased demand for yault cash thus lowers required deposits even
when reserve requirements are unchanged.
34. See "Monetary Policy and Open Market Operations1* for
further details.
-23-




Kasman

In two countries, the United Kingdom and Switzerland,
reserve requirements place no effective constraint on bank
behavior. In the United Kingdom, banks must place small nonliquid deposits at the Bank of England for six months at a time.
This requirement provides the BOE with operating income but is not
intended to play a role in the BOE's monetary policy operating
strategy.
Since effective requirements are lacking, demand for
reserves (operational deposits) is determined entirely by daily
clearing needs. In this environment, the BOE has developed an
operating strategy involving a number of daily market operations
to interest fluctuations and other intraday developments. In
addition, banks' uncertainty over their end-of-day clearing needs
is eased by the availability of BOE late-day lending facilities to
discount houses. BOE policies serve to stabilize reserve demand
and encourage banks to economize on reserve holdings (Table 6 ) .
Since the decline in reserve requirements in Switzerland in
1988, the SNB has placed greater emphasis on smoothing daily
fluctuations in interest rates in its daily activities. In
addition, central bank lending facilities in the form of Lombard
loans are available to banks without restriction to meet
unexpected liquidity shortfalls. Nonetheless, the SNB is much
less accommodative than other central banks in its approach to
offsetting temporary reserve disturbances, prohibiting overnight
overdrafts and setting a large spread (200 basis points) between
market and Lombard lending rates. In this environment, Swiss
banks have chosen to hold substantial reserve deposits in excess
of those required by regulations.
RELEVANCE POR FEDERAL RESERVE OPERATING PROCEDURES
The varied institutional and political environments facing these
central banks make it difficult to assess whether practices
followed in any one country would be useful to another.
Nonetheless, the comparison of operating procedures presented
above does provide interesting insights, some of which may be
relevant to U.S. policy makers.
The similarities in operating strategy among these central
banks dominate any existing differences. All six banks currently
gear their daily policies toward influencing money market interest
rates; all except the SNB use short-term interest rates as
operating objectives to guide their reserve management activities.

-24-




Kasman

Furthermore, none of the banks aims to control interest
rates rigidly. Although the tolerance for interest rate
divergences from objectives differs across banks, authorities
generally allow market forces to determine interest rates and
intervene only to limit short-term fluctuations or to alter rates
when changing economic conditions warrant.
Since interest rate operating objectives are transmitted to
economic activity largely through their linkage to longer term
interest rates and other financial prices, central bank
intervention strategies are designed to communicate information
about current and future policy that strengthens this
transmission. In most cases, interest rate objectives are changed
in small steps to stabilize expectations across the term
structure. In some countries, central banks intervene in assets
of varying maturities to influence the money market term structure
directly.
In addition, these central banks actively seek to limit the
daily volatility of targeted interest rates in order to reduce
uncertainty about the stance of policy. In some countries
(Germany, the United Kingdom) intervention rates under the tight
control of the central bank send a precise signal of central bank
intentions. Elsewhere, although some interpretation of money
market interest rate movements is necessary, the central banks
stabilize their targeted rates sufficiently so that the basic
thrust of their policies is clear.
Over the past decade, foreign central banks have increased
the role of open market operations as a reserve management
instrument, moving toward an approach long followed by the Federal
Reserve in the United States. At present, each of the central
banks reviewed employs some form of open market operation as an
instrument for controlling reserves. Some foreign central banks
conduct their operations through special arrangements with banks
or other counterparties. But where these arrangements exist, they
generally reflect the limited development of secondary security
markets.
More meaningful differences among the six central banks
emerge in the functioning of their credit facilities. To be sure,
the monetary authorities in all six countries extend credit to
banks with temporary clearing imbalances and to banks in financial
stress. The foreign central banks, however, differ from U.S.
practice by moving away from administrative controls on credit
allocation.

-25-




Kasman

In three countries -- Germany, Switzerland, and Canada -banks are able to access an open-ended line of credit for
temporary liquidity needs at their discretion. Borrowing rates
are set above the prevailing market rates and, in Switzerland and
Canada, rates adjust automatically to market rates. In Japan and
the United Kingdom, access to the discount window remains at the
discretion of the central bank. In practice, however, discount
houses in the United Kingdom can count on the central bank to meet
temporary liquidity needs at rates close to the Bank of England's
prevailing intervention rates.
These facilities provide foreign central banks with a
flexible instrument to contain interest rate pressures,
particularly late in a trading day when other intervention
instruments are unavailable. In addition, each of these foreign
central banks offers a facility to absorb late-day reserve
excesses and thereby moderate downward interest rate pressures.
The Federal Reserve's discount mechanism has considerably
less value as a device for smoothing interest rates. U.S.
discount window lending is provided at subsidized rates and in
accordance with administrative discretion. Partly because of this
subsidy, the Fed discourages frequent use of the window. In
recent years, banks have shied away from approaching the window,
fearing that the markets will perceive them to be dependent on
discount window support. The unwillingness of banks to borrow at
the discount window also reduces the ability of banks to shed
excess reserves through their repayment of outstanding credit.
In an environment of high, binding reserve requirements,
the methods employed by central banks to allocate credit might not
significantly affect their.ability to limit interest rate
variability. With sufficient averaging provisions in place, banks
can be expected to arbitrage away the interest rate effects of
transitory shocks to their reserve positions within a maintenance
period. Indeed, recourse to Lombard loans in Germany, the country
that has the highest reserve requirements and longest maintenur,:a
period of the six countries considered, is quite small under
normal market conditions.35

35. The Bundesbank estimates normal Lombard lending levels at
DM 0.5 billion, a level representing less than 0.2 percent of
total central bank assets. As noted earlier, Lombard lending has
risen sharply during short periods in which the Bundesbank allows
repurchase agreement rates to push up against Lombard rates before
it tightens policy.
-26-




Kasman

But in the United States, recent declines in reserve
requirements, coupled with increased demand for vault cash, have
sharply reduced reserve deposits at the Federal Reserve. In an
environment where overnight overdrafts are costly, the ability of
banks to take advantage of reserve averaging has- become more
limited as reserve deposits decline. These developments,
coinciding with the deterioration in the functioning of the
discount window, may have increased the sensitivity of the federal
funds rate to reserve shocks.
The central banks examined here that have faced similar
concerns about the effects of lower reserve requirements have
tended to revise their procedures to allow for a more elastic
late-day reserve supply. The BOE, operating for over a decade in
an environment where banks are effectively free from reserve
requirements, has developed a strategy combining the elastic
provision of central bank credit for late-day reserve imbalances
with frequent open market operations during the trading day. The
SNB has placed greater emphasis on interest rate smoothing in
daily operations since a reduction in reserve requirements in
1988. In addition, while maintaining a large spread between rates
on its Lombard lending and overnight rates, the SNB has increased
access to central bank lending facilities since the decline in
required reserves. In Canada, restrictions on bank access to BOC
credit have also recently been removed as part of the phased
elimination of reserve requirements.
The example of other central banks, then, raises a
question: Should the Federal Reserve consider revising its
operating procedures to adapt to lower reserve requirements? It
could be argued that some revision enabling the Federal Reserve to
supply reserves more elastically outside of the time it conducts
open market operations could help limit the variability of
interest rates from objectives.
To resolve this issue, an assessment of federal funds rate
variability and its effect on monetary policy transmission is
essential. The Appendix sheds some light on the issue by
presenting evidence on actual interest rate variability. The
interday volatility of the federal funds rate does appear to have
risen following the decline in reserve requirements in 1990.
However, U.S. federal funds rate volatility remains low in
comparison with the volatility observed in overnight interbank
rates in other countries. More important, perhaps, the evidence

-27-

Kasman

indicates that increased federal funds rate volatility, within the
range observed, has not diminished the response of three-month
money market rates to changes in interest rate objectives. Thus,
these results do not suggest that the reduction in reserve
requirements has weakened the effectiveness of the Federal
Reserve's policy transmission mechanisms.
CONCLUSION
Our analysis, while far from conclusive, provides insights that
may be useful in assessing monetary policy operating procedures in
the United States. Like the Federal Reserve in the United States,
several foreign central banks have lowered their reserve
requirements in recent years. Their experience indicates that
interest-rate-oriented monetary policies can be carried out in an
environment of low, nonbinding reserve requirements. Central
banks operating in such an environment have been able to achieve
their interest rate objectives using reserve management techniques
quite similar to those employed by the Federal Reserve System in
the United States.
Foreign central banks have, however, seen the need to
develop mechanisms that provide a highly elastic supply of
reserves tb restrict the intraday fluctuation of overnight
interest rates. In most, countries, the authorities have designed
their central bank lending facilities, with rates set at or above
current market interest rates, to achieve this goal.
The empirical evidence presented in this article indicates
that the recent decline in reserve requirements in the United
States, combined with the increased reluctance of banks to
approach the discount window, has been associated with greater
variability in the federal funds rates. Nevertheless, the
evidence suggests that this rise in variability has not diminished
the effectiveness of U.S. monetary policy operating procedures.
Within its current range, the variability of the federal funds
rate remains low and does not appear to have affected the linkage
between federal funds and other money market rates.
APPENDIX:




OVERNIGHT INTEREST RATE VARIABILITY

The review of central bank operating procedures presented in the
text suggests that foreign central banks, in contrast to the
Federal Reserve, employ their reserve management instruments,
particularly lending facilities, in a way that places strict

-28-

H

a 3 rt

»<

ID H - 3
iQ

H
0

*1

0
H>

* < *1 H - H VO P H i
*< ID 0 rt o M o
CA
3 * 1
1 0 3 3 *
^ h ^ l
rt p
i *rj
» H-PUD
0

o

0 H- 3* 3

OS

W3ftftM

i r p i i i i(I>QHa- H
H - a , a D 3*
v-i.
H- rt C O
ID
H CHID
en <n D PH O< C •A Co A*<O 3 ' frp VO W H i
£

p

rr

N

rt)

n
M
p
3

i t rt ID
H
M - ID H.
ID
< w 3 »i]rt
fl) • H - O ID
CA
iQ *1

ID3*MTJCA

P I I D M

P 3r

3* g o a vo
(nrt q o

f t H-

P
TJ
H- O P P
(0 ID H 3

a 3* 3 n

•

3 rt

i
t\>
VO
1

K 3 H vo * - ID

CA ID i Q r t n
a
rt
H
ID
0 ID
P M •-» M p i-h »i
3
• U> H-*»d
P
< TJ rt M
ID 3*
(D <D o
5 0 P ID H i
Ml"
i H - «i (D H
C
1 3
<
2 3
ID r t H- H - ID a
O a 3* ID 3 ?: W
p>
H- £
3M-CA« K : M

h--

<J
ID
3
ID

3
rt

CA
f t ht H - CL
C ( 0 ID H - 3 P>

n 2
o
ID w cID W
a

o a hID

M > 0

ID

-

0
H i CA

<J

pi

ID ID M

0>
TJ
P
3

(D ID ID C

H g artiii
H-TJ
IDIDHP> H H »
0* 3 rt

C
ID
»1
ID

3

H-

rt
a>
a
in
rt
P
rt
ID
W

3*
HO
3*

TJ < 0 rt
• ID ID CT 3*

cr o *i o IQ 3*t to £ >i
rtio
H*< O
3'ftao» aoo»wo>
M
30»OOrtH3rt0»MlD

H- H3
3
rt 3 rt tQ M»< rt 3* ID a P
IDID
ID 3
£ »1
H-3 a
p ID P 3 rt
CO CA f )
Hrt rt rt Hi

3

C 3'K

0 H P
r t pi H

ID

rt
ID




M- ID H- 3 3 X
w 3 rtO XlD*0
H-iQ p 3 • a ID
IQ
*i CA
O
3 !-»*<
H I rt
p vo
a ^ h i D

M V O » a C M O C L
H O
HiQ
3

rt.

3*
ID

HH-C

rt

h- 3 N rt M
O i Q ID 3 * P

>r}*<
0
H

rt
Q

H

H - 50
O ID
*< W
ID
O h

o

ID

C

H - r t »ij n r t
rt 3 p ID 3* ID
ID i Q M Q i 0» W
H.
K ID 3

r t h | ID • •
<
M - 0>
ID (D 3
?:w C D i

X

ID

tl

rt h O ID ID P
H h
a rt

ID

TJ

H-* M
»1IQ 0
M CA HC O < vo *v p ID 3
Di
3 a S 3 3
vo 0 M fX
r t 3 H i H - (D H ?JK)H
$

ID O

rt

rt
3*

2 0 P » '
rt 0 ( n O h
rt
ID c TJ 3 x ID a
*i P tl ID
P

hP r t (D
<
rt ID
rr 3
ID
rt

n

0
C
3

w ID a
H3
tQ

•
2:
0
rt
ID

*
3*
0

£

ID

hh
H

0

3
3

ID

0»
3
H-»
ID

<

ID
H-»
(fl

0

<

ID
M

H»
VO
00
00
1
vo
I- 1

rt
rt

3*

ID
(0
ID

rt
H-

0
3
i-»
ID
(0
(0

P

C

H-

CA

H
ID

3

m

P»
3

a
0
3
0

rt

O

1
3

0>
H

rt

o>

ID

O
ID
(A

3
*n HID
I-*
a
ID
0>

CA

H

M<Q

0»

3-

rt
ID
tfl

3*
0»

<

ID

a

*<

rt

H-

3

rt
ID
H
ID
10
rt

H-

<;

H

ID
»1
<Q
ID

0)

a

<>
P

«
0
3

<I

3*
P
3

ID

rt

0>

rt

3

H-

<
ID

0>

< aID
H
<
* H0*
rt

0

C
3
CL

"»

ID

3P

H>

qui

o 0 3 H TJ O 3 O O M H i O H
OTJOPOfDOOTJvoOTJP
3 (D<!3HH3C(!)U)h(D3
3iQ w
TJ H (HQ H - g (D 3 >1 H
(Dcrv
c P 3 < D *O< Pd r (tHrMt prt - nH»
r t rt(D
o S

ID
h

0»

<Q
ID

»
0)

cr
0
rct
H-»

**>
tr
o>
(A
H(0

rt
ID

H
H-

0»

cr

HH
Hrt

*<
H(fl
h
ID
H-»

3

rt
0)

a

0>
H-

!-•
><

*
0>

o

0

3

50

HW

<J
ID

•

0
3

rt

O

z
Hrt

30

rt

3*

ID
M
O

0
C
3

<;

H-

rt
ID
(A
rt

O
O

H>
Hi
CA
ID
rt
H

0»

ID
CA

*a
Hrt
ID
rt

3ID

3ID

rt

3-

£

O

3
ID
0>
CA

C

n
ID

3*

ID
HH

•<
CA

rt
O
H
ID
h
0*
3
O
ID

HCA

Hi

•

O

HCA

rt

O

H>
H-

3

rt
ID
H
ID
CA
rt

Hrt
ID

rt
ID'

ID
CA
ID
H

H-

h
H-

H

0»
3
0>

H

H
X

ID
rt

3

CA
rt
h
C
g
ID

3

rt
CA

0>

<>
0
0»

P>

tr
HH*
Hrt

*<
»d

H
ID
CA
ID

3

rt

tr

H
ID

P>

CA

3

O
O
rt

3O

<
ID
H
3

H-

*Q
3rt

H

0

<

H
0>
rt
ID
CA

3

rt
HHi
HID
Cb

V

H
ID
O
HCA
ID
H

0
tr

«_i.
ID
O
rt
H-

<
ID
•
tc
O
5:
ID

»< *
• O
£

O

3*
0>
3

*

h
ID

3

CA
H-

O

(Q
ID
CA

<Q
3- 3

ID
H

<

H-

HiQ
rt
H

0>
rt
ID

HHi
HO

0>
3
rt

<
O

h-»

0»

rt
H-

»-*

Hrt

^<
P

H
ID

3
0

rt

iQ

3-

H I

3

3

O
ID
CA

O

n
0
c
3
rt
ID

a

O

CA
rt

o

ID

3

H-

rt
N.

3

P»

CL

ID
1Q

H

ID
ID

CA
rt

0»
3

O
ID
CA
O

0»
3
3
O

rt

H-«

rt

3-

M
HO

3O K
C

a

ID

0>

*0
0> O

HHi
ID
h
ID

3

»

h-*
rt

tr
a»
3
x
CA
rt

3

H-

3

TJ

3

fD
fjj

O
i—

a
H-

tr
ID

3

O

0*

rt
H-

O
3
0

Hi

3*
O

< ^

ID
H

ID
H

0»

»< a
<
0»
H

rt

a

H

a
o»

!-••
3

ID
*1

171

a
ID

rt

H-«

H-

P

H-

O

H-

rt
ID
I

3*
CL 0»

H

rt

H>

CA

3
<
*< aM- ID
o 3
»-•
0
0» 0>

^
ID

TJ
HO
rt
C
h
ID

H

0>

K

h

0

0

3

CA
H-

•a
0*

rt
HID
H

H
p>
h-«

Hi
»i

3

ID
CA
ID
»i

rt
H
HID
(A

TJ
0
H-

3

3'

<

ID

H-

3

ct
ID
h
ID
CA
rt
H

0>
rt
ID
CA

H-

O

S *<
M

3
TJ
1
H-

CA
rt

0>
3
O
ID

3

-

»0

P
3

O
H-

M
ID

"•
rt

3-

M-.
CA

3

ID

0>
CA

c

!-•

H*
rt

*<

tr
0

Q
3

3

c
CA
a
rt

ID

P
3
l-*
ID

<

ID
M

c
H
a

Hi

rt.
ID

H
ID

O

TJ
H
0

a

<J

H-

a
ID
o>
M
ID

0

0)

x

o
o
c

CA

H
0»
rt
ID

c

0*

H

O

c
3
a
rt

3-

H(A

a

TJ

HiQ

3

3*

0>

H

O
ID

3

rt
H

30»
3

M

3

P
3
?r

<Q
<Q

<
ID

H
ID
CA
ID
rt

P

tr

*

CA

a

P

rt
ID
CA

O
3

H

3

rt

P

rt
(D
CA

C
3

a
ID
H

P

TJ
O

h-»
H-

CA
rt

O
3

a

•

H
XT 3O

3

rt

H
P

H-

tr
P
3
X

HCO

O
3

a

3

ID
p
CO

<• c
H

rt
ID
M
ID
(A
rt
H
p
rt
ID

O
TJ

ID
*1
p
rt
H-

3

iQ

O

Hi

ID
MCA
O

0
3

CA
rt
»i
C

a
rt
ID

3

O

<
H-

O

3

<

iQ

ID
tl

3

HIQ

3*

rt
CA
CD
O

CA

3

<

3

rt
ID
HID

P

O
ID

O
ID

ID

Hp
rt
H-

t
xr
p.
o
o ID
XT CA
ID

a

o

K

P
3

H-

O

O

P

a*

iQ
ID

TJ

3

P
3

HCA

ID

rt
N
H-»

H«
HO

tr
P
3

3

<
ID

o
ID

O

<
ID

cr

O

0»

TJ
0

H-

CA

3

CA

ID

H-

Hi

H

rt
H-

ID

H
O

!-•
<
P ID

0>

MCA

H
iQ
ID

C

0
tr

<_i.
ID
O
rt
H-

a »0< a
tr
< UJ.
0»
H
ID
•^
O
H3*

H

CL

Tl
0»
H
rt
HO

H

P

rt
ID
CA

H

O
C
3

a

P

3

ID
p
3
P
CL
i_».

c
CA
rt
ID

•
•^
3CA
ID
O

O
3

c

3

Hi

iQ
ID
CA

rt

•<
H3

iQ

H3

O
3
O

Hi

O

(A
0
M

c

rt
ID

ID

H-»

vo

00
00
1
vo
I-*

T3
ID
N

TJ

H
ID
CA
ID

3

rt
CA
rt

€

O

rt

a

•-9
3*

ID

P

*1

O
C
3
P

Hi
H-

'M

CA
rt

^1
rt
i

a

P
K

ID
1

O

O

<

3

TJ
P
M

<

p
3
O
ID

rt

3*

0

Hi

ID

<
P

3-

H
H-

Hi

O

P

P

C

O

tr
H-

3

M
H*
rt

c

rt

M

HID
CA

.»<
0

O

<

ID
CA

<

0
3

ID

ID
Hi
H-

Hrt
CA

o

O
HM
Mrt
HID
CA

H
3

MiQ

3rt

M-

3

rt
ID
»i
ID
CA
rt

Hi

£

O
3

CA
H-

a

ID
H
ID

a

Hi
H-

3

O
H

*i
P
rt
ID

X
P

<
P

p
3

rt
ID
*1
ID
CA
rt

3-

H
P

a
ID

50

*<
•

rt
ID
CA

rt
(D

^
ID
H
P

^

H
H-

P

cr

HM
Hrt

ID
*1

H-

3

3
rt

ID
CA
ID

3

3*

3*

H

3

ID

rt
ID
CA

H-

C
3

<
ID
»

P

Hrt
ID

Hrt

a

HCA

rt

C

o
Q

HiQ

TJ
C

rt

rt

3*

tr

O

Hi

H

O

rt

P

H
ID
CA

!-••
3

0

ID

CA

M

ID
H-»
ID

0

3

C

H-

cr

tr

3

< a

P

H*

< !-••
0

ID
H

P

ID

>

ID

CA

ID
h
P
iQ
ID

O

P

rt
H-

rt

tr

cr

^
ID

rt

rt
ID
CA

a

3*
P
3

O
H

P

*ID

K

^1

0

rt
ID

•^
P

rt

a
3*

cr

HH«
Hrt

rct

tr

ID

P

Hi

3

CA

P
3

<
^
HP

M

HiQ

Hi

HQi
ID

tr

CA

O

a

cr

H.

P

< aID
<
M-

ID
N.
p
«Q
ID

P

*<

N.
p
iQ
ID

ID

CL

O

<
ID

P

ID

P

P

rt

tr

ID

3

rt
ID
H
ID
CA
rt
H.

P

rt
ID

P

rt
ID
CA

CA
ID
Hi

c

!-•

M

CA
CA
ID
(A
CA
M-

3
iQ
rt

3ID

CA

3




Kasman

indicate the degree of intraday interest rate variability, an
issue of some concern to U.S. policymakers.
The evidence also points to a relationship between required
reserves and overnight interest rate variability. In the United.*
Kingdom and Switzerland, the two countries operating with low,
nonbinding reserve requirements, overnight rates are much more
volatile than the rates elsewhere. In addition, in the United
States and Canada, where reserve deposits held at the central bank
have fallen in recent years, the decline in reserves has been
accompanied by rising interest rate variability.
These findings support the view that central banks face
greater difficulty in stabilizing interest rates around desired
levels when reserve requirements are eased. Nevertheless,
increased overnight interest rate volatility, per se, need not
erode the effectiveness of monetary policy, particularly if
fluctuations in overnight rates are transitory and do not reduce
the ability of market participants to identify the authorities'
policy intentions.
To assess whether overnight interest rate variability has
influenced the monetary transmission mechanism, one must determine
whether the overnight rate variability impacts on longer term
market interest rates. Table A.2 presents regression results
estimating the effect of overnight rate variability (MAD°) on the
measured volatility of three-month money market rates (MAD").37
As the Table shows, overnight rate variability is not
systematically related to three-month money market rate
divergences in the United States. Indeed, of the countries
surveyed, only Switzerland has large and statistically significant
coefficient estimates for transmission.
Perhaps a more important issue is whether interbank rate
volatility influences the transmission of changes in central bank
operating objectives to money market rates. To resolve this issue
in the case of the United States, one can test whether the federal
funds rate variability measure affects the response of three-month

37. In Table A.2 the volatility of interbank (MAD°) and threemonth money market rates (MAD") are measured as the absolute
deviation of rates adjusted for changes in the monetary policy
stance. For Switzerland, however, deviations around a thirty-day
centered moving average are used. Note that the results are
qualitatively unchanged by the choice of volatility measures.
-30-

Kasman

Treasury bill rates immediately after a change in the Open Market
Desk's federal funds rate objective. In the regression
ARt = c + (bl + b2 MAD°t.i)Afft + ^
ARt is the change in the three-month Treasury bill rate; MAD°t_i is
the average absolute deviation of the federal funds rate from the
Desk's objective, measured over the preceding objective period;
and Afft is the change in the Desk's federal fund objective.38
The coefficient estimate for b2 provides an indication of how
variability has affected the transmission of federal funds rate
changes.
The regression results are presented in Table A.3.
Estimates are given for the responsiveness of the three-month
Treasury bill on both the day of the federal funds rate change and
the five days following the change. As the Table shows, the
three-month Treasury bill rates rose on average 22 basis points in
response to a percentage point rise in the federal funds rate
objective on the day the objective increased. This response
increased to 26 basis points after five days. The variability of
federal funds rates does not appear to have altered this response.
In both regressions, the coefficient on variability is not
significant and enters with the wrong sign. Taken together, the
results suggest that federal funds rate variability, within the
range observed has not altered monetary policy transmission in the
United States.




38. This analysis closely follows earlier work by Timothy Cook
and Thomas Khan, "The Effect of Changes in the Federal Funds Rate
Target on Market Interest Rates in the 1970s," Journal Of Monetary
Economics, vol. 24 (1989), pp. 331-51.
-31-

Kasman

REFERENCES
Bank for International Settlements. Financial
Monetary Policy,
BIS, 1984.

Innovation

and

Bank of England. "Bank of England Operations in the Sterling
Money Market", Bank of England Quarterly,
October 1988.
Bartholomae, Andre. "Some Operational and Instrumental Aspects
of Monetary Targeting in Germany". Unpublished paper
presented at XXVIII meeting of Technicians of Central Banks
of the American Continent.
Batten, Dallas S. and Michael Blackwell et.al. "The Conduct of
Monetary Policy in the Major Industrial Countries". IMF
Occasional

Paper.

No. 70, July 1990.

Bernanke, Ben and Frederic Mishkin. "Central Bank Behavior and
the Strategy of Monetary Policy: Observations from Six
Industrialized Countries". Mimeo.
Clinton, Kevin. "Bank of Canada Cash Management:
Technique for Implementing Monetary Policy".
Review, January 1991.

The Main
Bank of Canada

Cook, Timothy and Thomas Khan. "The Effect of Changes in the
Federal Funds Rate Target on Market Interest Rates in the
197 0s."

Journal

of Monetary

Economics,

vol. 24 (1989),

pp. 331-51.
Dale, Spencer. "The Effect of Changes in Official UK Rates on
Market Interest Rates Since 1987". Unpublished, Bank of
England.
Federal Reserve Bank of New York. "Monetary Policy Operations
during 1990". Quarterly Review, Spring 1991.
. "Monetary Policy Operations during 1991".
Review, Spring 1992.

Quarterly

Goodfriend, Marvin. "Interest Rates and the Conduct of Monetary
Policy". Carnegie-Rochester
Conference Series on Public
Policy,
no. 34 (1991), pp. 7-39.




-32-




Kasman

Kasman, Bruce and Anthony Rodrigues. "Financial Reform and
Monetary Control in Japan". FRBNY Quarterly Review, Autumn
1991.
Kneeshaw J.T. and P. Van den Bergh. "Changes in Central Bank
Money Market Operating Procedures in the 1980s". BIS
Economic Papers, No. 23, January 1989.
Longworth, David and Patrice Muller. "Implementation of Monetary
Policy in Canada with Same-Day Settlement: Issues and
Alternatives".

Bank of Canada Working

Meulendyke, Ann-Marie.

Markets.

U.S.

Monetary

Policy

Paper,
and

No. 91-3.
Financial

Federal Reserve Bank of New York, 1989.

Ministry of Finance and Bank of Japan. Japan's Short-Term Money
Market and Issues.
Money Market Study Group, August 1991.
Organization for Economic Cooperation and Development. OECD
Economic

Survey-Switzerland,

OECD (Paris, 1989).

Poole, William. "Optimal Choice of Monetary Policy Instruments
in a Simple Stochastic Macro Model". Quarterly Journal of
Economics, vol. 84 (1970), pp. 197-216.
Swiss National Bank.
Temperton, Paul.

U.K.

Annual Report
Monetary

various issues.

Policy:

The Challenge

for

1990s.

London, MacMillan 1991.
Thornton, Daniel L. "The Borrowed Reserves Operating Procedure:
Theory and Evidence". FRB St. Louis Review, January/February
1988.

-33-

1.

Structure of Short-Term Interest Rates

Country

Official Rates

Overnight Interest Rates

Other Key Interest Rates

United States

Discount rate

Federal funds rate

Treasury bill rate

Germany

Discount rate

Day-to-day money rate

Repurchase agreement
rate (one-to twomonth)

Lombard rate

Three-month interbank
loan rate

Treasury bill
selling rate
Interbank call money
rate

Japan

Certificate of deposit
rate (three month)

Discount rate
Bill discount rate
i

Overnight interbank
rate

United Kingdom

Bank of England dealing
rate

No posted rate
Commercial bank base
lending rate
Three-month interbank
loan rate
Bank Rate

Canada

Money market
financing rate

Three-month treasury
bill tender rate
Ninety-day prime
corporate paper rate

Switzerland1

Discount rate

Call money rate

Three-month Euro-franc
deposit rate

Lombard rate

Note:
1.



Each central bank's primary interest rate objective appears in bold face type.

The Swiss National Bank does not employ interest rate operating objectives.

1.

United Kingdom: Short-Term Interest Rates
Weekly Observations, Wednesdays

Percent

16

Base Rate,
14

i

p ^ W / i V " " ^

|f=*N

\ *

i

. :
lil '
y p a p f t s J 11

^
Stop Rate

^w
'

12
en
i




CA

'\

l< Overnight
I LIBOR

10

I
1988

I

1989

I

I

I

L
1990

J
1991

L

2.

Japan: Short-Term Interest Rates
Weekly Observations, Wednesdays

Percent
9

1,

.

-

Three-month certificate
of deposit rate . #

#.

biP^^i
wi
j v

•

•• • » A y \ /

?•••

J^\/*
/

0
m

Overnight
call money rate

J

1 1 i l l

#
#

\

^

•L

- — i

»

t
II i i

1988
* Values are month-end observations.




Discount rate *

»

^J^^l
I I I

Va

(A

--Y
I I I

*

••vv

• / If
i

•I

M

••••••• J

ay

* «L

1 i

i 1 i i 1 i
1989

i i

1 1 1 1 i l l
1990

1 1 1_L

i i 1 i i l.-LX-L„i. i !
1991

iill
1992

3

*

M W ^ u ? r l t n d : R e s e r v e Deposits and Interest Rat*
Monthly Averages
Billions of SF

Percent

12

U *,
,\

10 h

12

:

1

:•% « »
:
J
:,'*'-""»
'; . Lombard1
?
\
' i \ /%» Rate
Hio
» /
:
*»
\ ':
i
•-•-'•-»-•-»»••-«-....-....*
iI
;

\

x

x

Resen/e 'll
Deposits 1: \

v
^
'
V'i'
8 [~
h

.--ij

M

i
CO
I

/

/

i \

;/

^

\

; /

6 H"

Vv

i Overnight
• Call Money
• Rate

"j
H 6

/V

\:
\•
*

i

A

4 H

./:

(A

•

fV

H 8

1




1

H 4

/ \

"™ . ^ *"• • 1

J

H 2

2 K

Li

i

i
1987

i

r„.i

__L

1988

i

i

i

l
1989

i

i

t

i
1990

i

i

i

l
1991

l

i

1 1
1992

4.

Germany: Short-Term Interest Rates
Weekly Observations, Wednesdays

Percent

I




p»

1988

1989

1990

1991

1992

Canada: Short-Term Interest Rates
Weekly Average

Percent

i

u>




(A

1988

1989

1990

1991

At Thursday tender. The central bank lending rate (bank rate) is set 1/4 percentage point above this rate.

1992

2.

Instruments for Reserve Management

Country
United States

Germany

Japan

Primary Short-Term Reserve Management Tool
Instrument
Activity
Repurchase
agreement

Government
security

Matched
purchase
and sale

Government
security

Repurchase
agreement

Government
security

Repurchase
agreement

Commercial bills,
government
securities

Discount
window
lending

Other Operations
Activity
Instrument
Purchase
or sale

Government
security

Purchase or
sale

Government
security

Foreign
exchange

Swap

Purchase or
sale

Government
security

Repurchase
agreement

Commercial
paper

United Kingdom

Purchase or
sale

Government
security,
commercial bills

Repurchase
agreement

Government
security

Canada

Drawdown/
redeposit

Government
deposits

Purchase or
sale or
repurchase
agreement

Government
security

Switzerland

Foreign
exchange

Swaps

Purchase or
sale

Cantonal and
bank bonds

Drawdown/
redeposit

Government
deposits




3.

Central Bank Lending Facilities

United
States
1) Credit available at
below market rates
Access restricted by:
Q = quotas,
D = administrative
discretion

Germany

Japan

Yes

Yes

Yes

Q,D

Q

Q,D

Canada

Switzerland

No

No

Yes

Yes

Yes

Yes

P,D*

Interest rate setting:
P = posted rate
D = set at discretion
of central bank
2) Other credit
sources available

United
Kingdom

No

Yes

No

Access restricted by:
Q = quotas,
D = administrative
discretion
0 = other
Interest rate setting:
F = Floats in relation
to market rate
P = Posted rate
D = set at discretion
of central bank

1. The Bank of Japan provides credit at the official discount rate. The Bank can add or call loans at
will, however, and interest charged is calculated on the period of the loan plus one day. The effective
cost of borrowing thus rises as the maturity of a loan is reduced.
2.

Generally

non-binding.

3. Bank of Canada advances are provided only for overdrafts to meet a deficiency of clearing balances or
for an end-of-averaging period reserve deficiency.




4.

Central Bank Lending as a Share of Central Bank Assets
(Annual Average of End-of-Month Observations)

1985

1988

1991

United States

0.7

0.9

0.1

Japan

8.4

13.6.

12.1

29.4

22.5

25.0

United Kingdom

1.8

1.8

3.0

Canada

7.4
7.4

2.2
2.2

2.0
2.0

Switzerland

9.9

0.9

1.2

Germany

^




J"?

Q>

Reserve Requirement Regulations

i
Ui
i

United
States

Japan

Germany

United
Kingdom

Canada*

Length of reserve accounting
period

14 days

1 month

1 month

6 months

1 month

3 months

Length of maintenance period

14 days

1 month

1 month

6 months

15 days

1 month

Interval from end of accounting
period to end of maintenance
period

2 days

15 days

15 days

180 days

30/45 days

50 days

Highest reserve ratio for
demand deposits

10

1.3

12.1

0.5

10

2.51

1.2

4.95

0.5

Yes
No

Yes
No

No
No

Yes
No

Yes
No

No

Up to
50 percent

No

Yes

Yes

No

No

No

No

Highest reserve ratio for
other deposits
Averaging provisions
Carryover provisions
Vault cash satisfies
requirement

1.

0.5
w

Yes
Yes
Yes

Penalty for reserve deficiency
(percentage above central
bank lending rate)
Interest paid on reserves

Switzerland

3-5

No

No

Includes time deposits with a term to maturity up to three months.

* As of June 1992, reserve ratios were eliminated in Canada as part of a planned phaseout of required reserves.
Currently required reserves are set at a predetermined amount; this amount will decline to zero in 1994. The maintenance
period has been extended to one month. Banks incurring a reserve deficiency pay a penalty calculated at the Bank rate.




6.

Reserve Deposits Held at Central Banks as a Share of Total Bank Liabilities
(Year Average of End-Month Observations, in percent)

1980

1985

1988

1991

United States

1.6

0.8

1.0

0.6

Japan

1.6

1,1

1.0

1.0

Germany

7.2

5.6

5.5

5.5

Switzerland

4.0

3.1

1.7

0.7

United Kingdom

0.3*

0.1

0.1

0.1

Canada

3.9

1.4

0.8

0.4

* Figure is for year-end 1981.




to

A.l.

Overnight Interest Rate Variability
(Mean Absolute Deviation of Daily Observations, in Basis Points)

Deviations from Thirty-day
Centered Moving Average
1988

1989

1990

1991

Average
1988-91

1988

12.3

11.9

12.3

21.1

14.4

13.0

11.8

12.8

18.5

14.0

8.7

8.5

7.1

8.4

8.2

12.5

8.5

7.4

5.8

8.6

Germany

15.7

18.2

13.6

13.4

15.2

15.8

17.4

14.5

14.8

15.6

United Kingdom

50.4

32.9

14.8

25.3

30.9

52.5

39.7

14.2

25.0

32.9

9.7

13.4

21.3

28.7

18.3

11.0

15.7

21.3

28.8

19.2

33.8

34.8

37.8

35.5

United States
Japan

1

Deviations from Mean Adjusted
for Changes in Policy Stance1

en
i

Canada
Switzerland

1989

1990

1991

Average
1988-91

Note: Overnight interest rates are the effective overnight Fed funds rate (the United States), overnight
call rate (Japan), day-to-day money rate (Germany), London interbank offer rate (the United Kingdom), overnight
money market financing rate (Canada), and overnight call rate (Switzerland).
1. Values are average absolute deviations of overnight rates from a mean that changes along with estimated
shifts in central bank interest rate operating objectives.




(A

A.2.

The Transmission of Overnight Rate Variability to the Variability of Three-Month Money Market Rates
(Based on Monthly Observations, 1988-1991)

MAD",, = C + B MAD°t + Ut

.

B

j?

DW

United States

0,
0.12
.12
((4.79)
4 ,. 7 9 )

--0.16
0 ..16
(-0..95)
(-0.95)

-0.01

2.23

Japan

0,
.04
0.04
((0.90)
0 ,. 9 0 )

0. . 2 2
0.22
((0.41)
0 ,. 4 1 )

-0.01

2.34

Germany

0. . 0 5
0.05
( 1 .. 4 6 )

0,
.25
0.25
( 1 .. 2 8 )

-0.02

1.92

United Kingdom

0.
,14
0.14
((7.14)
7 ,. 1 4 )

--0.01
0 ,.01
((-.14)
-. .14)

-0.01

1.67

Canada

0,
0.05
.05
((3.71)
3. 71)

0,
0.04
.04
( 0 .. 5 8 )
(0.58)

0.10

1.90

0. . 7 0 *
0.70*
( 2 .. 0 7 )

0.23

2.32

Switzerland1

-0.13
- 0 . .13
(-0.,79)

Note: Equation is estimated using instrumental variables. Instruments include lagged MAD0 and lagged
levels of interbank interest rates. Overnight interest rates are those described in Table A.l. Three month
money market rates are the three-month Treasury bill rate (the United States and Canada), Gensaki rate
(Japan), three-month interbank loan rate (Germany, Switzerland) and the three-month Sterling interbank
deposit rate (the United Kingdom).
1.

Sample covers June 1989-December 1991.

* Significant at 5 percent level.




A.3.

I n t e r e s t Rate V o l a t i l i t y and t h e Transmission of Changes in Federal Funds Rate O b j e c t i v e s :

ARt = C + (Bl + B2 MAD°t_i)Afft

1988-1991

+ ji t

B,

DW

Response of three-month Bill Rates (ARt)
Day of Federal Fund Objective
Change

- 0 , .02
( - 1 , .51)

0.22**
(4.03)

0.06
(0.22)

51

Five Days Following Federal
Fund Objective Change

- 0 , .38
( - 1 . .39)

0.26*
(2.42)

0.58
(1.31)

.40

1.86
7*

i

i

* Significant at the 5 percent level
** Significant at the 1 percent level




2.25




COMMENTS

Stephen A. Meyer1

The goal of these two papers is to develop an understanding of
monetary pplicy operating procedures in countries other than the
U.S. in the hope that we can learn something applicable to U.S.
monetary policy.

Potentially the most useful insights in these

two papers come from examining operating procedures in those
countries, such as Switzerland and the U.K., that have had low,
non-binding reserve requirements.

In those countries, low or

nonexistent reserve requirements combine with tight restrictions
on daylight overdrafts to create a situation in which required
reserve balances are lower than the reserve deposits that banks
need to hold to clear payments through the central bank.
is moving toward such a regime.

The U.S.

Now that reserve requirements on

transactions deposits have been lowered, we are likely to find
ourselves in that situation during the early months of each year.
As Anne-Marie Meulendyke's paper for this conference notes,
those responsible for implementing monetary policy in the U.S. are
concerned that reserve requirements now are low enough that banks'
reserve deposits sometimes will be lower than the operating
balances they need to clear payments.

Under current operating

procedures, the Desk attempts to make the supply of non-borrowed
reserves roughly equal to the forecasted demand each day.

If the

forecast is wrong, especially near the end of a reservemaintenance period, the federal funds rate will deviate from its
target.

Non-binding reserve requirements, because they give rise

1. Vice President and Associate Director of Research, Federal
Reserve Bank of Philadelphia, and Adjunct Professor of Finance,
The Wharton School, University of Pennsylvania.

Meyer

to larger errors in forecasting the demand for reserves, will
generate more variability in the funds rate unless U.S. operating
procedures are changed to compensate.
Most economists, including those in the Federal Reserve
System, are not convinced that more variability in the funds rate
would be harmful.

But U.S. policymakers have revealed an aversion

to funds rate volatility, at least in part out of concern that
greater funds rate variability will reduce the Fed's ability to
communicate the stance of monetary policy to the markets.
Policymakers abroad share that aversion.

Given policymakers'

aversion to interest rate volatility, we would do well to learn
whether other countries with low reserve requirements have
designed operating procedures that yield less day-to-day interest
rate volatility than does our current procedure.

Or perhaps we

can learn how to construct operating procedures that clearly
communicate policymakers' intent to the markets despite volatility
in short-term rates.

THE PAPERS
The papers by Kasman and by Morton and Wood treat operating
procedures in major industrial countries, so they do overlap.
Nonetheless, the papers neatly complement one another.

Reading

the two papers together, we learn about the evolution of operating
procedures over time, including how central banks have adapted to
changing financial conditions, and also about current practices
for the day-to-day implementation of monetary policy.

Both papers

clearly indicate that operating procedures converged to a large
extent during the 1980s, with central banks of all the countries
examined now using a short-term interest rate as their operating
instrument.

Nonetheless, some important differences remain.

From Bruce Kasman's paper we learn about differences in the
day-to-day operating procedures used to smooth short-term interest




-2-




Meyer

rates.

Central banks rely mostly on open market operations in one

or more financial assets, but non-market techniques such as direct
lending to commercial banks (discount window lending, in U.S.
parlance) or shifting government deposits between the central bank
and commercial banks are used, too.

From the paper by John Morton

and Paul Wood we learn how the institutional context in which dayto-day operations take place has changed over time, and how
monetary policy procedures have changed in response.

While we

learn a great deal from the papers, we learn less that 1 would
like about the roles that the European Monetary System and
international integration of money markets have played in the
development of operating procedures.

I suspect they have been

important, but the papers do not tell us how important.
As the authors note, interest rate operating procedures are
appropriate under some conditions.

They were adopted or revived

during the 1980s by all seven of the countries examined in these
two papers as the seemingly robust empirical relationship between
money, interest rates, and economic activity appeared to break
down under the pressure of continuing deregulation of financial
firms, spreading financial innovations, increasing international
mobility of financial capital, and declining costs for financial
transactions.

The common movement toward interest-rate operating

procedures was also driven, in part, by policymakers' perceptions
that reserves or money supply targeting allowed too much interest
rate volatility.
I do not have a comparative advantage in knowledge of the
details of other countries operating procedures.

Furthermore,

reading some relevant literature and discussing the topic with a
few participants in foreign financial markets reveals that the
authors have done a generally good job of laying out those
details.

Thus I want to step back from the details and try to put

the information presented in these two papers into perspective by

-3-

Meyer

discussing, at least in a general way, some analytic issues. I
hope to focus the material in the papers on the search for
alternative operating procedures before we turn to the lessons
that we might draw from the experience of countries with nonbinding reserve requirements.
INTEREST RATE OPERATING PROCEDURES
I turn first to a brief discussion of the objectives of interest
rate operating procedures, and then to a broad-brush
characterization of different ways to structure such procedures.
I will focus on the implications of those structures for short-run
variability in interbank interest rates. That will lead to some
comments on the statistics presented in the two papers.

Finally,

I will offer comments on what we might adapt for use in the U.S.

The Objectives of Interest-Rate Operating Procedures
At the macroeconomic level, the objective of any monetary policy
operating procedure is to achieve policymakers' desired outcomes
for real GDP growth, inflation, or other macroeconomic variables.
At the tactical level, we can identify at least four not-always compatible objectives of day-to-day operating procedures:

(1) to

set the short-run operating instrument at a level believed
consistent with policymakers' desired outcomes for intermediate
targets such as money growth or the exchange rate, or for final
goal variables; (2) to smooth day-to-day variability in shortterm interest rates while nonetheless allowing the level of
interest rates to move in response to "goods market" shocks; (3)
to convey information about the stance of monetary policy to the
markets -- sometimes clearly and sometimes not; and (4) to extract
information about economic and financial shocks from the markets.
Trade-offs among these objectives condition the design of
interest-rate operating procedures.




-4-

Meyer

The Design of Interest-Rate Operating Procedures.
As both papers note, an interest-rate operating procedure is, in
practice, a set of techniques for managing the supply of bank
reserves so as to keep the supply of reserves equal to the demand
at the target interest rate.

I will focus on market-related

procedures.
The Simplest Procedure.

Perhaps the simplest procedure is one in

which the central bank creates a perfectly elastic supply schedule
for reserves by posting an interest rate and announcing that it
will supply any quantity, of reserves that banks want to obtain at
that rate. A central bank could implement such a procedure either
by posting a rate at which it will freely lend to banks, or by
posting a yield at which it will buy and sell some short-term
financial instrument.
Clearly this simple procedure allows policymakers to set the
operating instrument exactly at its target level. Changes in the
target level are immediately observable, so this procedure
provides full information about the stance of monetary policy to
financial markets.

Central banks in some of the countries

considered in these two papers used to follow such a procedure,
more or less, but no longer do so. The papers indicate that this
procedure was dropped because it was perceived to slow the
response of interest rates to shocks. That statement must be an
argument about policymakers' willingness to change the target
level of interest rates or about the need for political cover,
rather than an argument that the simple fixed-interest-rate
procedure does not extract information about shocks. The observed
change in the quantity of reserves that results from a shock under
a fixed-interest-rate procedure provides the same information
about the economy as would the change in interest rates under a
variable-interest-rate operating procedure, so long as we know the




-5-

Meyer

interest elasticity of the supply and demand for reserves.
Evidently the simple fixed-rate operating procedure yields too
little interest rate variability.

More Flexible Procedures.

As an alternative, we can construct

interest rate operating procedures that generate an inelastic or
less-than-perfectly elastic supply of reserves over some range of
interest rates, but that also keep rates from moving out of that
range.

Such procedures would allow shocks to immediately affect

interest rates, and would also provide less information about
policymakers intentions or targets by allowing some day-to-day
interest rate variability.
One possibility is a two-tier lending mechanism, as in
Germany.

The central bank posts one interest rate at which banks

can borrow up to a rationed amount, usually not quite enough to
satisfy their total demand for reserves at that rate, and a
second, higher interest rate at which banks can borrow freely.

A

second possibility uses a two tier intervention or repurchase rate
to produce the same result, as in France.

The central bank offers

to buy government securities at a yield it chooses, up to some
maximum quantity that is less than needed to satisfy banks' total
demand for reserves at that yield.

The central bank also offers

to buy securities at banks' initiative, but at a higher yield.
The central bank could either buy very short-term securities
outright, or buy longer-term securities through repurchase
agreements.
Both of these procedures would generate a supply schedule
for bank reserves that is a step function.

These two procedures

would allow shocks to reserve demand to affect the level of
interbank interest rates, at least within the range defined by the
lower and upper discount or intervention rates.

By

choosing the

spread between upper and lower rates, policymakers can control the




-6-




Meyer

maximum volatility in the targeted rate.

By allowing the targeted

interest rate to vary somewhat on a day-to-day basis, these
procedures can maintain some flexibility or ambiguity about the
central banks' exact target.

And by announcing changes in the

lower and upper rates, policymakers can make clear announcements
about the stance of monetary policy when they need to do so.
Policymakers might feel that a vertical step in the supply
schedule allows too much day-to-day variability in the targeted
rate.

In another variation on operating procedures, the central

bank can manage the supply of reserves to get virtually any
positive slope for the portion of the reserve supply schedule
between the lower and upper intervention rates.

By undertaking

short-term repurchase agreements or foreign exchange swaps in
response to forecasts of changes in the demand for reserves or to
observed variations in interbank interest rates, or by shifting
government deposits between commercial banks and the central bank,
those responsible for implementing policy can generate an upward
sloping supply schedule for bank reserves over a range of interest
rates between the lower and upper lending or intervention rates.
Of course the upper and lower intervention rates can be used
at the margin, rather than to provide the bulk of banks' reserves.
The bulk of reserve deposits can be provided through outright
purchases of securities, as in the U.S., Canada, and the U.K., or
through long-term repurchase agreements as in Germany, or through
foreign exchange swaps as in Switzerland.

Thus the central bank

can manage not only the slope of the upward-sloping portion of the
reserve supply schedule but also its position.

These mechanisms

for providing the bulk of reserves can be biased toward keeping
the banking system short of reserves on average, making it likely
that interbank rates will trade near the top of the range defined
by the upper and lower intervention rates.

Or they can be biased

toward keeping the banking system flush with reserves on average,

-7-

Meyer

so that interbank rates tend to trade near the bottom of the
range.
Do Flexible Operating Procedures Meet the Stated Objectives?
How does such an operating procedure stack up relative to the
goals 1 discussed earlier?

Such a mixed strategy can provide a

well defined trading range for the targeted interest rate, thus
limiting interest rate volatility.

The targeted rate will respond

to shocks in the demand for reserves, allowing policymakers to
observe such shocks. The provision of reserves can be biased so
as to keep the targeted rate near any desired level within the
range, allowing policymakers not only to hit a specific interest
rate target on average, but also giving them flexibility to make
adjustments to their target without making explicit announcements
of such changes. And changes in the top or bottom of the range
can provide clear signals of changes in the stance of monetary
policy.

Other Countries Use These Mixed Strategies.

This mixed strategy

of lower and upper discount or intervention rates with an upward
sloping supply of reserves between them is a reasonably good
characterization of operating procedures used in Germany and
France.

It is a less good, but still reasonable characterization

of operating procedures in Canada and the U.K.

In Germany the

spread between the lower and upper lending rates is quite large,
usually around 200 basis points.

In the other countries the range

is much narrower.

The U.S. Does Not. U.S. operating procedures, in contrast,
generate no clearly identifiable minimum or maximum for the target
interest rate. There is no minimum -- other than zero -- because
discount window borrowing is so sharply restricted by




-8-




Meyer

administrative controls; there is no maximum because the Desk
enters the markets at most once each day and there is no "Lombard
facility."

We can characterize the current U.S. operating

procedure as one in which a nearly vertical supply curve for
reserves is placed, daily, so that it intersects a forecast of
that day's demand for reserves at the target federal funds rate.
The reserves supply schedule is "nearly vertical," rather than
vertical, for three reasons:

(1) the Desk does shade its forecast

of the demand for reserves up or down in response to movements in
the funds rate; (2) dealers can and do withdraw from repurchase
agreements with the Fed when market yields fall; and (3) discount
window borrowing still responds a little to changes in the spread
between the fed funds rate and the discount rate. With this
nearly-vertical supply schedule, we sometimes see very large daily
movements in the funds rate, particularly on the last day of
reserve maintenance periods.

COMPARING OPERATING PROCEDURES AND INTEREST-RATE VOLATILITY
This discussion of operating procedures, along with the earlier
discussion of the effects of non-binding reserve requirements,
might lead us to expect that interbank interest rates in the U.S.
would be more variable than those in other countries, except
perhaps Switzerland and the U.K.

That conclusion turns out to be

half right, as the papers by Kasman and by Morton and Wood
indicate.

Interbank interest rates do seem to deviate more from

their targets in Switzerland and the U.K. than in the U.S., but
there is no apparent difference in the variability of U.S.
interbank rates and those of the remaining countries.
Three cautionary notes on the interpretation of the
statistics on interest rate variability presented in the papers
are in order.

First, as Kasman argues, we do not want to confuse

changes in interest rate targets with the variability in interest

-9-

Meyer

rates that occurs around unchanged targets.

Presumably it is only

the latter kind of variability that policymakers find disturbing.
As the charts presented in the two papers show, changes in target
rates were quite frequent in some countries.

For that reason I

find it difficult to interpret the statistics presented by Morton
and Wood; the average of standard deviations of daily interest
rate changes around monthly means, for example, would correspond
to unintended interest rate variability only if policymakers never
changed interest rate targets except at the turn of the month.
Second, as Kasman himself notes, his statistics for Switzerland
and Japan are based on a less accurate identification of the
central banks' target rates than is the case for other countries.
Thus we should not be all that confident that the variability
reported for Switzerland gives us an accurate measure of the
effects of non-binding reserve requirements.

Third, I suspect

that the observed differences in interest rate variability reflect
not only differences in operating procedures but also differences
in policymakers' aversion to interest rate variability.

Thus I am

reluctant to draw strong conclusions about the effects of
operating procedures on interest rate variability from the
statistics presented in these two papers without knowing more
about how much variability each country's policymakers find
acceptable.

WHAT MIGHT THE U.S. ADAPT FROM OTHER'S OPERATING PROCEDURES?
I will conclude with a possibly provocative suggestion on what the
Federal Reserve might adapt from other countries' operating
procedures.

I offer this suggestion in the hope that it will

stimulate discussion and lead to wide-ranging consideration of
alternatives.

I should make clear that I have not worked out all

necessary details, nor, I am sure, have I thought of all potential
problems.




-10-




Meyer

A Suggestion for U.S. Monetary Policy Operating Procedures
My suggestion is that the Federal Reserve augment its current
operating procedures by setting up either a second discount rate - a penalty discount rate modeled loosely on German practice --or
by setting up a penalty-rate repurchase facility modeled loosely
on French practice. Adding either of these could serve as a first
step in modifying U.S. operating procedures.
To set up a two tier discount rate, the Federal Reserve
would establish the equivalent of a "Lombard rate" --an
additional, higher discount rate at which banks could borrow
freely against eligible collateral.

That second rate would be

higher than the target federal funds rate and would exist
alongside the current subsidy discount rate.

Banks would be able

to borrow at the lower, subsidy discount rate only in the event of
truly unforseen reserve shortfalls due to events such as computer
failures or wire transfer delays. But banks would have ready
access to borrowing at the higher "Lombard rate."
To set up a penalty-rate repurchase facility, the Federal
Reserve could announce that it stands ready to provide reserves to
banks through short-term repurchase agreements arranged at banks'
initiative, but at a rate that would be set above the target
federal funds rate.

Such a facility would require that banks have

appropriate collateral; it also would require the Federal Reserve
to put in place safeguards to limit counterparty risk.
Potential benefits of Modifying U.S. Operating Procedures.
Either of these facilities would provide a backup source of
liquidity to the banking system when reserve deposits plus
clearing balances fall short of balances needed for funds transfer
purposes, or when the Desk underestimates the demand for reserves.
Either facility could prevent spikes in the federal funds rate
such as we have seen on some end-of-maintenance-period Wednesdays.

-11-

Meyer

By establishing clear, although perhaps broad, limits on
movements in the funds rate, either of these facilities might
reduce policymakers' concerns that day-to-day movements in the
federal funds rate could be misinterpreted as a change in monetary
policy.

That is particularly likely if market participants knew

that significant changes in the stance of monetary policy would be
signaled by changes in the penalty discount rate or repurchase
rate, and perhaps also by changes in the subsidy discount rate.
Seemingly paradoxically, a change in operating procedures that
would prevent large movements in the federal funds rate could
actually allow more day-to-day volatility by reducing markets'
reliance on changes in the funds rate as an indicator of the
stance of monetary policy.
In addition, the existence of a liquidity safety valve would
reduce the Desk's need to match the supply of reserves to the
predicted demand each day.

The Desk might well be more able to

focus on the reserve need for the maintenance period as a whole,
and thus be free to conduct fewer daily open market operations
aimed at smoothing the funds rate.
Finally, to the extent that the proposed changes allow
greater day-to-day variability in the funds rate, they will enable
market forces to move the average level of the funds rate more
readily than is the case today. Those movements, in turn, might
allow policymakers greater flexibility in making a series of small
changes in their federal funds rate target, at least within the
band defined by the subsidy discount rate and the penalty rate.




-12-

MONETARY TRANSMISSION CHANNELS IN MAJOR FOREIGN INDUSTRIAL COUNTRIES
Robert B. Kahn and Linda S. Kole

The past two decades have witnessed far reaching transformations of
financial markets in major foreign industrial countries. The process of
financial liberalization that has taken place abroad in many ways
parallels changes that have occurred in the United States. Although the
specific circumstances of individual countries have differed, there have
been important common elements, including the introduction of new
financial assets and markets, fuller integration of domestic and
international financial markets, greater reliance on market-determined
interest rates, and significant structural change in banking systems.
The changes in financial markets that occurred in most of the major
industrial countries in part reflected the response to a common set of
global economic forces. These changes, in turn, had an impact on the
conduct of monetary policy and how monetary policy actions fed through to
the real economy.




This paper characterizes the main financial channels through which
monetary policy affects real economic activity in major industrial
countries, and analyzes whether and how these transmission channels have
changed during the past two decades. Because of the broad nature of the
question, we have chosen to limit our country coverage. Our primary
focus in this paper is on Japan, Germany, and the United Kingdom, three
countries that have had a wide range of diverse experiences with
financial deregulation and monetary control. We find evidence that in
all three of these countries, wealth is crucial in the determination of
money demand, the first link in the monetary transmission channel.
Further the demand for broad money seems to have become more portfolio

1. Linda S. Kole is on the staff of the International Finance Division
of the Federal Reserve Board. Robert B. Kahn, formerly a member of the
staff of the International Finance Division of the Board, is now on the
staff of the International Monetary Fund. We thank Hali Edison, Neil
Ericsson, Mike Gavin, Craig Hakkio, Dale Henderson, Karen Johnson, Steve
Kamin, Eric Leeper, Eileen Mauskopf, and Larry Promisel for their
comments. Peter Fishman, John Maluccio, and Tina Sun provided exemplary
research assistance.
- 1-

Kahn and Kole

based and interest sensitive in the past decade.

We also find evidence

that different monetary variables affect real economic activity
differently across countries, but that the role of the term structure and
the exchange rate seems to be increasing in the transmission channel.
The next section outlines reasons why monetary policy transmission
mechanisms have changed in the past two decades.

The third section then

traces out a basic Keynesian model to illustrate the importance of asset
prices in the monetary transmission process.

Section 4 analyzes the

relationship between money, interest rates, income, and wealth by
estimating money demand functions and considering how they have changed
over time.

Section 5 then examines the links between various instruments

of monetary policy and real variables.

Finally, in the last section we

summarize our results and explore possible extensions for future work.

MONETARY POLICY TRANSMISSION CHANNELS,

1970-1990

The reasons why monetary transmission mechanisms may have changed in the
past two decades are well known.

They include:

Financial liberalization and innovation.

In the early 1970s, many

foreign industrial countries had relatively underdeveloped financial
systems that limited the channels through which monetary policy
influenced economic activity.

Subsequent changes in the transmission of

monetary policy in these countries generally reflected governments'
attempts to modernize and integrate their financial markets in a changing
world economic environment.

In Japan, financial markets were tightly

controlled and segmented in the early 1970s.

Major bank deposit rates

were regulated while lending rates were closely linked to the Bank of
Japan's discount rate.

Bond markets were small and underdeveloped, and

international transactions tightly controlled.

Thus, credit was

effectively rationed, and bank lending was the main channel through which
monetary policy influenced the economy.

Japan liberalized its capital

markets later than many other industrialized countries.
1980s, deregulation gradually gained momentum.

In the early

Controls on domestic

interest rates and on external capital flows were dismantled, and new
financial products proliferated.
There was also substantial liberalization in the United Kingdom
over the course of the 1970s and 1980s.




- 2 -

Domestic financial markets were

Kahn and Kole

quite segmented (e.g. between the clearing banks and the building
societies) in the early 1970s, restricting competition for both loans and
deposits.

In the 1970s, monetary policy mainly operated through

restrictive guidance of bank and building society lending.

Between 1973

and 1979, the "corset", in essence a tax on the expansion of U.K. banks'
liabilities (and thus credit) was intermittently activated, although
eventually it became ineffectual.

When the Conservatives came to power

in 1979, the new government recognized that the controls were
increasingly ineffective and initiated major financial market reforms.
Most of the restrictions on lending and exchange controls were abolished
in 1979.

The corset was scrapped in 1980, and reserve requirements were

abolished in 1981.

A series of other regulatory changes followed that

culminated in the collapse of the building societies' cartel in 1983 and
the Big Bang in 1986.

Deregulation progressively put banks and building

societies on a more even footing, broke down the separation of the
mortgage market from other forms of personal credit, and freed consumer
credit.
In contrast, German domestic financial market liberalization was
already advanced by 1970, even though markets remain cartelized and
underdeveloped.

All interest rate controls were abolished in 1967.

Exchange controls were removed in the 1950s.

Such an environment might

be expected to be one with particularly strong international linkages,
and international financial innovation has made a mark on German capital
markets, although relatively few new financial instruments have been
introduced there.

One special aspect of German financial markets has

been their proximity to Luxembourg, a financial center that is




considerably more developed than German markets.

During the past two

decades, Luxembourg has often served as an outlet for German portfolio
shifts, providing Germans with the full array of instruments (offered by
German banks) that their own market lacks.

Without Luxembourg, German

financial innovation would undoubtedly have proceeded at a faster pace.
With the approach of a fully integrated European financial market, the
process should accelerate.
It is often noted that one factor influencing liberalization of
financial markets during the 1970s and 1980s was the sharp increase in
industrial countries' interest rates and inflation rates after 1973.

- 3 -

Kahn and Kole

Chart 1 depicts short-term interest rates and consumer price inflation
from 1973 to the present in major foreign industrial economies. The
increase in nominal and real interest rates created incentives for asset
holders to reduce their balances bearing below-market interest rates. At
the same time, the growth in government deficits in a number of countries
contributed to the deepening of debt markets.

Japan, France, and Sweden

are countries where changes in heavily regulated bank systems were
stimulated, by the development of the bond market under the pressure of
2
rising government budget deficits.
Clearly other factors were also at work. Structural change in the
banking system has contributed to a changing role for banks. And
technological change, particularly in the advance of computer technology,
altered the environment of financial markets, decreasing costs and
increasing the speed with which financial transactions were transmitted
between markets.




Most of the countries we examined responded to these changes by
shifting monetary policy away from the use of credit constraints toward a
greater reliance on market-determined interest rates. With the rapid
development of financial markets, monetary aggregates became less stable
and less subject to control by the monetary authorities. In general
governments began to deemphasize monetary aggregates, by changing the
method of targeting while paying more attention to short-term interest
rates and their impact on the economy. For example, as the velocity of
the broad aggregate £M3 became more unstable, the response of the U.K.
government was to shift from targeting £M3 towards a narrower aggregate,
MO. In the second half of the 1980s, British monetary policy became more
interest-rate oriented, and at times, was used to pursue exchange rate
targets as well.
In contrast, when German CBM became unstable and its targets were
overshot in part as a result of volatility in currency flows, the
Bundesbank shifted its focus to M3, an aggregate that puts less weight on
currency. The Bundesbank was one of the first central banks to adopt
(both informally and formally) monetary targets after the breakdown of
the Bretton Woods system. Monetary policy relied heavily on interest

2. For a comprehensive discussion of these issues, see Germany and
Morton (1985).
- 4-

Kahn and Kole

rates, supplemented by window guidance.

Despite the uncertainties

surrounding the unification of western and eastern Germany, the
Bundesbank continues to put more weight on targeting broad money than do
central banks in most other major industrialized countries.
Although the last two decades saw a gradual move towards more
active use of short-term interest rates as instruments of monetary
policy, evidence is generally inconclusive as to whether foreign
authorities have acquired greater influence over long-term interest
rates.

Some have argued that financial innovation implies that short-

term and long-term rates move more closely together than previously.
However, as both government and corporate bond markets deepened, the
ability of monetary authorities to influence long-term rates may have
been reduced.

Because long-term rates are more important in the

determination of real economic variables, the nature of the relationship
between monetary policy and the term structure is crucial.
Financial liberalization has had conflicting effects on the
interest sensitivity of aggregate demand.

In Europe and Japan, there

have been significant structural changes that reduced liquidity




constraints on consumers and rationing of loans by banks.

To the extent

that there is now less disintermediation from the banking system
associated with a monetary contraction, interest rate changes will be
measured to have a smaller impact on economic activity than previously.
On the other hand, as discussed further below, individuals now carry more
debt and their cash flow is more vulnerable to interest rate changes.

In

addition, households now have more assets whose return can be squeezed.
In fact, one of the general conclusions that we suggest below is that
wealth channels for the transmission of monetary policy now matter more.
Greater international openness and integration.

Liberalization of

domestic financial markets had its counterpart in greater economic
integration of trade and financial relationships across countries.

Part

of this transformation reflected the continued liberalization of trade on
a multilateral basis.

Table 1 illustrates that, throughout the

industrial countries, merchandise trade as a share of GDP/GNP rose
between 1971-75 and 1987-91.

Indexed by this measure, Germany, the

United Kingdom, Canada and France appear to have been quite open by the
late 1980s.

In contrast, the external sector in Japan remains a smaller

- 5 -




Kahn and Kole

part of aggregate demand than is widely realized. Exchange rate
considerations have figured prominently in Germany in part because of the
relatively large size of the export sector (and-because the Bundesbank
has a statutory mandate to "safeguard the currency").
The greater openness of trade flows reflected a revolution in
technology, which contributed to greater international integration of
financial systems. Greater openness and financial market integration
meant that governments increasingly had to respond to global economic
disturbances. Spillover effects of one country's change in monetary
policy could no longer be ignored. The growth of the tradable goods
sector of these economies supports the common contention that monetary
policy now has a greater impact on economic activity through the exchange
rate channel than previously was the case. The usual argument is that a
monetary-policy induced change in interest rates at home causes nominal
interest differentials to change, leading to movements in nominal
exchange rates. With the growth in the size of tradable goods sectors,
movements in exchange rates will then have greater effects on aggregate
economic activity. Mauskopf (1990) finds some evidence for an increase
in the influence of the exchange rate on the volume of U.S. exports and
imports using the Board's MPS model.
However, there are a number of reasons why we may not be able to
observe this change clearly in the data. First, there is some evidence
that long-term rates in the major industrial economies have moved more
closely together in recent years than in the past. Whether this reflects
greater financial integration or governments' policy responses to similar
economic disturbances, the implication is that an interest rate
innovation in one country may on average be associated with smaller
movements in interest rate differentials and exchange rates than in the
past. Further, changes in the passthrough of exchange rates to import
prices (e.g., Hooper and Mann, 1989) will also affect the impact of
monetary policy monetary policy through this channel.
The large movements in exchange rates of the major foreign
industrial countries (especially against the dollar) during the 1980s
also affected the monetary policy reaction function in some countries.
The most obvious example was the creation of the EMS in 1979 and the
gradual convergence of interest rates in member countries during the

- 6-

Kahn and Kole

1980s, which for all practical purposes devoted the conduct of monetary
policy in France, Belgium, and the Netherlands to the goal of exchange
rate stability against the mark.

The progressive decline in the

frequency of realignments in the EMS along with the move towards fuller
EMU gradually enhanced the ERM's credibility and to some extent has made
3
monetary targeting in separate nations obsolete. Germany, arguably the
most zealous G7 nation in terms of monetary targeting, occasionally
subordinated its monetary objectives to external goals.

For instance,

after meeting its CBM targets each year between 1979 and 1985, the
Bundesbank significantly overshot its target in 1986, when the mark
appreciated 35 percent against the dollar and the current account reached
4
a record level.
Japan also tolerated faster money growth at times due
to exchange rate considerations.
have been inappropriate.

Occasionally, exchange rate targets

For example, it is by now well recognized that

the British experiment with shadowing the ERM at a rate of DM3 between
1987 and 1988 led to overly expansionary monetary growth, a boom, and
excessive inflation.
The rise in consumer and business indebtedness.

During the 1980s,

the deregulation of financial markets and the provision of new financial
instruments in many industrialized countries allowed both households and
firms to accumulate more debt than ever before.

The removal of

liquidity constraints for less well-off households and small firms and
the wider selection of financing methods available to the private sector
changed monetary transmission mechanisms in the following ways.

Lower

liquidity constraints allowed more consumers to smooth consumption and
enabled them to react more to changes in permanent rather than current

3. It is interesting to note however that most of the ERM countries
continue to target some aggregate, despite the essentially fixed nature
of their exchange rates. One justification for the seeming inconsistency
between exchange rate and monetary targets is that the anchor of the
system, once the mark, has become less clear as Germany has had to adjust
to the inflationary pressures of unification.




4. Note that domestic considerations were not in conflict with a
monetary expansion, as consumer price inflation was negative for the
first time since the 1950s.
5. In some countries a change in tax incentives contributed to the
buildup of debt as well.

- 7 -

Kahn and Kole

income.
Deregulation lowered the cost of financial intermediation,
particularly for households, causing a shift towards borrowing. The
larger proportion of households in debt meant that the private sector's
sensitivity to interest rates might increase, thereby making monetary
policy more potent. For example, empirical work by Dicks (1990) found
that the interest elasticity of U.K. consumption increased during the
1980s, increasing the leverage of monetary policy.




Table 2 shows that the increase in debt of the household sector
since the 1970s has been most dramatic in Japan and the United Kingdom.
In both these countries, it could be argued that the conjunction of easy
monetary policy and the rapid accumulation of mortgage debt fueled a real
estate boom in the latter part of the 1980s. While land and equity
prices were rising, the increase in consumers' liabilities seemed to be
matched by the increase in their gross wealth. However, when tighter
monetary policy finally burst the asset price bubbles in late 1989, both
firms and households were left with an excessive stock of outstanding
debt. After two years of recession, U.K. spending is still quite weak as
firms and households continue to adjust their balance sheets. Debt
overhang may have made U.K. consumers less responsive to monetary easing
than previously was the case. For similar reasons growth in Japanese
spending is now expected to be sluggish for some time. Note that these
episodes have meant that asset prices, including those of tangible assets
such as land, may be playing more of a role in the monetary transmission
mechanism than previously.
The increase in household indebtedness has been more moderate in
continental European countries such as Germany and France. The
comparatively lower increase in the debt burden may mean that interest
rate sensitivity has not changed as much in these countries. Also these
countries did not experience the rapid asset price inflation seen in
Japan and the United Kingdom in the late 1980s, so monetary policy may
have worked less through the asset price channel.

6. Bayoumi (1990,199?) empirically documented this in his work on
financial deregulation and consumption in the United Kingdom. He
estimated that during the 1980s, the U.K. personal savings rate fell by
2-1/4 percent as a result of financial deregulation.
- 8-

Kahn and Kole

The cycle of debt growth and asset price inflation may have
changed monetary reaction functions as well.

In both the United Kingdom

and Japan, the asset price inflation preceded periods of general consumer
price inflation, giving monetary authorities an early warning that
7
tightening would be a good idea.
The responsiveness of firms to interest rates also was altered by
the rise in business indebtedness, which again has been greatest in Japan
and the United Kingdom as shown in the middle panel of table 2.
Financial market deregulation expanded the access to credit and lowered
the cost of borrowing for non-financial firms. With the development of
corporate bond markets and innovations such as securitization, businesses
have resorted increasingly to non-bank financing. However, as discussed
with respect to the United States (in Bernanke and Blinder, for example)
there are good reasons to believe that bank credit remains an important
channel for monetary policy.
The switch to a less active stabilization policy in a lower
inflation environment. In the 1980s, there appears to have been
increased recognition on the part of foreign industrial governments that
there are limits to what can be achieved through countercyclical fiscal
policy and that monetary policy should be devoted more seriously to
achieving price stability. As shown in the bottom panel of table 2,
public sector debt as a share of GDP leveled off or declined in a number
of countries between 1985 and 1990, although it probably has risen
recently. Further, to some extent the drop in inflation and inflation
expectations throughout the industrial world during the course of the
1980s can be credited to effective monetary policy. Lower inflation
should have contributed to a decline in the velocity of various monetary
aggregates and increased economic efficiency.




A CONVENTIONAL KEYNESIAN MODEL WITH WEALTH AND EXCHANGE RATE DYNAMICS
In light of our hypothesis that asset prices and wealth have become more
important in the transmission of monetary policy, this section develops a

7. The Bank of Japan explicitly tightened policy in 1989 to quell asset
price inflation. It appears that the U.K. government did not adjust
monetary policy quickly enough in response to the asset price signal,
leading to the boom and subsequent bust that occurred.
- 9-

Kahn and Kole

model of the open economy in which wealth plays a key role along with
income in the determination of aggregate demand and prices.

The model is

Keynesian; the price level is sticky and output is demand determined in
the short run.

Wealth, modelled for simplicity as the value of the stock

market, affects output directly through its effect on aggregate demand




and is also a factor in determining money demand.

Output, through both

profitability and interest rate channels, in turn is a key determinant of
asset prices.

The model is a variant on Gavin (1989) who built on the

closed economy analysis of Blanchard (1981).
Table 3 presents the equations of the model.

The model is

formulated so that all parameters in the structural equations are
positive and all variables other than interest rates are in logs.

In

equations 1-2, real aggregate demand, D, depends on the real value of the
stock market, q, real output, Y, the real exchange rate, E, and fiscal
policy, G.

Output then, adjusts to the difference between aggregate

demand and output, where the coefficient d is a measure of the speed of
g
adjustment of Y to D.
Equation 3 can be interpreted as a liquidity
formulation for money demand, solved for the nominal interest rate i,
where the money supply is exogenously given.
wealth affect money demand.

Money, income, prices, and

Alternatively, this equation could be

thought of as a policy reaction function of the monetary authority.

In

this case, a positive coefficient on wealth would suggest that monetary
authorities respond to asset price inflation by tightening policy.

For

simplicity, equation 4 defines the real interest rate as the nominal rate
less the rate of change of home goods' prices; the real exchange rate
does not enter.

We also have assumed that perfect foresight prevails.

The dynamics of the price process are given in equations 5. The
price adjustment parameter is simplistic, and abstracts from Phillips
curve or staggered wage contracting considerations, but still allows for
significant interactions in the results described below.

The steady-

state price level (equation 6) equilibrates the money market at the
steady-state interest rate and the equilibrium level of output.
Asset market equilibrium is given in equations 7-9.

In equation

7, the expected real return on a share of the stock market consists of

8. One could allow d to equal 1.
equilibrium value.

in that case, Y jumps to its new

- 10 -

Kahn and Kole

both capital gains, q/q, and the rate of profit, Z/q.
equal the real return on domestic bonds.
parity condition, where r
exogenously given.

This return must

Equation 8 is the open interest

is the foreign real interest rate and is

Equation 9 specifies real profits as a function of

real output and equation 10 gives the relationship between the long-run
real interest rate and the short-run real interest rate.
Equations 1-9 can be solved for the steady state values of Y, q,
E, and P, and the two negative roots that describe the adjustment of the
economy after a shock.

Chart 2 plots possible dynamic paths of these

variables after a monetary shock occurs.

Consider an unanticipated

monetary expansion at time 0, with the economy initially at a steady
state.

The new steady state is identical to the old, except that the

price level is increased by the amount of the monetary injection.

As

shown in chart 2, the simple price adjustment mechanism in this model
ensures that prices adjust monotonically to their new equilibrium level.
The dynamic responses of output, interest rates, and the real
exchange rate are more complicated.




Output is "humpbacked", rising from

the initial.steady state to a point and then falling toward the unchanged
steady state.

Stock market wealth jumps up at the time of the monetary

expansion to equilibrate expected stock market returns with returns in
the bond markets, then gradually falls as interest rates and income
adjust to the shock.

In the left middle panel, the normal pattern of

short- and long-term interest rates following the monetary expansion is
shown.

Interest rates initially fall following the monetary shock, then

rise as output expands.

Rates eventually rise above their baseline level

before beginning to fall again as output returns to baseline.

However,

the introduction of stock market prices into the model creates the
possibility that long-term interest rates could rise following the
monetary expansion.

Short-term interest rates always fall on impact of

the monetary impulse, but rise as activity responds.

If the monetary

expansion has a strong effect on the stock market, and the increase in
wealth in turn has a strong effect on activity, short term rates may
respond quickly and soon rise above baseline.

In this case, the long-

term interest rate as the weighted average of current and future shortterm rates might rise immediately, implying a stronger movement in the
term structure.

- 11 -

Kahn and Kole

Chart 2 also shows possible paths for the real exchange rate. The
left and center bottom panels show the usual depreciation of the real
exchange rate following a monetary expansion.

As discussed by Gavin, the

possibility exists in this model for the exchange rate to exhibit
perverse behavior and appreciate following the monetary expansion.

The

logic is similar to the reason that long-term rates can rise following




the innovation to money:

If the stock market response is strong enough,

the actual and expected increase in interest rates causes the exchange
rate to appreciate.

This model highlights the importance of wealth in

the monetary transmission channel.
MONEY DEMAND IN JAPAN, GERMANY, AND THE UNITED KINGDOM
Although wealth plays a crucial role in the monetary transmission
process, most of the vast literature on money demand in industrialized
countries excludes this variable from the analysis.

Especially when one

considers the demand for broad aggregates, that are less oriented towards
making transactions and have more of a role in portfolio management, the
omission of some measure of wealth is likely to seriously bias one's
results. M2+CDs, the aggregate most closely monitored by the Bank of
Japan, and M3, the aggregate targeted by the Bundesbank, are both likely
to be determined primarily by portfolio flows rather than fluctuations in
the transactions demand for money. The same is true of British M4, which
we chose to analyze here because of its closeness in definition to £M3,
9
the aggregate targeted over most of the period.
Corker (1989) noted the importance of including Japanese wealth in
order to achieve a stable equation for the evolution of real M2+CDs and
Hall, Henry, and Wilcox (1992) found personal sector wealth to be
significant in the long-run determination of U.K. M4. In the spirit of
these studies, we estimated money demand functions for broad monetary
aggregates that allow wealth to play a role, both in short-run
fluctuations of money demand and in a long-run (error-correction
mechanism) context.

9. We chose not to model British MO, the aggregate targeted by the
British government for the past few years. This aggregate, which
includes notes, coin,, and banks' operational deposits with the Bank of
England, is extremely narrow and more of a coincident indicator of
economic activity than a indicator of the government's monetary policy.
- 12 -

Kahn and Kole

As we mentioned above, Japanese financial markets were deregulated
later than those in Germany and the United Kingdom, and the proce'ss of
removing interest rate regulations is still ongoing.

As a consequence,

among the three countries considered, Japan is the most likely candidate
for instability in the demand for money.

This is indeed what we find.

In part, the problem in estimating a stable money demand function for the
entire period between 1973 and 1991 is that different interest rates were
probably relevant at different periods.
the relevant interest rates:

The top panel of Chart 3 shows

the Gensaki rate was the most market-

related rate in the 1970s whereas the own rate was very sticky and
unrelated to market interest rates.




Both the own rate and the postal

savings rate became more, important over the course of the sample, while
the Gensaki rate became progressively less significant.

The chart also

shows that the opportunity cost of M2+CDs, the gap between either outside
rate and the own rate continued to narrow until quite recently.
Using quarterly data, we regressed the change in the log of M2+CDs
deflated by the log GNP deflator, A(m-p), the change in log GNP (Ay), the
rate of GNP price inflation (Ap), the growth in the log of real wealth,
A(w-p), and the change in various interest rate measures: an own rate,
i , the postal savings rates i

, the Gensaki rate, and the long rate, i

as well as the error correction terms.

We started with a specification

that included 4 lags of each of the first-differenced variables, and
gradually eliminated those variables that were not significant at the 10
percent level.

The wealth variable used was an updated version of that

used by Corker (1989), the total stock of assets held by the personal and
corporate sector.

We originally started with error correction

mechanisms similar to Corker's, (m-w)

-and (m-p-y)

-, but found we were

almost able to reject the implied restrictions, so we allowed the long
run relationship between real money, real wealth and real output to be
estimated freely.

10. We followed Corker (1989) in the construction of the own rate on
M2+CDs as a weighted average of the interest rate on 3-month CDs and the
average rate on savings deposits.
11. Unfortunately, this measure of wealth only includes financial wealth,
not tangible wealth such as real estate.

- 13 -

Kahn and Kole

The first column of table 4 presents the results.

Note that the

wealth variable, w-p, is quite significant, both in terms of its rate of
change and in its lagged level.

The rate of inflation, Ap, and its first

lag, are strongly significant, and we were able to reject the hypothesis
that the relevant demand is for nominal rather than real money.

The rate

of change of GDP, Ay, and its lags, did not add to the explanatory power
of the equation, but y

., is significant.

Both the postal savings rate,

a rate on assets outside M2+CDs, and the own rate, a weighted-average
rate on assets inside M2+CDs, have estimated coefficients that are
significant and of the correct signs. Assuming price stability, the
long-run money demand function implied by the error correction variables
is:




m-p - .43y +.48(w-p) + 6.28i° - 5.01iPS
We then estimated the equation over two subperiods; 1973-1982 and
1983-1991. We broke the sample after 1982, before financial deregulation
and innovation really started to take off in Japan.12 The Fisher test
indicates that one can reject the hypothesis that the demand for M2+CDs
remained stable between these two periods at the 1 percent level of
significance. Note that the dynamics have changed considerably between
the two periods, and that during the first period this specification
leads the errors to be serially correlated. There is some evidence that
the lags were longer in the first period than in the second, perhaps
indicating that money holders have become quicker to adjust their
balances in response to changes in wealth, interest rates, and other
explanatory variables.
Turning to the estimated coefficients for the two subperiods,
several points are worth noting. First, both the short-run and long-run
interest rate elasticities are estimated with the wrong sign during the

12. Kasmin and Rodriques (1992) chose to break their sample in 1984.
13. We were able to find a stable money demand function between the 2
periods when we left many lags in the equation. These lags were more
significant during the first period. Their addition however, did not
improve the fit of the equation in the overall period, so they were
discarded.
- 14 -

Kahn and Kole

earlier period; with the change in the own rate significant, whereas in
the later period all the estimated interest rate elasticities have the
correct sign.

Second, although the lagged wealth variable is significant

in both periods, the long-run coefficients indicate that the proportion
of an increase in wealth allocated to M2+CDs fell from 60 percent in the
early period to about 50 percent in the later period, which seems
consistent with the wider diversity of assets available outside the
aggregate.

Finally the higher significance of the lagged level of income

in the earlier period is evidence that the transactions motive for
holding money has become less important over time.




In contrast to the Japanese case, due to the lack of financial
deregulation and the consistency of monetary policy, German money demand
before 1990 was, a priori, the most likely to be stable among the
countries we considered.

The middle panel of chart 4 shows the interest

rates relevant for German money demand.

It is immediately apparent that

they move together during the entire sample.

Unlike Japanese savings

instruments, German time deposits have had market-related interest rates
for some time, reducing the incentive for financial innovations such as
those which change the nature of the relationship between Japanese
interest rates during the sample period.
We used the same estimation technique employed for Japanese money
demand to determine the proper lag lengths of the included variables.
Table 5 presents the results for the overall period and two subperiods:
1970-79 and 1980-89.
the EMS.

The split in the sample is close to the creation of

We ended the sample in the fourth quarter of 1989 because of

the possibility that the fall of the Berlin wall induced instability in
the demand for M3.
The rate on public bonds was found to be more important than
short-term rates as an opportunity cost of M3.

This may be due to the

fact that M3 contains assets of maturities of up to 4 years, so that the
marginal investor might substitute into a longer-term asset such as
public bonds.

Although some studies of German money demand (see for

example, von Hagen and Neumann (1988)) have included foreign interest
rates and/or the exchange rate, (spot and forward), Frowen and Schlomann
(1992) found them not to be very important for M3.

The rate on public

bonds is more important in the 1980s than it is in the 1970s, as is the

- 15 -

Kahn and Kole

own rate. Note that the coefficients on own rate and the public bonds
rate are of equal and opposite sign in the later sample, indicating that
an opportunity cost specification using the interest rate differential
may be best for the period of the 1980s.
The wealth variable for Germany was constructed from flow-of-funds
data of the household sector. Since data on households' net assets are
available only on a biannual basis, we interpolated resulting in a smooth
version of quarterly wealth.
(This measurement error associated with
interpolation and constructing stock data from flow data may account for
the lower significance of wealth in German money demand, at least in
comparison to Japan and the United Kingdom.) However, its significance
increases between the two periods, indicating that portfolio management
plays an increasing role in the determination of the demand for M3. The
long-run solution implied by the error correction coefficients is:
m-p - .19y + .54(w-p) + 3.75i° - 5.41iPB
Finally, it is evident that the transactions demand for money has
become less important as the lagged level of real GDP has become
progressively less significant. The fit for the earlier period improves
slightly if one adds in the change in GDP. None of these changes are
statistically significant, however, as is indicated by the fact that the
Fisher test does not reject the hypothesis that the two subsamples can be
pooled. As we expected, and as others have found, it does not take much
to arrive at a stable demand for German money in the pre-unification
period.




It is difficult to say much about how the monetary transmission
mechanisms have changed since the unification of east and west Germany.
The lack of a sufficiently long data set post-unification rules out
econometric analysis of the period. Nonetheless, a few preliminary
judgments can be ventured. First, German money demand has become less
certain following the 1990 currency conversion of Osmarks for Deutchmarks
that added eastern German demand, and at times movements in interest
rates and money growth have sent conflicting signals regarding the stance

- 16 -




Kahn and Kole

of policy.

In this environment, the Bundesbank has recognized that

monetary policy changes will be at best a blunt instrument for
influencing prices and activity.
Second, a case can be made that aggregate demand in Germany
currently is less interest sensitive than before unification.
Specifically, a significant portion of German domestic demand is
associated with the surge in investment, and specifically construction,
in the eastern portion of Germany.

Much of this investment appears to be

limited by capacity constraints, suggesting that changes in interest
rates may have a limited effect on these flows in the near term.

In

addition, a substantial portion of this investment is publicly
subsidized, and thus less responsive to monetary policy.

Further, to the

extent that investment in eastern Germany is subject to substantial
systemic risks associated with the region's transition to a market
economy, nominal interest rates will be a smaller part of the total
perceived cost of investing than in western Germany.
Another argument that has been made is that with the completion of
the internal market in the European community, barriers to entry into
German banking markets have been reduced.

New entrants, actual or

potential, are expected to enhance competition in these markets and
should lead to more competitive pricing of deposits.

This may lead to

less disintermediation associated with changes in Bundesbank policy,
meaning that interest rate changes may have less effect on activity
through the bank lending channel than previously.
Table 6 presents the results of estimating the demand for real
U.K. M4.

Here we adjusted the M4 data to account for a break in the

fourth quarter of 1981 when the monetary sector replaced the banking
sector and many new institutions such as the Trustee Savings Banks were
for the first time included in the broad monetary aggregates.

We tried a

variety of interest rates including long-term rates, local-authority
rates, and deposit rates, and found that over the entire sample, the rate
on 3-month Treasury bills fit the best.

From the bottom panel of chart

14. Some have argued that strong eastern German money demand suggests a
combination of a high income elasticity and strong income growth due to
large transfer payments from the government.

- 17 -

Kahn and Kole

4, it appears that the opportunity cost of M4 has not varied
substantially over the sample.
No interest rates seemed to matter in the- rate of change form, and
it was difficult to find specifications with the rate of change of GDP
entering the equation significantly and with the correct sign.

The

change in real wealth (and its lags) also had nothing to add to the
explanatory power of the other variables.

Wealth is total wealth of the

personal sector, including tangible assets. We also tried financial
wealth, and it fit, but not as well, especially in the latter period.
The long-run money demand function implied by the error-correction
terms is:
m-p - -.51y + 1.58(w-p) + 5.4i° - 5.84iTB
The estimated long-run elasticity of real M4 demand with respect to GDP
is about -1/2. The equation can be rewritten in the following way:
m.p-y

-

1.58(w-p-y) + .07y + 5.4i° - 5.84iTB

The inverse velocity of M4 is positively related to the wealth/GDP ratio
and negatively related to the opportunity cost of money.
We chose as a sample breakpoint the end of exchange controls, and
the beginning of the Thatcher government's reorientation of monetary
policy from restrictive lending towards more market and monetarist goals.
The equation passes the Fisher test for stability between the two periods
at the 5 percent level of significance. It seems intuitively plausible
that demand for U.K. M4 would be less stable than that for German M3, but
more stable than the demand for Japanese M2+CDs. However, the dynamics
seem to have changed between the two periods, as do the signs and
significance of some of the coefficients.
The estimated coefficient on the growth of GDP (two quarters
earlier) switches from positive to negative between the two periods.
It is interesting to note that the error correction mechanism implies the

15. Note that this result does not disappear if current and once-lagged
GDP are included in the equation. The sum of the estimated coefficients
still adds up to a negative number for the second period.




- 18 -




Kahn and Kole

opposite change in sign between the two periods.

The long-run elasticity

of real M4 demand with respect to real GDP changes from an insignificant
negative number to a significant positive 1.5. . (Note that the negative
long run income elasticity estimated for the full period seems to be
coming from the 1970s rather than the 1980s.)

The implication is that

between the two periods the relationship between transitory changes in
income and real M4 changed, but income became more important as a longrun determinant of money demand.
As in Japan, the long-run wealth elasticity declines between the
two periods, from 1.4 in the 1970s subperiod to .5 in the latter period.
It appears that in the earlier period, investors were moving into M4 on
balance, while in the later period they were diversifying into assets
outside M4.
The estimated coefficients for 3-month Treasury bill rate and the
own rate decline in significance between the first and second period.
This could be a result of other interest rates being more relevant.
There is some indication that long-term rates seem to become more
important between the two periods.

This may be an indication that

investors became more sophisticated as financial markets deepened.
To summarize our results thus far, we find evidence that wealth is
a significant determinant of money demand in Japan, Germany, and the
United Kingdom, and may have become more important than in the past.

In

all three countries, there seems to have been a shift from holding money
for transactions purposes to holding it as part of a portfolio between
the 1970s and 1980s.

We find that the demand for broad money remained

fairly stable in both Germany and the United Kingdom during the past two
decades.

However, in Japan, where financial deregulation was more

gradual and is still ongoing, we find more instability.
THE RELATIONSHIP BETWEEN MONETARY AND REAL VARIABLES.
What is the relationship between monetary variables, be they monetary
aggregates, interest rates, or exchange rates, and the real economy?
This is an age-old question that has been analyzed most extensively for
the United States without resulting in anything close to a consensus.
surveyed the evidence for Japan, Germany, and the United Kingdom.

Some

of the changes in the transmission mechanism discussed above could be

- 19 -

We

Kahn and Kole

expected to change the reduced form relationship between monetary
variables and various activity variables.

For example, if consumers have

become more interest sensitive because of lower liquidity constraints,
interest rates may be more related to real spending.




Estimation of the structural model laid out in section 3 is beyond
the scope of this paper, but in this section we take a first pass at
analyzing the relationship between monetary and real variables by looking
at reduced-form results.

Specifically, we ran a battery of regressions

to assess the predictive power of monetary variables in the determination
of real economic activity, over the 1973-91 period and two subperiods for
Japan, Germany, and the United Kingdom.'

For the activity variables for

which monthly data are available, industrial production (IP), retail
sales (RS), the trade balance (TB), and the unemployment rate (U), we
included six lags of the relevant variable, a trend, six lags of various
price indices to proxy for supply shocks (and to stack the deck against
us), and six lags of the three monetary indicators shown as column
heading in tables 7-9.
The monetary indicators chosen were the major aggregates (in most
cases the targeted ones), the 3-month -interbank rate, and the term spread
between a long-term rate and the 3-month rate.

The latter variable has

been shown by Stock and Watson (1989) and others to have remarkable
explanatory power for U.S. output.

We also included the exchange rate

as a monetary indicator, when the activity variable in question was an
external balance.
Quarterly regressions were run with the following variables:

real

GNP or GDP, real consumption expenditures (C), real gross fixed
investment (I), and real net exports (NX). These regressions included
four lags of the activity variable, four lags of the monetary variable
(on an average quarterly basis), and four lags of the GNP/GDP deflator,
the CPI, the PPI, and the world commodity price index, respectively.
Because we were primarily interested in the cumulative impact of a
given money or interest rate innovation, we tested whether the addition
of the lags of the relevant monetary indicator variable summed to a

16. Our equations for industrial output are similar to those estimated
for the United States by Stock and Watson (1989).

- 20 -




Kahn and Kole

quantity significantly different from zero.

When we find the sum of

the coefficients on a monetary variable to be significant, we can say
that that variable helps predict the real variable.

Table 7 presents the

marginal significance levels for rejecting the hypothesis that the
estimated coefficients on monetary indicator variables (in the columns)
sum to zero in regressions where the activity variables (in the rows) are
the dependent variables.

A low marginal significance level means that it

is easy to reject the null hypothesis that the 6 lags of a monetary
variable sum to zero, i.e. low significance levels are associated with
monetary variables adding to the predictability of real variables.
For Japan, it is evident that of the monetary variables
considered, M2+CDs has the strongest relationship with real activity
variables over the entire period in question.

In addition, the

additional explanatory power of M2+CDs (as well as bank credit) rose
between the two periods, at least for retail sales, consumption,
industrial production, and GNP.

The 3-month rate is only a good

additional predictor for industrial production and real GNP for the whole
period, although it seems to be an important variable for investment in
both of the subperiods, and for retail sales and net exports in the later
period.

The term spread is a good predictor of GNP, consumption and the

trade balance, but only for investment in the earlier period.

The fact

that it increases in significance for several activity variables in the
later period could mean that the sensitivity of activity variables
(excluding investment) may have increased between the periods.

The

amount of financial deregulation and innovation in Japan during the past
decade may have made the term structure relevant for the first time.

As

financial markets deepen, interest rates and term spreads may transmit a
clearer signal of both monetary and real disturbances.
For Germany, the 3-month rate is the best indicator for the entire
period, although M3 is a good indicator of industrial production,
unemployment, GDP, and investment.

The term spread is a good indicator

for the trade balance, retail sales, GDP, and consumption.

In the full

17. Note that this is a weaker test than Granger causality in that it
does not require that each individual coefficient equal zero, but only
that they sum to zero. However, we wanted to rule out cases where
monetary variables entered significantly, but with equal and alternating
signs.

- 21 -

Kahn and Kole

sample, the regressions yield coefficients with the correct signs. One
interesting finding is that money is negatively (and 3-month interest
rates are positively) related to real net exports and the nominal trade
balance.




Not surprisingly, real net exports are also positively related

to the DM/$ exchange rate especially in the case where the regressions
are specified in terms of rates of change.
Comparing the two subperiods, it looks as if M3 and the exchange
rate have somewhat more significance for real net exports in the later
period, while the 3-month rate and the term structure have more
significance for several real variables in the earlier period.

It

appears that interest rates were more important in the earlier period.
Retail sales were much more related to monetary variables, including bank
credit, in the earlier period indicating that they may have played more
of a role in the monetary transmission mechanism.

Overall, there are

less differences to point to than in the case of Japan.
In the United Kingdom, interest rates are more strongly related to
real activity variables than the two monetary aggregates we considered.
The best overall indicator for the entire period seems to be the 3-month
rate, although the term-spread is quite significant as a predictor of IP,
GDP, investment and net exports.

There is little evidence that

consumption became more interest-sensitive over the two periods, and
unfortunately, we only have retail sales data on the latter half.

There

is evidence that industrial production became more related to the 3-month
rate, the term spread, and M4 while becoming less related to MO.

The

short rate and the term spread also have more impact on GDP and
investment in the 1980s, i.e. the marginal significance levels decline
between the two periods.
We also repeat these tests with first-differenced variables as
well.

These results were not as strong (not surprisingly), but there

were a few points worth noting.

In the case of Japan, the growth of

M2+CDs is the best additional predictor of real economic growth,
especially in the later period.

There is again evidence that the term

spread increased in importance between the two periods.

The rate of

change of the dollar exchange rate (the variable most likely to be nonstationary) has considerable significance for the rate of change of net
exports for the full period and the early period, but not for the latter.

- 22 -




Kahn and Kole

In contrast, the rate of DM depreciation against the dollar has more
significance for German real net exports in the later period than in the
earlier one, even though it its marginal significance level for the full
sample is .0002.

The 3-month rates still stands out as the best

additional predictor for German economic activity.

Finally, the result

for the U.K. rate of change regressions were the worst of the three
countries.

The change in the dollar exchange rate did not add

explanatory power in regressions involving the the trade balance or net
exports, but both the change in the 3-month rate and the term spread were
significant.
Next we considered the role of bank credit for each of the
countries and the monthly regressions in tables 7-9.

To summarize, bank

credit did not play much of a role in predicting German activity
variables, except retail sales.

The marginal significance of bank credit

was .0001 for retail sales in the earlier period.

Bank credit appeared

to matter for all the activity variables in Japan, primarily in the later
period.

For the United Kingdom, available data only allowed us to

conduct the tests on the later period, and we found that bank credit did
not have significant explanatory power for real activity in the 1980s.
To get at the issue of the change in international linkages, we
decided to test whether a foreign monetary policy variable could add
explanatory power to domestic monetary variables in Japan, Germany, or
the United Kingdom.

We reran all the regressions included in tables 7-9

adding 6 lags of the U.S. term spread to the monthly equations and 4 lags
of the U.S. term spread to the quarterly Equations and tested whether the
U.S. term spread had additional predictive power for foreign real
economic variables once their own monetary variables were accounted for.
For Japan, we found that the U.S. term structure was most
significant as an additional explanatory variable for retail sales, the
trade balance, the rate of change in unemployment, and GDP.

In Germany,

the U.S. term spread had additional explanatory power for GDP,
consumption, net exports, and retail sales.

The U.S. term spread is most

significant in the U.K. regressions involving the trade balance, net
exports, GDP, and investment.
One result we found across countries was that the U.S. term spread
seemed to have more significance in the earlier period than in the later

- 23 -

Kahn and Kole

period.




This could be evidence that international interest rates were

less integrated in the earlier period, so that the U.S. term structure
conveyed information lacking in domestic monetary variables.

Another

finding of note was that for most of the regressions involving net
exports or the trade balance, the significance of the U.S. term spread
substantially diminished when the exchange rate was chosen as the
monetary variable, indicating that the U.S. term spread may be proxying
for the exchange rate.

In summary, the U.S. term spread had some

additional explanatory power for foreign real variables, but nowhere near
the power is seems to have for the U.S. real economy.
CONCLUDING REMARKS
This paper attempts to characterize the ways in which monetary
transmission mechanisms abroad have changed in the past decades. We find
that despite a major transformation in global financial markets, the
demand for German M3 (pre-unification) and British M4 seems to be stable
during the past two decades. Only in Japan, where financial deregulation
was later and is still ongoing, do we fail to find a stable demand for
broad money. We find some evidence that financial deregulation and
innovation may have changed the interest rates relevant for money demand,
and that the elasticity of money demand with respect to the opportunity
cost of holding money has increased. There also seems to have been a
shift from holding money for transactions purposes in the 1970s to
holding it for portfolio motives in the 1980s.
We find that wealth plays an important role in the determination
of money demand in each of the countries we considered. While this paper
does not take the further step of empirically measuring the importance of
wealth channels in the transmission of monetary policy to the real
economy, we believe that more work on this area is crucial. Furthermore,
our analysis suggests that a greater role for asset prices as indicators,
instruments, or targets of central bank operating procedures.
The reduced form nature of our tests do not allow us to draw many
conclusions about the transmission of monetary shocks to the real
economy. However, we do find evidence that corroborates the view that
the interest rate sensitivity of spending (especially that of
consumption) has increased in the past decade. This is not surprising

- 24 -

Kahn and Kole

given the wider availability of consumer and mortgage credit for
households and the development of new financing methods for firms*.




Further investigation into the impact of the debt buildup of the late
1980s on the monetary transmission mechanisms in Japan and the United
Kingdom is clearly warranted.
We also find some evidence that the exchange rate is an important
channel for monetary policy.

In addition, the term structure of interest

rates seems to be gaining significance as a predictor of real economic
activity in Japan and the United Kingdom.

However, the term spread is

not found to be nearly as significant in Japan and Germany as it has been
found to be as a predictor of U.S. real output, which raises the question
of why such a difference has emerged.
Short of estimating a structural model across these foreign
countries, there are several extensions to this paper that would be
worthwhile.

Estimating monetary reaction functions and attempting to

pinpoint how they have changed over the past several years could give us
valuable information on how money supply processes have evolved.

- 25 -

Kahn and Kole

APPENDIX:

Major Events or Policy Changes Affecting the Monetary
Transmission Mechanism

01
1973
The Bretton Woods system of fixed exchange rates collapsed.
1979
The exchange rate mechanism of the EMS began operation in March.
1985
G7 Finance ministers met at the Plaza Hotel in New York in September
and agreed on the Plaza Accord, that the dollar (which peaked in
February 1985) was fundamentally overvalued and that a further fall in
its value was warranted.
1987
G6 Finance ministers met at the Louvre in February and decided to
manage major exchange rates within unannounced target zones.
1989
At the Madrid summit, EC governments agreed to start Stage One of
European Economic and Monetary Union (EMU) on July 1, 1990.
1991
At the Maastricht summit, EC governments agree to start Stage Two of
EMU in 1994, and to begin full monetary union sometime between January
1997 and the end of 1999.
Japan




1974
Following the oil price shock, inflation rose sharply and the
government began to run substantial deficits that contributed to
growth in the government securities (Gensaki) market. Also, interest
rates on foreign currency deposits were liberalized.
1975
Bank of Japan announced the "Money supply policy" in July
acknowledging a shift in emphasis to focus on M2.
1978
Bank of Japan began to present a public forecast of the money supply.
At the beginning of each quarter, a forecast (not an official target)

- 26 -




Kahn and Kole

for the year-on-year growth rate of the monetary aggregate (initially
M2) during the quarter was announced.
The government began to issue, by public tender, 2-4 year medium-term
bonds.
1979
Bank of Japan shifted to announcing target for M2+CDs.
Banks were allowed to issue Certificates of Deposit (CDs) worth at
least -P500 million.
1980
"Chuki-Kokusai" funds (similar to money-market mutual funds) were
introduced by securities firms.
"Foreign Exchange and Foreign Trade Control Law" led to significant
relaxation of capital controls.
1985
Continued financial market reform occurred such as the introduction of
money market certificates (MMC) for banks and credit associations,
with minimum denominations (-50 million) and maturity 1-6 months.
Minimum denomination and maturity of CDs were shortened, and the
ceiling on issue size was raised. Interest rates on deposits of -1
billion or more with maturity of 3-24 months were deregulated.
1986
Interest rates on deposits of -300 million or more were decontrolled.
Further liberalization on ceilings, size, and maturity of CD issues
took place.
Germany
1972
The annual report of the Council of Economic Experts advocated the
control of the money stock to combat inflation.
1973
The Bundesbank discarded its previous monetary indicator, "free liquid
reserves" and replaced it with "central bank money" (CBM). In order
to maximize control over CBM, it reduced to "practically zero" the
formerly generous reserves that allowed banks easy access to CBM. The
mark began to float.

- 27 -




Kahn and Kole
1974
The Bundesbank announced a target for CBM for the first time in
December. The growth of CBM was to be held to 8 percent between
December 1974 and December 1975.
1978
As a result of a large appreciation of the mark and a substantial
overshooting of its 8 percent CBM target for 1978, the Bundesbank
announced that it would adopt a target range of 6-9 percent between
the fourth quarters of 1978 and 1979.
1981
The Bundesbank made Lombard credit available only under a Special
Lombard Facility at a cost higher than the ordinary Lombard rate.
1985
In January, the Bundesbank raised the Lombard rate to a level that
applied in its temporary security operations and began to use market
operations to supply reserves more liberally.
1986
Reserve requirements were restructured to ensure that they covered
liabilities in the form of bearer securities at up to two years, when
German banks were authorized to issue CDs.
1987
Minimum reserve ratios were increased for the first time since 1979,
to offset the effect on bank reserves of large Bundesbank purchases of
foreign exchange.
After the substantially exceeding its CBM target, the Bundesbank
decided to switch to an M3 target in December to decrease the
influence of notes and coins on the aggregate.
1988
A withholding tax on capital incomes to take effect the beginning of
1989 was passed and reportedly caused substantial capital outflows.
1989
The withholding tax was withdrawn.
The Berlin wall fell in November.
1990
Monetary union, including the swaps of Osmarks for DM occurred in July
and was followed by full unification of east and west Germany in
October.
- 28 -

Kahn and Kole

The United Kingdom




1972

Sterling floated in June.
1973
Supplementary Special Deposits scheme (better known as the "corset)
was introduced, under which banks were required to place supplementary
special deposits with the Bank of England if their interest bearing
eligible liabilities grew faster than a specified rate (8 percent in
the first 6 months). This scheme was suspended and reactivated
intermittently throughout the rest of the decade.
1979
The Conservative party came to power in May and essentially ended
restrictive guidance on building society lending. In its first
budget, the new government announced relaxation of exchange controls,
continuation of the "corset", and a £M3 target range of 7-11 percent
from mid-June 1979. Starting with the 1978-79 fiscal year it was
announced that targets would be rebased every six months.
1980
The green paper on Monetary Control was published in March, and in the
medium term financial strategy (MTFS), the government announced a
gradual reduction in money supply growth. The 1980-81 target for £M3
was to be 7-11 percent, and the range was to decrease by 1 percent
each subsequent fiscal year to 4-8 percent in 1983-84. The "corset"
was discontinued in June.
In November, the government published a note on Methods of Monetary
Control, in which it advocated phasing out the reserve assets ratio,
under which banks had to hold at least 12.5 percent of their deposits
in a specified range of liquid assets, and considering a cash ratio
instead. The Bank of England was to change its money market
intervention to emphasize open market operations rather than discount
window lending, to try and keep very short-term interest rates within
an unpublished band, and to gear daily operations primarily towards
offsetting cash flows between the Bank and the money market.
1981
In the MTFS, a new target range for £M3 was set at 6-10 percent for
the 14 months starting in mid-February 1981. It was announced that
the minimum reserve assets ratio would be abolished, and that all
banks would be required to hold non-operational non-interest bearing
deposits with the Bank of England. The new arrangements for monetary
control took effect in August.
1983
The conservative government of Margaret Thatcher was reelected. Nigel
Lawson, new Chancellor of the Exchequer, reviewed monetary policy and
- 29 -




Kahn and Kole

stated that narrow measures of money were linked more closely to
inflation.
The building societies' cartel collapsed in October.
1984
A target range for MO was set for the first time, at 4-8 percent for
the 14-month period beginning in mid-February 1984. The target range
for £M3 was set at 6-10 percent.
1985
Lawson suspended the 1985-86 £M3 target of 5-9 percent, because it had
been set too low. Overfunding (the policy of issuing more government
debt than required by the budget deficit which had started in 1981)
was dropped and a full-funding policy was adopted in order to ease
persistent shortages in the money market.
1986
The Building Societies Act was passed and mortgage lending guidance
was withdrawn. In October, financial reforms known as the "Big Bang"
eased entry into dealing on the stock exchange and in the gilts
market.
A new target range for £M3 was set at 11-15 percent for the 1986-87
fiscal year, much higher than the 4-8 percent target range set out in
the previous MTFS. The target range for MO was set at 2-6 percent.
1987
Target ranges for broad monetary aggregates were abandoned. The MO
target was set at 2-6 percent. The U.K. authorities begin to "shadow
the ERM", by attempting to keep the value of the pound below DM3.
1988
Sterling was allowed to appreciate above DM3 beginning in March.
the MTFS, M4 replaced £M3 as the main measure of broad money.

In

1989
At the Madrid summit in June, Britain committed itself to becoming a
member of the ERM by the end of Stage One of EMU.
1990
In October, Britain joined the ERM at a central rate of DM2.95 and
with 6 percent bands.

- 30 -

Kahn and Kole

REFERENCES
Bayoumi, Tamim (1990), "Financial Innovation and Consumption in the
United Kingdom", IMF Working Paper WP/90/95, forthcoming in
Review of Economics

and

Statistics.

, Tamim (1992), "Financial Deregulation and Household Saving",
mimeo.
Bernanke, Ben, and Alan Blinder (1990), "The Federal Funds Rate and the
Channels of Monetary Transmission."

National Bureau of

Economic

Research Working Paper no. 3487, October.
Bernanke, Ben and Frederic Mishkin (1992), "Central Bank Behavior and
the Strategy of Monetary Policy:

Observations from Six

Industrial Countries", mimeo.
Blanchard, Olivier J. (1981), "Output, the Stock Market, and Interest
Rates," American Economic Review, vol. 71, no. 1, pp. 132-143,
March.




Boughton, James M. (1992), "International Comparisons of Money Demand,
IMF Working Paper no. WP/92/7.
Brayton, Flint, and Jaime Marquez (1990), "The Behavior of Monetary
Sectors and Monetary Policy," in Financial
Sectors in Open
Economies:
Empirical
Analysis
and Policy Issues, P. Hooper, K.
H. Johnson, D. L. Kohn, D. E. Lindsey, R. D. Porter and R. Tryon
(eds.), Board of Governors of the Federal Reserve System,
Washington, D.C., pp. 365-393.
Bryant, Ralph C. (1990), "Model Representations of Japanese Monetary
Policy," mimeo.
Corker, Robert (1989), "Wealth, Financial Liberalization and the Demand
for Money in Japan", IMF Working Paper WP/89/85.

- 31 -

Kahn and Kole

Dicks, M.J. (1990), "Interest Elasticity of Consumers' Expenditure", in
S. G. B. Henry and K. D. Patterson (eds.) Economic Modelling
the Bank of England,

Chapman and Hall, London,

at

pp. 73-106.

Estrella, Arturo and Gikas A. Hardouvelis (1991), "The Term Structure as




a Predictor of Real Economic Activity", Journal

of Finance,

vol.

46, pp. 555-76.
Frowen, Stephen F. and Heinrich Schlomann (1992), "Financial Innovations
and the Stability of the Demand for Money in Germany since 1974",
in Monetary Policy
Countries:

and Financial

Innovations

in Five

Industrial

The UK, the USA, West Germany, France and Japan,

S.

F. Frowen and D. Kath (eds.), St. Martin's Press, New York, pp.
59-81.
Fukui, Toshihiko (1992), "The Recent Development of the Short-Term Money
Market in Japan and Changes in the Techniques and Procedures of
Monetary Control Used by the Bank of Japan," mimeo
Gavin, Michael K. (1989), "The Stock Market and Exchange Rate Dynamics"
in Journal of International
Money and Finance, vol. 8, pp.181200.
Hess, Gregory D. and Richard D. Porter (19.92), "Comparing Interest-Rate
Spreads and Money Growth as Predictors of Output Growth: Granger
Causality in the Sense Granger Intended", mimeo.
Germany, J. David, and John E. Morton (1985), "Financial Innovation and
Deregulation in Foreign Industrial Countries", Federal
Reserve
Bulletin,
pp. 743-753.
Hall, S. G., S. G. B. Henry and J. B. Wilcox (1990)

"The Long-run

Determination of the UK Monetary Aggregates" in S. G. B. Henry
and K. D. Patterson (eds.) Economic Modelling
England,

Chapman and Hall, London, pp. 127-166.

- 32 -

at the Bank of




Kahn and Kole

Hooper, Peter and Catherine L. Mann (1989), "Exchange Rate Passthrough in
the 1980s: The case of U.S. Imports of Manufactures",
Papers

on Economic Activity,

Brookings

no. 1, 297-329.

Jappelli, Tullio and Marco Pagano (1989), "Consumption and Capital Market
Imperfections:
Economic Review,

An International Comparison" in

American

vol. 79, no. 5, December, pp. 1088-1105.

Kasman, Bruce, and Anthony Rodrigues (1991), "Financial Liberalization
and Monetary Control in Japan."
York Quarterly

Review,

Federal

Reserve

Bank of New

Autumn, pp. 28-60.

Kole, Linda S. and Michael P. Leahy (1991), "The Usefulness of P*
Measures for Japan and Germany", International Finance Discussion
Paper No. 414.
Shigehara, Kimiharu (1992), "Current Monetary Policy Issues in Japan,"
mimeo.
Stock, James H. and Mark W. Watson (1989), "New indexes of coincident and
leading indicators", NBER Macroeconomics

Annual 1989,

pp. 351-93.

Tamura, Tatsuya (1992), "Monetary Control in Japan," in Monetary
Policy
and Financial
Innovations
in Five Industrial
Countries:
The UK,
the USA, West Germany, France and Japan, S. F. Frowen and D. Kath
(eds.), St. Martin's Press, New York, pp. 101-119.
Temperton, Paul (1991), UK Monetary Policy:
St. Martin's Press, New York.

The Challenge

for

the

1990s,

Topping, S. L. and S. L. Bishop (1989), "Breaks in Monetary Series", Bank
of England Discussion Paper, Technical Series No. 23.
Ueda, Kazuo (1990), "Financial Deregulation and the Demand for Money in
Japan" in Financial

Sectors

in Open Economies:

Analysis

Issues,

Edited by P. Hooper, K. H. Johnson,

and Policy

- 33 -

Empirical

Kahn and Kole

D. L. Kohn, D. E. Lindsey, R. D. Porter and R. Tryon, Board of
Governors of the Federal Reserve System, pp. 175-199.
Von Hagen, Jurgen and Manfred J.M. Neumann (1988), "Instability versus
Dynamics: A Study in West German Demand for Money", Journal of
Macroeconomics,
vol. 10, no. 3, pp. 327-49.
West, Kenneth D. (1991), "An Aggregate Demand - Aggregate Supply Analysis




of Japanese Monetary Policy, 1973-1990," National Bureau of
Economic Research Working Paper no. 3823, August.

- 34 -

Kahn and Kole

TABLE 1
Measures of Openness
(Annual Average, in Percent)

Exports
GDP

Exports + Imports
GDP
1971-75

1987-91

197J-75

1987-91

Japan

24

33

10

17

Germany

50

.74

25

38

United Kingdom

47

67

23

32

France

38

51

18

24

Italy

10

12

4

6

Canada

40

58

21

28

United States

14

21

6

10




- 35 -




Kahn and Kole

TABLE 2
Gross Debt as Percent of GNP/GDP

1960-70

1970-80

1985

1990

47.4
57.1
51.4
59.0
50.2
38.1
6.9
55.5

63.2
52.6
73.8
71.7
63.2
47.1
9.8+
66.9

Households
Japan
Germany
United Kingdom
United States
Canada
France
Italy
Sweden

20.3
34.7
31.0*
48.0
43.9
34.0*
n.a.
46.0*

34.2
43.2
33.6
50.3
53.2
38.1
6.5
51.1

Non-Financial Firms
Japan
Germany
United Kingdom
United States
Canada
France
Italy
Sweden

85.6
54.6
44.3*
35.6
62.0
63.6*
n.a.
n.a.

92.1
65.1
46.7
35.6
65.8
62.0
48.9
59.9

100.6
72.9
46.5
40.7
69.6
60.1
57.2
68.1

134.7
74.4
78.7
48.9
76.4
68.9
60.9
99.6

Public Sector
Japan
Germany
United Kingdom
United States
Canada
France
Italy
Sweden

22.5
20.3*
87.1*
56.
94.
45.
36,
n.a.

45.8
24.5
64.9
44.9
82.8
36.4
56.6
55.0

* Data for only part of period
+ 1989
Source: BIS

- 36 -

88.9
41.4
58.6
56.6
106.8
42.5
84.7
80.1

75.7
43.3
45.3
64.1
108.9
44.2
102.5
55.7




Kahn and Kole

TABLE 3
The Model

(1) D - a1q + a2Y + a^E +

D

-

value of stock market

(2) Y - d(D-Y)
(3)

aggregate demand

i - gY - h(M-P) + jq

Y

—

real output

(4) r - i - P

E

-

real exchange rate

(5) P - -f(P-P)

G

-

government spending

(6) P - M + (r - gY)/h

i

-

nominal interest rate

(7) q/q + Z/q - r

M

-

money stock

(8) r - r

P

-

price level (P in
equilibrium)

(9) Z - bj+ b 2 Y

r

-

real interest rate (in terms
of home goods)

(10) R - r + R/R

Z

-

profits

r

-

foreign real interest rate

R

-

long-term real interest rate

X

-

dx dt

X

-

equilibrium value of X

+ E

- 37

Kahn and Kole
TABLE 4: Estimates of Japanese Demand for Real M2+CDs, A(m-p)
1973:IV 1982:IV

1973:IV 1991:11

Variable

1983:1 1991:11

.464
(.58)

.065
(.06)

.571**
(5.07)

.328**
(2.86)

.365*
(2.04)

A(w-p) t

.150**
(3.44)

.269**
(3.92)

.196**
(3.00)

A(w-p)tl

.075*
(2.04)

.001
(1.59)

.073
(1.53)

-.860**
(-10.13)

-.552**
(-7.32)

-.952**
(-5!l8)

.530**
(3.44)

.089
(-.72)

.462
(1.54)

1.196*
(2.10)

-1.292*
(-2.33)

2.193*
(2.20)

-.988*
(-2.11)

.486
(1.20)

-1.277
(-1.24)

-.253**
(-4.84)

-.413**
(-5.01)

-.311*
(-2.28)

Intercept

1.005*
(2.31)

A(m-p) t _ 1

Ap,
Ap

t-1

"°t
A -PS
Al
t
m

t-r

p

t-i

yt-i
w




t-r pt-i

.0
t-1

L

.PS
1

t-l

Regression

.103*
(2.16)

.182**
(2.98)

.179
(1.54)

.121**
(4.88)

.252**
(3.20)

.157**
(2.72)

1.588**
(4.80)

-1.271
(-1.79)

1.107
(.86)

-1.268**
r-4.811

.651
(1.31N

-.907
(-.75)

statistics

R2
.895
Standard error
.00405
Sample size
71
2
Serial correlation (X )
lrst order
.070
lrst-4th order
2.843

.973
.00234
37
9.493**
12.271*

Fisher test

F

13.45 - 4 ' 6 2 **

T statistics are in parentheses
•^Significant at the 5 percent level
**Significant at the 1 percent level
- 38 -

.850
.00362
34
2.545
6.547




Kahn and Kole
TABLE 5: Estimates of German Demand for Real M3, A(m-p)
1970:11 1989:1V

Variable

1970:11 1979:IV
-1.610
(-1.49)

.331
(.88)

Intercept

1980:1 1989:IV
.381
(.94)

A(w-p) t

.456**
(4.02)

.367*
(2.19)

.708**
(4.04)

A w

.159
(1.69)

.244
(1.59)

-.060
(-.44)

Ay t

.050
(.63)

.289
(1.97)

-.073
(-.73)

Ap t

-.481**
(-3.15)

-.509*
(-2.60)

-.153
(-.47)

m

-.099
(-1.66)

-.327*
(-2.31)

-.105
(-1.25)

( -P>t-1

t - r Pt-i

yt-i
w

t - r Pt-i

.019
(.31)

.349
(1.95)

-.026
(-.40)

.054
(1.76)

.077
(1.35)

.102
(1.91)

.o
t-1

.371*
(2.04)

.PB

-.536**
(-4.03)

L

1

t-1

Regression

.072
(.23)
-.484*
(-2.41)

.586*
(2.26)
-.588**
(-3.34)

statistics

R2
.594
Standard error
.00620
Sample size
Serial correlation (X )
lrst order
.044
lrst-4th order
1.776
lrst-8th order
3.307

.601
.00706
39
.184
7.865
10.094

Fisher test

F

10,59 "X - 2 3

T statistics are in parentheses
^Significant at the 5 percent level
^^Significant at the 1 percent level

- 39 -

.554
.00499
40
1.586
4.333
8.875

Kahn and Kole

TABLE 6:

Estimates of U.K. Demand for Real M 4 , A(m-p)

162
(-' 36)

-.147
(-1.08)

Intercept

A(m-p)tl

A

1970:1
1979: III

1970:1 1991:11

Variable

^t-2

278

.337**
(4.01)

a. 78)

.182**
(2.74)

.265*
{2. .45)

-.944**
(-13.67)

(-s'• 79)

.972**

1979:IV 1991• ; H ,.
-2. 583**
(-2.94)
057
(• 36)

-2..87*
(-2..08)
-1..116**
(-8 • 26)

Ap t-1

.315**
(3.00)

(1 .41)

.091
( • 52)

Ap t-2

.190**
(2.66)

.128
(1 .21)

. .314
(-1 • 97)

-.055**
(-3.44)

.058
(-1 .67)

. .187**
(-4 .68)

m

t-rpt-i

-.028
(-1.45)

't-1
w




L

.285

.019
(- • 72)

.282*

(2 .52)

t-rpt-i

.087**
(4.71)

.082*

.094*

(2 • 15)

(2 • 53)

t-i

.297*
(2.49)

(1 • 61)

.TB
L
t-1

Regression

.696

-.321**
(-3^80)

.553**

(-3 • 14)

.087
( • 42)
- .013
(- • 10)

statistics

R<
.881
Standard error
.00663
Sample size
86
Serial correlation <*2>
lrst order
.920
Irst-•4th order
2.329

.906
.00726
39
2.437
3.957

Fisher test

11,64

T statistics are in parentheses
*Significant at the 5 percent level
**Significant at the 1 percent level

- 40 -

- 1.88

.805
.00531
47
2.065
10.802*

Kahn and Kole
TABLE 7: Marginal Significance Levels of Monetary Indicators for
Forecasting Alternative Measures of Economic Activity: Japan
Activity
Variables

M2+CDs

3-month
Rate

Term
Soread

Dollar
Exchange
Rate
«^i^___^»~»a~

1973:111-1992:1
.014

.086

.168

2) Retail Sales (RS)

.00003

.020

3) Trade Balance (TB)

.144

.302

.771
.064

.081

4) Unemployment (U)

.015

.303

.115

--

5) GNP

.001

.023

--

6) Consumption (C)

.005

.365
.304

.059

--

7)

.022

.395

.759

--

.286

.264

.273

.021

1) IP

.135

.380

.432

2) RS

.677

.189

.329

--

3) TB

.050

.875

.737

.499

4) U

.930

.959

.930

--

5) GNP

.253

.187

.604

--

6) C

.149

.298

.962

--

7) I

.912
.191

.001

--

.420

.003
.594

1) IP

.033

.678

.069

2) RS

.007

.005

.006

--

3) TB

.005

.972

.003

.771

4) U

.095
.022
.0003
.660

.144

.232
.002

--

1)

Industrial Production (IP)

Investment (I)

8) Net Exports (NX)
1973:111-1982:IV

8) NX

--

.001

1983:1-1992:1

5) GNP
6) C
7) I
8) NX

.522

.302
.491
.015
.029

--

.376
.200

--

.484

.040

--

For each activity variable, entries across the rows are the marginal significance
levels for the F-test that the coefficients on 6 lags of the monetary indicators
(columns) sum to zero in an unrestricted OLS prediction equation that also
included a constant, trend, 6 lags of the forecast variable, and 6 lags of a price
variable. The price variables used were the PPI in regressions including IP or U,
the CPI in regressions with RS, and a world commodity price index in regressions
with TB. Data are monthly -and all variables but interest rates, the trade
balance, and net exports are log levels. The term spread is the 10-year rate less
the 3-month rate. Quarterly data include real GNP, consumption, investment, and
net exports and are in 1985 yen.




- 41 -

Kahn and Kole
TABLE 8: Marginal Significance Levels of Monetary Indicators for
Forecasting Alternative Measures of Economic Activity: Germany
Dollar
•Exchange
Rate

3-Month
Rate

Term
Spread

.080
.005

.405
.006
.039

--.164

.442

--

.010

.006
.004

.013

.681
.001
.341

.00002
.016
.008

.002
.106

---

.103

-.007

1) IP
2) RS
3) TB

.779

.770

.995

--

.018

.0004

.006

--

.585

.720

.077

.542

4) U
5) GDP
6) C

.131*
.072

.035*
.071

---

.949

.00008*
.460
.041

.509

--

7) I
8) NX
1981:1- 1989:IV

.611

.044

.517

--

.647

.375

.019

.231

.084
.018
.514

.623
.833

.415*
.404

.667
.429
.605
.219*
.189

.475
.153*
.064

--.209
---

.690

.612

.555

--

.530

.272

.588

--

.021

.276

.299

.019

Activity
Variables

Ml

L973:Ii:I-1989:IV
1) Industrial Production (IP)
2) Retail Sales (RS)
3) Trade Balance (TB)
4) Unemployment (U)
5) GDP
6) Consumption (C)
7) Investment (I)
8) Net Exports (NX)

.036
.035
.913
.056

.048

1973:III-1980:IV

1) IP
2) RS
3) TB
4) U
5) GDP
6) C
7)

I

8) NX

For each activity variable, entries across the rows are the marginal
significance levels for the F-test that the coefficients on 6 lags of the
monetary indicators (columns) from an sum to zero in an unrestricted OLS
prediction equation that also included a constant, trend, 6 lags of the
forecast variable, and 6 lags of the PPI. Data are monthly and all variables
but interest rates, trade balance, and net exports are log levels. The term
spread is the rate on 5-7 year public bonds less the 3-month rate. Quarterly
data include real GDP, consumption, investment, and net exports and are in
1985 DM.
* 12 lags used for unemployment, PPI, and monetary variable.




- 42 -




Kahn and Kole
TABLE 9: Marginal Significance Levels of Monetary Indicators for
Forecasting Alternative Measures of Economic Activity: United Kingdom

Activity
Variables
1974:1-1991:11
1) IP

MO

M4

.186

.023

016

3-Month
Rate

Term
Spread

Dollar
Exchange
Rate

.0002

.003

--

.065

291

--

.028

.149
.105

5) GDP
6) C
7) I

658

.049

.042

.053

182
307

.130

.750
.074

8) NX

109

.145
.299

.372
.011
.051

.053

1974:1-1979:111
1) IP

.015

.660

641

.796

.875

--

.022

.545

.747

.016
.687
.070
.886
.021

-.191
.065
.317
.071

.961
.434
.628
.056
.328

.252
.466
.197
.092
.640

.123

.336

.046

.0001

.009

.0008

.209

.016

.136

.336
.009
.407
.768
.546
.360

.013
.603
.256
.910
.530
.669

.190
.561
.131
.424
.016
.842

.213
.411
.442
.545
.018
.063

2) RS
3) TB
4) U

2)
3)
4)
5)
6)
7)

RS
TB
U
GDP
C
I

8) NX
1979:IV -1991:111
1) IP
2) RS
3) TB
4) U
5) GDP
6) C
7) I
8) NX

.068

.101

.131

.751

For each activity variable, entries across the rows are the marginal
significance levels for the F-test that the coefficients on 6 lags of the
monetary indicators (columns) from an sum to zero in an unrestricted OLS
prediction equation that also included a constant, trend, 6 lags of the
forecast variable, and 6 lags of the PPI. Data are monthly and all variables
but interest rates, trade balance, and net exports are log levels. The term
spread is the 10-year rate less the 3-month rate. Quarterly data include real
GDP, consumption, investment, and net exports and are in 1985 pounds.
- 43 -

Kahn and Kole
Cham

INTEREST RATES AND INFLATION IN SELECTED G-10 COUNTRIES
UNITED STATES

JAPAN

Percent

Percent
25

25
•3-month Interbank Rate
— — - . - 12-month Rate of CPI Inflation
20

L

*

15

10

* X.Wv ~r**\
* ' • % , /

I M
1970

M M
1975

I I I I M
1980

M ll
1985

GERMANY

I I I M
1990

I 1I I I I I I I I I I I M I 1I I I I 1I 5
1970

1975

1980

-985

FRANCE

Percent

1990
Percent

20

' I

16

I—

16

12

f—

12

I I I I I 1 I I I I I I I t I t I I I I

1970

1975

1980

1985

UNITED KINGDOM

I M M I I I I I I I I I I I I I I I I I

1990

1970

"™^™l

A

L^

1975

CANADA

Percent

r-"

20

1980

1985

1990
Percent

30

25

^ J 25

11

20

1 •

U-

i 1

•

—| 20
15

J s& A t
It

XI \

15

if 1 h

10

y

¥-V

v-\-

''vl
\J
£1

1 1J_U
1 _LU
M l I 1 1 1 1 1 1 1 1 L-LL
1970

1975




1980

1985

10
5
0

1990
- 44

1 M
1970

IIIIIIIIIIIM IIIII I
1975
1980
1985
1990

Kahn and Kole

Chart 2: Model Dynamics
An Unanticipated Monetary Expansion

Time

To

Time

To

Time

Interest Rates
r.R

To

Time

Case a. "baa news" (normal case)

Case D: "good news"

Possible Exchange Rate Paths

To

Time

Case a* norma! case initial deoreciation.
then graoual return to baseline



To

Time

Case 0: initial depreciation, then
overshoot
-

(Ef = depreciation)

45

-

To
Casec: initial appreciation

Time

Kahn and Kole

Chart 3

Market Interest Rates and Own Rates of Return On Broad Money
14

Japan

T

12

1

H
6

l

4

i

2-r

w

' M2 + CDS

1976

1979

1982

1985

1988

1991

1988

1991

Germany

T979




i^IT

United Kingdom

20i

q 973

1982

197T

1979"

1982
-

U(y -

1985

MONETARY TRANSMISSION CHANNELS IN
MAJOR FOREIGN INDUSTRIAL COUNTRIES: A COMMENT
Craig S. Hakkio

In "Monetary Transmission Channels in Major Foreign Industrial
Countries," Robert Kahn and Linda Kole address "whether and how
these [monetary] transmission channels have changed during the
past two decades."

In studying the U.S. economy, Friedman (1989,

p. 96) finds that changes in the U.S. economy have "led to major
changes in standard reduced-form relationships of the kind that
often stand behind quantitative analysis of monetary policy at
either formal or informal levels."

Kahn and Kole have the more

difficult challenge of finding whether the transmission channels
have changed in Japan, Germany, and the United Kingdom.
In the first part of their paper, Kahn and Kole discuss why
the channels of monetary policy may have changed.

They then

present a stylized Keynesian model that supposedly "captures the
basic relationships that [the authors] hope to capture in the
empirical work."

However, the changes that occurred during the

last two decades are probably more complex than what can be
captured in a simple Keynesian model.

In these comments, I will

not discuss either of the first two sections.

Instead, I will

discuss the empirical results--presented in sections 4 and 5.
The evidence presented in "Monetary Transmission Channels in
Major Foreign Industrial Countries" suggests that the

1. Craig S. Hakkio is an assistant vice president and
economist at the Federal Reserve Bank of Kansas City. He thanks
Sean Becketti for comments on an earlier draft.



Hakkio
transmission channel may, or may not, have changed.

The authors

estimate money demand functions for Japan, Germany, and the U.K.
for the whole period and two subperiods.

They find that money

demand functions are stable in Germany and the United Kingdom,
but unstable in Japan.

The authors then estimate reduced form

equations for 8 activity variables, using 3 measures of monetary
policy, for the whole period and two subperiods.

They "find

evidence that corroborates the view that the interest rate
sensitivity of spending has increased over the past decade."
In these comments, I will first make some specific comments
on the paper by Kahn and Kole. Then, I will extend their results
by looking at whether financial markets have become more
integrated.

Finally, I will test whether their results are

sensitive to specification problems.

SOME SPECIFIC COMMENTS ON THE KAHN-KOLE PAPER
The authors1 finding that "the interest sensitivity of spending
(especially that of consumption) has increased in the past
decade" is surprising.
(1989, p. 30) states:

In studying the U.S. economy, George Kahn
"Empirical evidence suggests a reduction

in the economy's overall interest sensitivity.

This reduction in

interest sensitivity is not spread equally across all sectors of
the economy, however."

In particular, Kahn finds that

consumption is less interest sensitive, not more interest
sensitive as found by Kahn-Kole.




-2-

Benjamin Friedman (1989, p. 97)

Hakkio
reports similar findings.
In discussing their results, Kahn-Kole seem to argue that a
smaller marginal significance level means that the effectiveness
of monetary policy is larger.2 This is not necessarily true.
The effectiveness of monetary policy depends on the size of the
coefficient, in addition to its significance. And the marginal
significance level says nothing about the size of the effect;
rather it says something about the size of the coefficient
relative to its standard error.

If the coefficient falls and the

standard error falls more, the marginal significance level will
rise even though monetary policy has become less effective.

HAVE MARKETS BECOME MORE INTEGRATED?
Kahn and Kole argue that the transmission channels of monetary
policy have changed due to financial liberalization and greater
international openness.

Since financial liberalization often

took the form of opening financial markets to international
competition, I will consider these two explanations as one. With
more integrated financial markets, interest rates are determined
in a single world capital market.

Therefore, we would expect

German interest rate changes to be highly correlated with U.S.,
Japanese, and U.K. interest rate changes.

In addition, we would

2. The authors recognize this problem when they state:
"The fact that it increases in significance for several activity
variables in the later period could mean that the sensitivity of
activity variables (excluding investment) may have increased
between the periods."




-3-

Hakkio
expect monetary policy would be less able to influence interest
rates.
Therefore, instead of determining whether money demand
functions have shifted, or the sum of lag coefficients have
changed, we can look directly at whether changes in interest
rates have become more or less correlated over time.

If markets

are more integrated, then we would expect interest rate changes
to be more highly correlated.
To test this hypothesis, I collected daily interbank bid
rates from FAME for Japan, Germany, the U.K., and the United
States.

The interbank rate is a short-term interest rate. Since

the timing may be important, the June 24 interest rate quote is
at 10:00 am (local time) in Germany and the U.K., and at closing
(local time) in Japan; in the United' States, the interest rate is
the effective federal funds rate.

I then calculated the Spearman

rank correlation coefficient between changes in interest rates
and between the level of interest rates. The Spearman rank
correlation coefficient is a robust measure of association
between two variables; it is simply the correlation between the
ranks, as opposed to between the actual values.

I did not

include exchange rates in calculating a covered interest rate
because the results would be dominated by the exchange rate.
Table 1 gives the Spearman rank correlation coefficients.
The top half of the table gives the correlation coefficient for
the first difference ir* short-term interest rates, while the




-4-

Hakkio
bottom half gives the correlation coefficient for the level of
short-term interest rates. The first number in the cell is for
the whole period, while the other two numbers are for the 2
subperiods.

The subperiods are chosen to match those used in the

Kahn-Kole paper.

The breakponts are: Germany, December 31,

1979; Japan, December 31, 1982; and the United Kingdom, September
28, 1979.
The table shows that the rank correlation is about zero for
the first difference and between 1/3 and 3/4 for the level. For
example, the correlation between German and Japanese interest
rate changes, for the whole sample, is -0.001; the correlation
between the level of German and Japanese interest rates is 0.77.

There is little evidence that the correlations are bigger in
the second subperiod.

For example, for the first differences, 3

correlations become bigger in absolute value, 2 becomes smaller,
and 4 remain the same.
become smaller.

For the levels, 5 become bigger and 4

Of course, without standard errors we cannot

determine whether the changes are significant.

3. The marginal significance levels for the correlation
coefficients equal 0.00 for the correlation of the levels, and
are generally greater than 0.30 for the first differences. The
only exceptions for the first differences are: Germany and Japan
in the first subperiod (correlation = 0.07, msl = 0.10); Germany
and the U.S. in the whole period (correlation = 0.03, msl =
0.09), and in the second subperiod (correlation = 0.03, msl =
0.08); the U.K. and the U.S. in the whole period (correlation =
0.05, msl = 0.00), in the second subperiod (correlation = 0.05,
msl = 0.01).




-5-

Hakkio
Another way to test for changes in the extent of financial
market integration is to calculate whether big changes in shortterm interest rates are independent.

Define "big" to mean a

change in the upper or lower 5 percent tail of the distribution.
With this definition, 10 percent of interest rate changes are
"big."

Table 2 shows the results for changes in U.S. and German

interest rates. Table 2 is a two-way classification of interest
rate changes. The table shows that of 3978 observations, 56 (or
1.4 percent) had a big change in U.S. interest rates and a big
change in German interest rates. Given the definition of "big,"
we would expect 1 percent of the observations to fall in the BIGBIG cell if big changes in U.S. interest rates were independent
4

of big changes m German interest rates.
is a test of independence.

Fisher's exact test

According to the table, big changes

are correlated with big changes.
Table 3 reports similar results for all countries and for 2
subperiods.

The periods were chosen to match those used in the

Kahn-Kole paper.

The table reports the marginal significance

level of Fisher's exact test statistic for independence.

A small

marginal significance level means you can reject the hypothesis
that big changes are independent of big changes; less precisely,
4. Actually, we would expect 0.95 percent of the
observations to fall in the BIG-BIG cell. Since there are
missing observations for each variable and the table is
constructed for non-missing observations for both variables, the
BIG row and column sums do not equal 10 percent. As a result, if
the changes are independent, we expect to find 0.107*0.089 =
0.0095 =0.95 percent of the observations in the BIG-BIG cell.




-6-

Hakkio
a small marginal significance level means that big changes are
correlated with big changes.
The results in Table 3 suggest that financial markets did
become more correlated in the second subperiod.

In most cases,

the marginal significance level fell in the second subperiod.
The two exceptions were (1) Germany and the U.S., where the
marginal significance level rose from 0.00 to 0.01, and (2) the
U.K. and Japan, where the marginal significance level rose from
0.46 to 0.49.

Also, in many cases the marginal significance

levels are less than 10 percent in the second subperiod.

For

example, we can reject the hypothesis that big changes in German
interest rates are independent of big changes in Japanese and
U.S. interest rates.
To summarize, the results in this section complement the
results in the Kahn-Kole paper.

There is some weak evidence that

financial markets have become more integrated.

As a result, we

would expect that monetary policy transmission channels would
change.

However, since the evidence on greater integration is

weak, the change in transmission channels is probably also weak.

WHY ARE THE RESULTS WEAK?
The inconclusive or weak results could truely reflect little or
no change in the transmission channels, or they could reflect
statistical problems.

If we can minimize the chance of

statistical problems, then we can be more confident that the




-7-

Hakkio
results really reflect little or no change in the transmission
channels.

Therefore, this section looks at some potential

statistical problems.
Are the Results Sensitive to Outliers?
The presence of outliers could produce the inconclusive
results reported by Kahn-Kole.

To check for outliers, I first

estimate a reduced form equation for industrial production. The
general form of the equation is similar to that used by the
authors:
K

log (ind prod)

t

- <x0 + a1TIMEt + ]£ P i (monetary policy)

t_i

i-l
X

+ ]T Yilog( ind prod) t_± + et
2-1

P

-EPi
1-1

Monetary policy is measured by a monetary aggregate and by shortterm interest rates. The lag length, K, is determined from
Akaike's Information Criterion and Amemiya's Prediction
Criterion.

I search for influential observations in this

regression in several ways.
An influential observation, or a small influential subset of
data, is one which "can have a disproportionate influence on the
estimated parameters or predictions" of a regression equation

5. Generally, 2 to 4 lags were required, fewer than used by
the authors.




-8-

Hakkio
(Belsley, Kuh, and Welsch (1980), p. 6). An observation may have
a big influence on the fitted values of a regression, on the
variance-covariance matrix of the coefficients, or on the sum of
lag coefficients. Accordingly, three statistics are used in this
paper to detect influential observations.6 The first, Cook's
distance, measures the influence of the t-th observation on the
fitted values from a regression.

The second, COVRATIO (Belsley,

Kuh, and Welsch), measures the influence of the t-th observation
on the variance-covariance matrix of the coefficients.

The last,

a variant of DFBETA (Belsley, Kuh, and Welsch), measures the
influence of the t-th observation on the sum of the lag
coefficients of the regression.
each statistic.

Critical values are given for

In the results presented below, I focus on only

the largest value of the statistic (which is also greater than
the critical value).
To find an influential observation, the regression is
estimated with all observations and with all but the t-th
observation.

Then, a normalized difference in some statistic is

calculated with and without the t-th observation.

Finally, the

normalized difference is compared to a critical value. A large
normalized difference means the observation is influential.
As an example, consider the variant of the DFBETA statistic.

6. See Chatterjee and Hadi for a discussion of influential
observations in linear regressions. They state that these three
measures "seem sufficient for detecting influential observations"
(p. 387).




-9-

Hakkio
I calculated a time series of sum of lag coefficients obtained
from omitting one observation at a time. More specifically,
DFBETA

« (j8 - $t)/aR

(t) ' w h e r e & i s t h e

sum of the

la

9

coefficients, j8. is the sum of the lag coefficients obtained from
omitting the t-th observation, and a*. . is the standard error of
the sum of lag coefficients.

In other words, DFBETA

is like a

t-statistic: it equals the difference in coefficient estimates
divided by the standard error.

If results are sensitive to an

outlier at observation t, then DFBETA

will be "large."

Table 4 shows the dates of the influential observations in a
reduced form with monetary policy measured as money and shortterm interest rates.

Each cell in the table gives the date of

the most influential observation, the fraction of observations
that are influential (the number to the left of / ) , and the size
of the largest statistic relative relative to the critical value
(the number to the right of / ) .
As expected, the different statistics find different
influential observations.

The fraction of influential

observations ranges from 1 percent to 10 percent; and the size of
the largest statistic ranged from 1.3 times the critical value to
31.8 times the critical value.
Unfortunately, no simple conclusions can be drawn from the
table.

While no simple conclusions can be drawn, it is clear

7. See Belsley, Kuh, and Welsch (1980) for an extended
discussion of the DFBETA statistic.




-10-

Hakkio
that with so many influential observations, some of which are
very "large," the results may be due to influential observations.
Influential observations can be due to improperly recorded data.
Alternatively, they can be legitimately extreme observations that
contain valuable information about the parameter estimates.
However, even in this situation it is important to identify the
observations and determine the extent to which they are
responsible for the results. That is, we want to know whether
the results are due to this one observation, rather than the
entire dataset.

Unfortunately, such an analysis is beyond the

scope of these comments.

Are the results sensitive to choice of subperiods?
The results could also be inconclusive because the authors
split the sample at the wrong place. As a result, estimating a
reduced form equation over two subperiods may miss the change.
In addition, while it may be reasonable to think the change
occurred at a single point in time, it is probably more likely
that there have been several changes or that the changes occur
gradually over time.

If the effectiveness of monetary policy is

changing gradually or has changed more than once, then estimating
a reduced form over two sample periods may again miss the change.
To allow the change in monetary policy to occur over time, I
estimated a series of rolling regressions.

I estimate the same

reduced form as in the previous section. Monetary policy is




-11-

Hakkio
measured as either a monetary aggregate or the short term
interest rate. The sample period is a fixed 20 percent of the
observations.

Charts 1-3 plot the sum of the lag coefficients

with a 2 standard deviation confidence band.

The top panel is

the sum of money lag coefficients and the bottom panel is the sum
of short-term interest rate lag coefficients.
In Japan, the sum of money lag coefficients changed from
negative in the first part of the sample period to positive in
the second part.

Furthermore, the standard errors have become

smaller over time.

The sum of interest rate lag coefficients has

changed over time, but there is no pattern.

Except for two

episodes, the sum was about 0 for most of the 1980s; the sum then
turned positive in the 1990s.
There is little evidence of a change in the effectiveness of
German monetary policy.

The sum of both money and interest rate

lag coefficients has fluctuated around 0 for most of the sample
period.
In the United Kingdom, monetary policy seems to have become
less effective.

The sum of money coefficients was positive in

the early part of the sample, and has been zero since then.
However, the sum of interest rate coefficients has fluctuated
around zero for most of the sample period.
To summarize, while the sum of lag coefficients have changed
over time, the changes may not have been significant. Therefore,
the weak results reported are probably not due to the authors1




-12-

Hakkio
choice of subperiods.

CONCLUSIONS
The authors present a comprehensive study of changes in the
monetary policy transmission mechanism in Japan, Germany, and the
United Kingdom.
change.

The find weak evidence that there has been a

In looking at different data and techniques, I have

confirmed their results.




-13-

Hakkio
REFERENCES

Belsey, David A., Edwin Kuh, and Roy E. Welsch, Regression
Diagnostics:

Identifying Influential Data and Sources of

Collinearity, John Wiley & Sons, New York, 1980.

Chatterjee, Samprit and Ali S. Hadi, "Influential Observations,"
Statistical Science. August 1986, pp. 379 - 392.

Friedman, Benjamin, "Changing Effects of Monetary Policy on Real
Economic Activity," in Monetary Policy Issues in the 1990s,
a symposium sponsored by the Federal Reserve Bank of Kansas
City, 1989.

Kahn, George A. "The Changing Interest Sensitivity of the U.S.
Economy," Economic Review, Federal Reserve Bank of Kansas
City, November 1989, pp. 13 - 34.




-14-

Hakkio
l.

Correlation of Short-term Interest Rates
(Spearman rank correlation matrix of first differences and
levies)

Germany

Japan

United
Kingdom

U.S.

First difference of short-term interest rates
Germany

-0.00

0.01

0.03

-0.01
0.00

-0.02
0.03

-0.02
0.06

0.01

-0.01

0.01
0.01

0.01
-0.01

*

-0.00
|

||

*

Japan
-0.01
0.01
United
Kingdom

0.01

0.05

0.01
*

|

0.01
0.02

0.05
0.05

0.03
0.00

Levels of s hort-term interest rates
1

Germany

|
||

0.77

0.61

0.47

0.74
0.73

0.42
0.59

0.13
0.42

0.63

0.43

0.73
0.49

0.61
0.12

•

0.77
*

Japan
0.72
0.66
United
Kingdom

0.61

0.45

0.63
*

1

0.34

0.33

0.10

P_._5_9.__

0___6_4.

0.40

1

Note:
Each number is a rank correlation coefficient. The numbers
are for different subsamples:
row 1
whole subsample
row 2
first subsample, determined by row variable
row 3
second subsample, determined by row variable
The German/Japan correlation does not equal the Japan/German
correlation in rows 2 and 3 because the breakpoints for the
subsamples are different.




-15-

2.

Normal and Big Changes in German and U.S. Short-term
Interest Rates

Change in U.S. interest rates

Change
!

row
sums

normal
change

BIG
change

3253

300

3553

81.8 %

7.5 %

89.3 %

369

56

425

9.3 %

1.4 %

10.7 %

3662

356

3978

91.0 %

8.9 %

in
German
interest

normal
change

rate
BIG
change

column
siims

;

Test of independence of row and column variables:
Fisher's marginal significance level = 0.002




100 %

Hakkio
3.

Normal and Big Changes in Short-term Interest Rates
Two-way Table of Frequency Counts

Germany

Germany

Japan

U.K.

U.S.

0.00

0.26

0.00

0.50
0.00

0.67
0.26

0.00
0.01

0.69

0.37

1/1/80
*

Japan
1/3/83

U.K.
10/1/79

0.00
0.29
0.00

*

0.26

0.77

0.66
0.17

0.46
0.49

0.11
0.10

1.00
0.06

0.03
*

0.41
0.08

Note:
A big change is defined as a change in the top or bottom 5
percent of changes. Therefore, 10 percent of the changes
are big.
The number in2the cell is the marginal significance level of
the Pearson x test of independence between the row and
column variable. Each cell has 3 numbers, corresponding to
different sample periods:
row 1
whole subsample
row 2
first subsample, determined by row variable
row 3
second subsample, determined by row variable
Note, the German/Japan pair does not equal the Japan/German
pair in rows 2 and 3 because the breakpoints for the
subsamples are different. The first observation of the
second subsample is given below the country name; the date
corresponds to the breakpoint used in Kahn-Kole.




-17-

Hakkio
4.

Detecting Influential Observations in a Reduced Form of
Industrial Production
Germany

Japan

United
Kingdom

Feb 1991

May 1991

June 1985

5% / 20.4

7% / 2.8

10% / 3.2

July 1984

Feb 1976

Jan 1974

4% / 15.6

6% / 6.0

6% / 12.3

March 1991

June 1989

May 1990

6% / 31.8

11% / 4.3

7% / 5.3

May 1981

June 1975

Sept 1973

8% / 9.5

10% / 8.0

6% / 11.6

June 1984

Oct 1982

Sept 1985

4% / 2.4

3% / 1.3

5% / 2.1

April 1979

April 1989

Jan 1974

1% ../ 1.6

3% /._1=_8

3% / 4.7

Statistic:
moneyCook•s
distance
interest
• rates
money
COVRATIO
interest
rates
money
DFBETA
interest
rates

Note:




j

The first number in each cell is the date of the
largest statistic. The pair of numbers on the second
line give the fraction of observations greater than the
critical value (the number to the left of /) and the
size of the statistic relative to the critical value
(the number to the right of / ) .

-18-

Chart 1
Sum of Lag Coefficients - Japan
15

15

M2
H 10

10

H5

-5

-10

-10

I

I

I

I

l

l

1878-1

1979-1

196*1

1991*1

1912:1

1993*1

-15

1994*1

1995:1

199*1

l

l

I

l

l

I

1997:1

1999*1

1999:1

1990:1

1991*1

1992:1

-15

Year
0.06

0.06

Interbank Interest Rate
0.04

h

0.04

0.02

h

0.02

-0.02

-0.02

-0.04

-0.04

-0 06

-0.06




Chart 2
Sum of Lag Coefficients - Germany
M3
H2

-2 h

J
197T1

tfTKI

L
HT«1

Itaaci

l

I

1WV.1

IMtl

l
1W*1

i

l

l

1«MC1

1«Mct

1MK1

t

i

1

I

I

I

1WK1

1WT1

IMMC1

tMVI

t«t£1

Year
0.04

0.04

Interbank interest Rate

0 02

H 0.02

-0.02

-0.02

J

-0.04
tITM




t»7fl

t«71-1

L
ItN-l

J
IM1 t

tMBM

INTI

J

L
10««'1

IMf:f

Year

1MT1

tttT"!

I
J9W1

I
tNTf

'
ItNtl

'

'

IHt 1

t»S2:f

-0.04

Chart 3
Sum of Lag Coefficients - The United Kingdom
15

15

10

H 10

-5

-5

-10

-10

t

!

-15

i

l

l

J

I

L

'

I

!

-15

1977:1 1978:1 1979:1 1980:1 1981:1 1982:1 1983:1 1984:1 1985:1 1986:1 1987:1 1988:1 1989:1 1990:1 1991:1 1992:1

Year
.015

0.015

Interbank interest Rate
0.01

H 0.01

h
t \

i

.005

hh

t

H 0.005

A

A

*/

J

A

V

/w

/
/

x

\

\

r

v
t

005

VY

v

*

hh

v

\/ '

--V/

H
J

» /

1

r

l

.015

V

h

-J

I

L_

1

1

!

1

1

1

1

1-

-L

'

1

/V/'

\ /'

V
0.01

i 1

. *' /

'

V

'

H -0 01

'

! I -0.015

1977:1 1978:1 1979:1 1980:1 1981:1 1982:1 1983:1 1984:1 1985:1 1986:1 19871 1988:1 1989:1 1990:1 1991:1 1992:1




Year

-0.005

ANOTHER HOLE IN THE OZONE LAYER:
CHANGES IN FOMC OPERATING PROCEDURE
AND THE TERN STRUCTURE
William Roberds, David Runkle, and Charles H. Whiteman1

To economists schooled in the Walrasian tradition, there could be no
more enigmatic ritual than the practice of central banking. The
classical models of this tradition show how Pareto-optimal allocations
can be realized in competitive equilibrium, in which prices reflect the
fundamentals of tastes, technology, and endowments. In response to
changes in these fundamentals, prices must also change if competitive
allocations are to remain optimal. By contrast, since 1914 the unifying
theme of real-world monetary policy has been the elimination of shortrun movements in short-term interest rates, typically via open market
operations in government securities. The obvious implication of this
practice, which has become known as interest-rate "smoothing," is that
central banks (and their sponsoring governments) find such fluctuations
in the time price of money to be inherently undesirable.
This apparent contradiction between high theory and everyday
practice has hardly gone unnoticed by the economics profession. In
fact, this contradiction has formed one of the traditional jumping-off
points for much of the monetarist and neoclassical criticism of the
policies of the Fed and other central banks. Yet the constant criticism
of interest-rate "smoothing" from this quarter seems to have had almost
no effect on the practice of monetary policy. A recent survey of
operating procedures in five major industrialized countries (Batten et
al., 1990) found that short-term control of interest rates was tightened
in virtually all of these countries during the 1980s.
Perhaps in response to the continued popularity of interest-rate
smoothing, a number of papers have appeared in the macroeconomics
literature, in which economists working in the Walrasian tradition*have
taken a more benign view of interest-rate smoothing. These papers run
the gamut from Sargent and Wallace (1982), which presents a model where

1. William Roberds is on the staff of the Federal Reserve Bank of
Atlanta. David Runkle is on the staff of the Federal Reserve Bank of
Minneapolis and on the faculty at the University of Minnesota.
Charles H. Whiteman is on the faculty at the University of Iowa.
We thank David Stowe for help in data collection. This paper was
written while Whiteman was a visiting scholar at the Federal Reserve
Bank of Atlanta; it reflects only the views of the authors, and not the
Federal Reserve Banks of Atlanta and Minneapolis or the Federal Reserve
System.




Roberds, Runkle, and Whiteman

the best monetary policy is an interest rate peg, to Poole (1991), which
suggests that the practice of interest-rate smoothing could in some
instances serve as a potentially useful method for communicating a
central bank's intentions to the public. While these papers have
offered a number of insightful explanations for the smoothing
phenomenon, it is fair to say that none of the explanations has been
widely accepted within the economics profession. Instead, various
factions within the profession have supported disparate views of
smoothing, reflecting the more general professional quandary over the
proper role for money in macroeconomic models.2
At the same time, there has been some acceptance on the part of
the Federal Reserve System of the idea that it is possible to be too
aggressive in the smoothing of short-term interest rates. In
particular, the operating procedure of the late 1970s, which was almost
completely focused on smoothing the Federal Funds rate, is typically
viewed as a mistake. In an article in the Federal Reserve Bulletin,
Heller (1988, p.425) notes that the emphasis on the funds rate
contributed to a loss of control over the growth of the monetary
aggregates. Essentially identical sentiments are voiced in a later
issue of the Bulletin, in an article by Donald Kohn (1990, pp. 4-5). In
a New York Fed-sponsored survey of various Fed approaches to open market
operations, Meulendyke (1990) observes that during the late 1970s, "[Fed
funds] rate moves during the week were so limited that they provided
little or no information about reserve availability or market forces."
This combination of a desire on the part of policymakers to
avoid past mistakes, together with the ongoing professional impasse over
the conduct of monetary policy appears to have led to an "eclectic" or
"compromise" approach to the day-to-day conduct of open market
operations. The essence of this approach is perhaps best summarized in
the survey of Batten et al. (1990, pp.30-32). Describing the general
approach adopted during the 1980s, Batten et al. note that
"operating procedures in [the U.S. and other major industrialized
countries] generally allow short-term rates to be primarily marketdetermined, while at the same time, permit monetary authorities to limit
the range within which these rates fluctuate. Each monetary authority
sees the need for interest rates to adjust expeditiously to reflect new
economic developments but also recognizes the importance of maintaining

2.




See Goodfriend (1991) for a survey.

-2-

Roberdsf Runkle, and Whiteman

some discretionary control over interest rate movements to avoid
excessive volatility." [pp. 30-31]
An inherent limitation of this approach is that, lacking any real
theoretical guidance, it provides no specific definition as to what
level of interest rate volatility is "excessive." Without some specific
criteria that define a well-functioning credit market, the "avoidance of
excessive volatility" in short-term interest rates cannot be construed
as a meaningful objective for monetary policy.
One reasonably objective and recently popular metric for
evaluating the impact of interest-rate smoothing on the bond markets has
been to compare the information content of the term structure,
particularly at horizons of less than a year, across periods of time
associated with different regimes for monetary policy. The intuition
behind the use of the term structure for this purpose is fairly simple.
If interest rates are to "adjust expeditiously to reflect new economic
developments," then the spreads between long and short rates should
contain some useful information about the future course of interest
rates. This is because interest rates represent intertemporal prices,
prices one would expect to be affected by news about the likely future
course of the economy. One of the most widely cited papers in this area
is by Mankiw and Miron (1986), who consider the performance of the short
end (less than one year) of the term structure over various periods
ranging from 1890 to 1979. They find that the term structure was more
informative prior to the founding of the Fed. They hypothesize that
this result is due to interest-rate smoothing activities on the part of
the Fed after 1914. The salient claim of their paper is that there
exists a tradeoff between the desire to smooth interest rates on the one
hand, and the informativeness of the term structure on the other.
Other papers in this tradition include Cook and Hahn (1990), Hardouvelis
(1988), Mankiw, Miron, and Weil (1987), and Simon (1990).
In what follows, we seek to apply the term structure yardstick
to the Fed operating procedure that has been in place since early 1984,
technically known as borrowed-reserves targeting with contemporaneous
reserve accounting. Specifically, we are interested in the ability of
the term structure in the Fed funds market to predict subsequent moves
in Fed funds rates, at horizons ranging from one to six months. We also
try to measure the information content of the spreads between Fed funds
rates and closely related rates on Treasury bills and repurchase
agreements (repos). We are especially interested in comparing the term
structure during the current operating procedure to the term structure




-3-

Roberdsf Runkle, and Whiteman

under other recent operating procedures* Our results should be of
interest to policymakers, given the widespread acceptance of the idea
that successful monetary policy should not incorporate interest-rate
smoothing to the same extent as was the case during the late 1970s. Our
results should also be of interest to monetary theorists, as the set of
stylized facts presented below presents a challenge to any theory that
would attempt to explain the interaction between a central bank's open
market operations and the information contained in the term structure.
Our study differs from previous studies in this area primarily
in that we make use of daily data on yields for Fed funds and related
markets. Previous studies that have attempted to measure the impact of
interest-rate smoothing on the term structure have made use of weekly or
lower frequency data. Since a major emphasis of the Fed's open market
operations has traditionally been the smoothing of day-to-day interest
rate changes, the use of daily data is necessary to fully capture the
dynamics of the yield curve.
INSTITUTIONAL BACKGROUND
Although the smoothing of short-term interest rates has always
been an important component of Federal Reserve policy, this practice
reached a new stage during the 1970s. The development of the overnight
market for bank reserves, popularly known as "Fed funds" provided the
Fed with an efficient vehicle for large, frequent, short-lived
interventions in this market.3 Particularly during the latter half of
the 1970s, short-run Fed policy came to focus almost exclusively on the
funds rate target.
From October 1979 through October 1982, nonborrowed reserves
(bank reserves not borrowed from the Fed) replaced the funds rate as the
official short-run operating target. This change in operating targets
was accompanied by a marked increase in the volatility of the funds
rate. As is discussed in further detail below, the standard deviation
of daily changes in the funds rate increased roughly threefold. Despite
this degree of volatility, however, it is doubtful that fluctuations in
the funds rate were completely ignored during this time period. A
recent study by Cook (1989), found that despite the nominal adoption of
the nonborrowed reserves target, two-thirds of the variation in the
funds rate during the October 1979-October 1982 period can be directly
attributed to policy actions on the part of the Fed; that is, these

3. Goodfriend and Whelpley (1986) present a useful historical
summary of the Federal funds market.




-4-

Roberds, Runkle, and Whiteman

movements in the funds rate were not necessary to meet the nonborrowed
reserves target.
In October 1982, the Fed's short-term operating target was
changed from nonborrowed reserves to borrowed reserves. Despite the
continued nominal use of a reserves target, this change has been widely
perceived (e.g., by Friedman, 1988) as a retreat towards the funds ratetargeting of the 1970s. Statistical comparisons of the two periods are
somewhat problematic due to a change in reserve accounting that was
instituted by the Fed in early 1984. Under the pre-1984 accounting
procedures (commonly referred to as "lagged reserves accounting")
required reserves were computed over a week-long computation period, and
had to be maintained with a two-week lag. Under the post-1983
accounting procedures (commonly known as "contemporaneous reserves
accounting"), required reserves are computed over a two-week period, and
must be maintained with a two-day lag. A characteristic feature of the
new accounting procedure has been the introduction of an occasional
spike in the overnight funds rate on alternate Wednesdays, i.e., the
last day of the two-week reserve maintenance period.4
STATISTICAL FINDINGS
Data and Suaaary Statistics
In what follows, we use daily data on Fed funds, repo, and
T-bill rates at maturities of one day, 30 days, 90 days, and 180 days to
examine the predictions of yield spread about future movements in short
term interest rates.5 The sample starts in the fall of 1974 and runs
until the summer of 1991. Within that sample, we examine three of the
different operating regimes:6 the Fed-funds targeting regime (using a
sample from January 2, 1975 to October 3, 1979); the nonborrowed

4. For a more detailed description of the mechanics and implications
of the change in reserve accounting procedures, see Goodfriend (1984).
Additional details concerning recent approaches to open market policy
can be found in Meulendyke (1989,1990) and Heller (1988).
5. Repurchase agreements, or repos, are short-term loans collateralized by a fixed-income security. For more information on repos, see
Lumpkin (1986) or Stigum (1989). One problem in comparing repo rates is
that different rates can be quoted for repos using different types of
collateral. To minimize this problem, we look at data specifically for
repos that are collateralized by Treasury securities.
6. We exclude the borrowed reserves targeting-lagged reserves
accounting regime (October 1982-January 1984) because there are too few
observation to conduct meaningful inference about the term structure of
interest rates.




-5-

Roberds, Runkle, and Whiteman

reserves targeting regime (using a sample from October 11, 1979 to
October 6, 1982); and the present regime-borrowed reserves targeting
with contemporaneous reserve accounting (using a sample from February 2,
1984 to July 24, 1991.)
The overnight Federal funds rate we use is the effective Federal
funds rate computed by the Federal Reserve Board, which is a transaction-weighted average. All other data for Federal funds rates and
repurchase agreement rates represent the daily closing quotes from the
Bank of America at 5:00 p.m. Eastern Time. The repurchase agreement
quotes are for transactions collateralized by Treasury securities.
Since both Fed funds and repo rates are originally stated on a 360-day
basis, they are all converted to bond-equivalent yields for comparison
with other data. Data for one-month, three-month, and six-month7
Treasury bill rates come from the Federal Reserve Board. These data are
stated as discounts for an average of bid quotations for the most recent
issue, and are also converted to bond-equivalent yields. A brief
description of our dataset is presented in Table 1.
Summary statistics for daily changes in the various interest
rates during the four operating regimes are summarized in Tables 2 and
3, and Figures 1-5.8 The results in Table 2 reveal four characteristics
of fluctuations in the rates.
First, the sample standard deviations of daily changes in the
rates document the well-known increased volatility in interest rates
across all maturities during the nonborrowed reserves targeting period.
For example, the standard deviation of the daily change in the effective
Federal funds rate FFEY increased from 0.301 (30.1 basis points) during
the funds targeting period to 0.823 during the nonborrowed reserves
period. The volatility dropped markedly after 1982, to 0.316 in the
borrowed reserves-lagged accounting period, and 0.378 during the most
recent contemporaneous accounting period.
Second, the higher-order moments summarized in the skewness (Sk)
and kurtosis (Ku) measures are consistent with the view that while
volatility increased during the nonborrowed reserves period, outliers
were less important. With few exceptions, Sk and Ku are smaller during

7. Since bills are auctioned only once weekly, their maturities
fluctuate on a periodic basis. For example, a "91-Day" T-bill typically
has a maturity of 91 days on its issue date (Thursday), a 90-day
maturity on the following day, etc.
8. The complete set of tables of summary statistics (112 pages in
16.67 pitch type) is available from the first author for a nominal fee
to cover reproduction and postage.




-6-

Roberds, Runkle, and Whiteman

1979-82 than in the funds-rate targeting regime or the borrowed-reserve
contemporaneous-reserve-accounting regime* The similarity between the
higher moment measures pre-1979 and post-1984 provides further evidence
that these periods were alike, and highlights the difference between the
practice of permitting only infrequent changes in the Federal funds
target during 1975-79 and post-1984 and the "practice" of permitting the
Funds "target" to change daily during 1979-82.
Third, except for 1979-82, there is a tendency for the
volatility in daily changes to fall with the maturity of the underlying
contract. For example, in the 1984-91 period, the standard deviations
of overnight, thirty-, sixty-, ninety-, and one hundred eighty-day
Federal funds rates were 0.378, 0.151, 0.136, 0.151, and 0.16. During
1979-82, rates on daily contracts were more volatile than those on
longer contracts, but otherwise the volatility-maturity relationship
seems absent.
Fourth, the Federal funds rate tends to be more volatile than
the repo rate at each maturity. An exception to this is the 1979-82
period, when the two rates were about equally volatile.
The composite statistics of Table 2 conceal large interday
differences displayed in Tables 3 and 4. Table 3 presents statistics by
day of the Federal Reserve maintenance period for overnight Federal
funds, and shows characteristics shared by the other overnight rates and
to a lesser extent by rates on weekly contracts; Table 4 presents
statistics for 3-month T-bills, and shows the characteristic pattern of
other medium-term rates.
Table 3 displays the striking effects on short-term rates of
reserve requirements. Even during 1979-82, volatility in daily changes
is noticeably higher on the first and last days of the settlement
period. Before 1979 and after 1982, the differences are quite large:
pre-1979, volatility increases by a more than factor of four from Monday
(day 1 of the five-day settlement period) to settlement Wednesday. From
1988 on, volatility increases by a factor of greater than six between
nonsettlement Wednesday (day 5) and settlement Wednesday (day 10) and
the following Thursday (day 1 ) .
Table 3 also hints that the large movements on Wednesdays and
Thursdays tend to be offsetting. Skewness switches sign between these
days, as does the midpoint between the minimum and maximum values.
The fat tails and otherwise odd features of the distribution of
daily changes is apparent in Figures 1-5. Figure 1 displays the
distributions of daily changes for the four operating regimes, and
illustrates clearly the leptokurtic nature of the distribution of




-7-

Roberds, Runkle, and Whiteman

changes during the nonborrowed reserves period. The remaining figures
illustrate the distributions by selected days of the settlement period,
and indicate the generally fatter tails which occur on settlement day.
The results in Table 4 indicate that the interday pattern of
daily rates does not extend to rates for longer horizons. Volatilities
across days are quite similar, and the distributions more nearly
symmetric. Furthermore, the results differ little across operating
regimes—a characteristic not shared by the term structure restrictions
investigated below.
Tests of the Tern Structure Restrictions
The most direct method for testing the implications of the
expectations theory of the term structure is the so-called VAR (vector
autoregressive) approach.9 While the results of the VAR tests are less
easily interpreted than those of the other tests presented below, they
do provide summary measures of the overall validity of the expectations
model of the term structure over the various policy regimes. To
describe these tests, let R^ denote a longer-term, n-period rate of
interest, and let RVB denote a shorter, m-period rate of interest, where
a divides n. The risk-adjusted expectations hypothesis then states that
the n-period interest rate at time t, R^ is the average of the current
m-period interest rate R^, and current expectations about future
m-period rates, plus a time-invariant risk premium; i.e.,
(1)

R^ « (l/k)^:i tRt+H^+c, k = n/m,

where tRt+krin is the expectation at time t of the m-period interest rate
starting in period t+k. In the textbook case, the {R^} and {R^}
processes are jointly Gaussian and covariance stationary.10 Then it is
a straightforward, though some what tedious exercise to apply the
standard techniques of rational expectations to derive the implications
of equation (1) on the fundaunental moving average representation for the
U\», Km)} process.

9. The approach is discussed in more detail in Campbell and Shiller
(1987,1991) and Hodrick (1991).
10. These processes must satisfy other technical requirements in
order to apply standard rational expectations methodology. See Hansen
and Sargent (1991) for a thorough discussion of the econometric issues
associated with tests of the expectations model of the term structure.




-8-

Roberds, Runkle, and Whiteman

We adopt a number of modifications on this basic strategy,
following the approach taken by Campbell and Shiller (1987). First,
equation <1) is approximated by assuming that n/m is large, and taking
an infinite horizon counterpart, i.e.,
(2)

R ^ « (1-5) ST.* 6> ^ ^

-he

where 6 is a discount factor in (0,1). This modification allows the
restrictions imposed by the expectations model to be expressed in a
linear form, which reduces computational complexity* Second, in this
application we assume ««1 day, that is, the short rate is taken to be an
overnight rate. Third, due to evidence in favor of differencestationarity of the various interest rate processes, it is advantageous
to rewrite (2) as
(3)

R^ - R ^ H S*-> = Zmiml 6l (tRt+nri4--Rt+«(M),») +C

Equation (3) states that the spread between the long rate and the short
rate, S ^ must equal a discounted sum of expected future changes in the
short rate. The fourth and final modification is to assume that the
bivariate, stationary process (R^-R^, Ru-Rt-u) h a B a V A R representation.
Under these modifications, the implications of the expectations model of
the term structure can be shown to be equivalent to a set of linear
restrictions on the coefficients of the VAR for (R^-R^, RU-RM.I). 11
Representative results for these tests are shown in Table 5. In
these applications, the short interest rate was taken to be the effective overnight Fed funds rate, and the long rate was taken to be the
3-month Treasury bill rate. A VAR model was fit to daily observations
on first differences in the overnight funds rate and the spread between
the T-bill rate and the funds rate. Missing observations were filled in
by repeating the previous day's values. Standard tests for lag length
revealed that 21 lags were sufficient to capture the model dynamics
after October 1979. For the funds rate targeting period, however, these
tests were somewhat ambiguous. Hence, for this period, Table 4 presents
results for a 42-lag VAR system as well as for a 21-lag system.

11. See Campbell and Shiller (1987, pp. 1066-1068). Following
Campbell and Shiller, the expectations-model restrictions are tested
using a Wald test. The variance-covariance matrix of the coefficients
is the heteroskedasticity-consistent estimator suggested by Hansen
(1982).




-9-

Roberds, Runkle, and Whiteman

The results in Table 5 show that the expectations model can be
rejected at arbitrary significance levels for the funds rate targeting
period. The expectations model cannot be rejected on the basis of the
data from the 1979-82 period, although the smaller sample size associated with this period makes this finding somewhat less informative than
might be the case otherwise. The test results for the post-1984 sample
fall in an intermediate range: the expectations model can be rejected
at the 5% but not at the 1% level of significance*
The VAR results accord with other studies of the short end of
the term structure that report subsample results for different Fed
operating procedures, e.g., Simon (1990) and Hardouvelis (1988). They
indicate that the expectations model can be taken literally only over
the relatively short subsample associated with nonborrowed reserves
targeting, that it is unlikely that much information can be recovered
from the short end of the yield curve during the late 1970s, and that
some information may be present in the yield curve after 1984.
Information in Spreads
Although the testable implications of the expectations theory of
the term structure are rejected for two of the subsamples by the using a
VAR model, that rejection does not necessarily mean that there is no
information in the term structure. The expectations theory implies that
current spreads between interest rates at different maturities predict
future interest rate changes. This implication of the expectations
theory warrants separate examination.
Campbell and Shiller (1991) show how under the expectations
hypothesis, yield spreads can be used to predict changes in both shortand long-term interest rates. To use the hypothesis to predict short
rates, follow the approach of the previous section by subtracting R ^
from both sides of (1) and reverse sides, giving
(4)

(l/k)2j:i tRt+ni,m - R ^

=

Run-R^n

+ c, k = n/m.

The right-hand side of (4) is just the current spread between n- and •period interest rates. Equation (2) thus suggests that the difference
between the average expected m-period rate and the current m-period rate
is equal to the current spread between n- and m-period rates plus a risk
premium.
Equation (4) can be tested by regressing the realized difference
between average m-period rate and the current m-period rate, (l/k)2^;i
Rt+mMn - R^B S St(n-m)" on the current spread, R^-R^ = SfK
The expecta-




-10-

Roberds, Runkle, and Whiteman

tions theory implies that the coefficient on the current spread should
be unity. Thus, the current spread should be a good predictor of the
future average change in short-term rates.
Campbell and Shiller also consider the implications of equation
(1) for changes in future long-term rates. They note that
(5)

sj*»>

s (m/(n-m))S?^>

*

^^-K^.

This implication of the expectations hypothesis can be tested by regressing the realized value Rt+BMMIrRtta on s ^ . The expectations theory
predicts that the estimated coefficient on s ^ will be unity. Thus, a
known multiple of the current spread should be a good predictor of the
future change in long-term rates.
Campbell and Shiller test both (4) and (5) using Mcculloch's
(1990) monthly data on U.S. Treasury bill, note, and bond prices from
1952:1 to 1987:2. Their analysis is especially complete: they look at
all possible combinations of short and long maturities that are multiples of each other from one month to ten years. By conducting such an
exhaustive analysis, Campbell and Shiller are able to pinpoint those
maturity combinations for which the expectations theory of the term
structure works well, as well as those combinations for which it works
poorly.
One of the Campbell-Shiller findings is that for any two
maturities, n and a, equation (5) performs abysmally. That is, the
current spread between n- and a-period rates has no power in predicting
the difference between the (n-a)-period rate a-periods from now and the
current n-period rate. In fact, equation (5) performs so poorly that
the coefficient on s ^ is usually negative, while the expectations
theory predicts a value of one for that coefficient.
The Campbell-Shiller estimates of equation (4) are somewhat more
promising for the expectations theory. For maturities beyond three or
four years, they cannot reject the hypothesis that the coefficient on
This means that the current spread between n- and
S(iwm) ^g unity.
m-period rates predicts how the average m-period rate will change over
the next n-periods.12 But for shorter maturities, especially those
below one year, Campbell and Shiller's tests reject the hypothesis that
the coefficient on S ^ is unity. Their results are consistent with

12. Or to be more precise, it says how the average a-period interest
rate every a periods from the current period to the n-ath period will
change from the current a-period interest rate.




-11-

Roberds, Runkle, and Whiteman

earlier research by Shiller, Campbell, and Schoenholtz (1983), Fama
(1984), and Mishkin (1988).
Although the Campbell-Shiller results are useful for analyzing
the predictive power of the yield spread over the entire post-Treasuryaccord period, they do not tell us much about how different Federal
Reserve operating procedures have affected the term structure. There
are two ways in which we believe the Campbell-Shiller results must be
extended to understand the effect of different operating procedures.
First, we must examine the predictive power of yield spreads during each
different operating regime, since the amount of information contained in
the spread could differ greatly across the different regimes. Second,
we must examine the predictive power of yield spreads for each different
day of the maintenance period for the different operating regimes, since
operating procedures and volatilities vary greatly by day of the
maintenance period.
Since the Campbell-Shiller results show that there is almost no
hope for equation (5), we concentrate our efforts on equation (4). We
want to see whether differences in operating procedures can explain the
Campbell-Shiller finding that average future short-term interest rates
do not change as much from the current short term rate as the current
yield spread predicts that they will. To do this, we estimated ex post
versions of equation (4) using data consolidated according to operating
regime, as well as broken out by day of the settlement period within
each regime.13
Results for term Fed funds rates under the non-borrowed and
borrowed reserves operating regimes are presented in Table 6. In the
table, three characteristics of the term structure emerge. First, under
the current operating regime, the short end of the term structure
displays the characteristic pattern found by Campbell-Shiller for intrayear rates: the bias of the term structure forecast increases with the
maturity of both the long and short term rates. Second, under the
nonborrowed reserves targeting regime, the short end of the term
structure was substantially more informative about movements in future

13. Because we use daily data, the errors in our term-structure
regressions are serially correlated, for reasons noted by Hansen and
Hodrick (1980). We correct for both serial correlation and conditional
heteroskedasticity using the methods suggested by Hansen (1982).
Missing observations are dealt with in the following manner. We repeat
missing observations in order to calculate the forward averages on the
LHS of (4). However, these repeated observations are not used to
calculate the regression results in Tables 6-10. This procedure de
facto extends the maturity of pre-holiday short rates by one day.




-12-

Roberds, Runkle, and Whiteman

short rates. With the exception of the overnight—30-day connection,
each of the slope coefficients in the nonborrowed reserves table is
within one standard error of unity, the value predicted by the
expectations hypothesis. Third, the fraction of the variation in the
spread between average future short rates and the current short rate
which can be explained by the current long-short spread is much lower at
the shortest end of the term structure during the nonborrowed reserves
period. For example, the R2 in the spread regression using the
overnight—60-day spread was 0.45 after 1984, but only 0*16 between 1979
and 1982. However, this deterioration in quality of fit does not
characterize the longer end of the term structure—primarily because so
little of the variation is explained even in the best cases.
For the borrowed-reserves/contemporaneous-reserves accounting
regime, the results for estimating (4) using data on repo rates are very
similar to those using Fed funds data, as can be seen in Table 7. There
are only two important differences between the repo results and those
for Fed funds: First, the slope coefficients are somewhat higher using
repo data if the long-rate maturity is 90 days or less. Second, the
slope coefficients are actually negative if the long-rate maturity is
180 days and the short-rate maturity is 30 days or more.
For the results for estimating (4) using repo data are very
different from those using Fed funds data for the nonborrowed-reserve
targeting regime, as can be seen in Table 7. As noted above, the slope
coefficients in the Fed funds regressions were all within one standard
error of unity, except with a short-rate maturity of one day and a longrate maturity of 30 days. With the repo data, only two of the slope
coefficients are within two standard errors of unity.
The relatively poor performance of the term-structure regressions on repo data in the nonborrowed-reserve targeting period may well
be explained by institutional factors affecting the repo market during
this time. Before the fall of 1982, the repo market was quite immature
and contained many legal uncertainties. One indication of the repo
market's immaturity is the fact that most trades during this period were
done at 1/4 percentage point increments. Fed funds were trading on 1/16
or finer percentage point increments. Because Fed funds rates had a
higher resolution than repo rates, the term structure of repo rates
could not contain as much information as the term structure of Fed
funds.
Other institutional issues besides market immaturity may also
explain the poor performance of these regressions. Until 1982, courts
did not decide who actually owned the pledged securities if a broker




-13-

Roberds, Runkle, and Whiteman

went bankrupt. Also, pricing by dealers was not uniform until October
1982, when the New York Fed required repo pricing to be based on accrued
interest on the pledged securities. This ruling came after abuses of
alternative pricing mechanisms lead to nearly $300 million in losses to
Chase Manhattan when Drysdale Government Securities collapsed.
Since term Fed funds data are not available for the funds-rate
targeting period, Table 8 replicates some of the results in Table 6,
using observations on the overnight funds rate and the 1, 3, and 6-month
T-bill rates. The Table 8 results for the nonborrowed and borrowed
reserves regimes are generally not as favorable to the expectations
hypothesis as the analogous results in Table 6. We speculate that this
deterioration in the fit of the expectations model is driven by the
existence of a secondary market for T-bills that does not exist for term
Fed funds. The existence of a secondary market implies that the price
of T-bills should reflect the value of this "put-option" feature. It
also seems likely that the value of this feature of T-bills would
incorporate factors other than the conditional first moment of the
short-term interest rate.
The Table 8 results also differ from the analogous figures in
Table 6 in terms of the patterns displayed by some of the statistics
over the various maturities. For both the borrowed and nonborrowed
reserves subsamples, the bias of the term structure is less at 91 days
than at 28 or 182 days. However, the two subsamples still differ
substantially in terms of the bias and fit of the term structure
equations. The results for the nonborrowed reserves subsample still
dominate those for the borrowed reserves subsample in terms of bias,
i.e., the slope coefficients are closer to unity. At a horizon of
182 days, the R2 statistics are still larger for the nonborrowed
reserves period. At the shorter horizons, the R2 statistics are roughly
the same for both periods.
Results for the funds-rate targeting period are also shown in
Table 8. For the funds-rate targeting regime, the information content
of interest-rate spreads appears to be uniformly low, as evidenced by
the very low R2's obtained for equation (4). The slope coefficients are
also generally quite small, and in most cases are within a standard
error of zero. The exception is the slope coefficient on the 6-month/
3-month T-bill spread, which is greater than the analogous estimates for
the borrowed- and nonborrowed-reserves regimes, though it still falls
well within two standard errors of zero.




-14-

Roberds, Runkle, and Whiteman

Periodicity and Information
As noted above, since 1984 the overnight funds rate has
displayed a marked periodic pattern over the two-week reserve
maintenance period, while these same periodic patterns are absent for
term Fed funds rates. To investigate the effect of this periodicity on
estimates of equation (4), we estimated versions of equation (4) over
subsamples that consist of observations on particular days of the
reserve maintenance period* Representative results from this exercise
are displayed in Tables 9 and 10.
Table 9 shows that the choice of subsample makes a tremendous
difference in the fit of equation (4), when the short rate is the
overnight funds rate and the time period considered is the post-1984
borrowed-reserves/contemporaneous-reserve accounting regime. In these
cases, i.e., in the first column of Table 9, the results for settlement
Wednesdays display markedly higher values for both the estimate of the
slope coefficient and for R2. This "settlement-day" effect is more
muted or virtually nonexistent for short rates having a maturity of 30
days or more. It is also much harder to detect for the 1979-82 period,
in that the results for Wednesdays (which were all settlement days) are
extremely close to the results for the period as a whole (cf. Tables 6
and 9).
Table 10 replicates the results of Table 9 for the repo market.
The post-1984 results follow essentially the same pattern as for Fed
funds, i.e., of better fits for the overnight rates on settlement
Wednesdays, worse fits on non-settlement days, and few differences
otherwise. Note that all of the slope coefficients when comparing
overnight repos to longer-maturity repos are within two standard errors
of unity. The 1979-82 results for Wednesdays differ little from the
results over the entire 1979-82 sample (cf. Table 7).
INTERPRETATIONS AND FINDINGS
The results reported in Tables 1-10 are entirely consistent with
the idea advanced by Mankiw and Miron (1986) that the information
content of the term structure is strongly linked to the volatility of
short-term interest rates. This effect shows up in two ways in our
results. First, the estimates of the slope coefficient for equation (4)
in Tables 6-10 are generally larger for the volatile 1979-82 period than
is the case for the other subsamples. Second, both higher slope
coefficient estimates and better fits are obtained for the relatively
volatile subsample of settlement Wednesdays during the post-1984 period.
The first of these two observations should be uncontroversial, as it has




-15-

Roberds, Runkle, and Whiteman

already been reported by earlier studies, notably Hardouvelis (1988) and
Simon (1990). To our knowledge, the second observation is unique to the
present paper and thus merits additional discussion.
From the standpoint of policy and evaluation of the current Fed
operating procedure, the key question is "does the current operating
procedure allow some information about short-term interest rates to be
reflected in the term structure?" The answer to this question provided
by Tables 5-10 is an unambiguous "yes, but ... •" The post-1984 results
for equation (4) are certainly more favorable than the available 1975-79
results to the idea that rate spreads contain information about the
future course of short-term rates. The down side of this generalization
is that much of this information is of a limited, short-term nature.
Tables 9 and 10 indicate that after 1984, equation (4) fits best for a
particularly volatile subsample of our dataset, i.e., on settlement
Wednesdays when the short rate is an overnight rate.14 This pattern of
results is consistent with the widely held notion that while the Fed may
loosen its grip on the overnight funds rate on settlement Wednesdays,
the day following settlement will generally see the return of the
overnight funds rate to previous target value. Our settlement-day
results might be considered encouraging in the sense that it shows that
the markets are not "spooked" by settlement-day pressures in the
overnight Fed funds market. On the other hand, the value of this type
of information is likely to be nil at other than very short horizons.
Inspection of various entries of Tables 6-10 shows that such is
in fact the case. For example, the second column of Tables 6,7,9, and
10 shows that in the post-1984 period, the 30-day/60-day and 30-day/
90-day spreads do contain some information for future movements in
30-day rates. However, the 30-day/180-day spreads on Fed funds and
repos do not have any forecasting power for future movements in 30-day
rates. Similarly, the 60-day/180-day spreads on Fed funds and repos
never provide information on the future course of 60-day rates. The
same is true for the 90-day/180-day spreads on Fed funds and T-bills.
In the case of repos, the 90-day/180-day spread does provide some
information on future 90-day repo rates, but the sign of the slope

14. Since 1984, the overnight funds rate has also tended to be quite
volatile around year-end, due to holiday cash demand and "window-dressing" pressures.
Point estimates very similar to those obtained for
settlement Wednesdays were obtained for a post-1984 subsample consisting
of the two-week periods beginning on a Thursday and spanning the
Christmas and New Year's holidays.




-16-

Roberds, Runkle, and Whiteman

coefficient is the opposite of that predicted by the expectations
hypothesis and the amount of variation explained is quite small.
These last results suggest that despite the nominal distinctions
between the post-1984 operating procedure and the funds-rate targeting
regime of the late 1970s, relatively little information is being
captured at the short end of the term structure. We find that what
information available in the short end of the term structure vanishes at
a horizon somewhere between 90 and 180 days, a finding consistent with
the results of Hardouvelis (1988), whose data set extended only to 1985.
That is, the net position of the markets, as reflected by the term
structure, cannot be interpreted as having any predictive power beyond a
horizon of roughly 90 days.
To obtain a better idea of how this finding impacts on the
market's expectations of monetary policy, we make use of an idea
suggested by Simon (1991). Using 1983-88 data on 30- and 60-day term
Fed funds rates, Simon (1991, p.334) finds that a version of equation
(4) fits particularly well during the days immediately preceding and
following FOMC meetings. Simon interprets this finding as supporting
the notion that the policy intentions of the FOMC are quickly
transmitted to financial markets. To implement this idea for our data
set, we fit equation (4) to both Fed funds and repo data after 1984,
restricting ourselves to the days immediately following FOMC meetings
(or the second day of the meeting for 2-day meetings). These results
are displayed in Tables 11 and 12, along with the analogous results for
the 1979-82 period.
In general the post-1984 results in Tables 11 and 12 do not
differ radically from those reported in Tables 6 and 7. This is
particularly true for the Fed funds market (cf. Tables 6 and 11). For
repos, there is a somewhat better fit immediately post-FOMC for versions
of equation (4) where the short rate is the overnight rate, or where the
long rate has a horizon of 30, 60, or 90 days (cf. Tables 7 and 12). At
a horizon of 180 days, there is no improvement in fit for the equations
with a short rate having terms of 30, 60, or 90 days. These results
suggest that interest rate spreads directly attributable to policy
actions are not likely to be more informative than is usually the case,
especially at horizons beyond 90 days.
To summarize, our results indicate that in the current
(post-1984) policy environment the information implied by the short end
of the term structure vanishes at horizons beyond 90 days. This result
is consistent with the Mankiw-Miron hypothesis in the sense that the
available evidence from the 1979-82 period (which is necessarily limited




-17-

Roberds, Runkle, and Whiteman

because of the short duration of the nonborrowed-reserves operating
procedure) suggests that this was likely not the case when the Fed was
less aggressively smoothing the funds rate. The fact that some
information is contained in the post-1984 term structure for the very
short term is consistent with the Mankiw-Miron hypothesis, in contrast
to conjecture of Hardouvelis (1988, p.355). As is documented above, the
volatility of the overnight funds rate on reserve settlement days is
accompanied by an increase in the informativeness of the term structure.
Since longer-term Fed funds and repo rates are generally not subject to
the settlement day volatility, the Mankiw-Miron hypothesis would predict
that the fit of equation (4) would fall with the maturity of the short
rate. This is exactly what happens in the post-1984 sample.
Our results are also complementary to those obtained by Campbell
and Shiller (1991). Recall that Campbell and Shiller are unable to
reject the expectations hypothesis restriction on equation (4) (i.e.,
that the slope coefficient equals one) when the long rate has a maturity
greater than three years. For the post-1984 Fed funds and repo markets,
our results imply that interest rate spreads are quite informative at
very short horizons, although we can still formally reject the
expectations hypothesis, excepting the repo market on settlement
Wednesdays. Campbell and Shiller (1991, p.507) also note that for the
Treasury market, the forecasting ability of equation (4) falls with the
length of the forecasting horizon (the long rate maturity) at horizons
of less than one year. We document a similar effect in the post-1984
Fed funds and repo markets. The information content of the yield curve
in these markets begins to decline at a horizon of no more than two
months, and vanishes at six months.
CONCLUDING REMARKS
The results discussed above, together with the term structure
results obtained by Campbell and Shiller (1991), Fama (1984), and
related papers, point to a remarkable empirical regularity associated
with the recent U.S. term structure. In terms of the ability of the
term structure to predict subsequent movements in short rates via
equation (4), there is an "ozone hole" in the term structure beginning
at a horizon of roughly six months and extending out to a horizon of two
or three years. That is, the ability of the implicit forward rates to
anticipate the future course of interest rates is severely curtailed at
horizons between 3-6 months and 2-3 years. Our conjecture is that the
cause of the "ozone hole" is the Fed's historically accommodative stance
towards seasonal fluctuations in the demand for credit. At a horizon of




-18-

Roberds, Runkle, and Whiteman

roughly six months, a policy incorporating seasonal accommodation has to
come into conflict with the market-determination of short-term rates.
Put another way, no one has yet invented a seasonally adjusted credit
market•
Without a well-specified model, it is not possible to analyze
welfare implications of the results presented above. However, the
operating procedure in place since 1984 has been only partially
successful in terms of providing information to credit market
participants concerning the future course of short-term interest rates.
Further, the greatest amount of yield-curve information has been
available during episodes associated with higher volatility of the
overnight Fed funds rate. Finally, we conjecture that the Fed's
historical policy of seasonal accommodation poses an inherent
limitation, for better or for worse, on the ability of implicit forward
rates to forecast future interest rates at horizons close to one year.




-19-

Roberds, Runkle, and Whiteman

REFERENCES
Batten, Dallas S., Blackwell, Michael P., Kim, In-Su, Nocera, Simon E.,
and Ozeki, Yuzuru. "The Conduct of Monetary Policy in the Major
Industrial Countries: Instruments and Operating Procedures."
International Monetary Fund Occasional Paper No. 70 (July, 1990).
Campbell, John Y. and Shiller, Robert J. "Cointegration and Tests of
Present Value Models," Journal of Political
Economy, vol. 95
(1987), pp. 1062-88.
Campbell, John Y. and Shiller, Robert J. "Yield Spreads and Interest
Rate Movements: A Bird's Eye View," Review of Economic
Studies,
vol. 58 (1991), pp. 495-514.
Cook, Timothy and Hahn, Thomas. "Interest Rate Expectations and the
Slope of the Money Market Yield Curve," Federal Reserve Bank of
Richmond

Economic

Review,

vol. (1990), pp. 3-26.

Cook, Timothy. "Determinants of the Federal Funds Rate 1979-82."
Federal Reserve Bank of Richmond Economic Review, vol. 75 (January/February 1989), pp. 3-19.
Fama, Eugene F. "The Information in the Term Structure," Journal
Financial
Economics, vol. 13 (1984), pp. 509-28.
Friedman, Benjamin M.
Journal

of Economic

of

"Lessons on Monetary Policy from the 1980s,"
Perspectives,

vol. 2 (1988), pp. 51-72.

Goodfriend, Marvin. "Interest Rates and the Conduct of Monetary
Policy," Carnegie-Rochester Series on Public Policy 34 (Spring,
1991), pp. 7-30.
Goodfriend, Marvin. "The Promises and Pitfalls of Contemporaneous
Reserve Requirements for the Implementation of Monetary Policy."
Federal Reserve Bank of Richmond Economic Review, vol. 70 (May
1984), pp.
Goodfriend, Marvin and Whelpley, William.




in Instruments

of

the

Money Market,

-20-

"Federal Funds." Chapter 2
6th edition, edited by Timothy

Roberds, Runkle, and Whiteman

Q. Cook and Dimothy D. Rowe.
of Richmond (1986).

Richmond, Va.:

Federal Reserve Bank

Hansen, Lars Peter. "Large Sample Properties of Generalized Method of
Moments Estimators," Econometric*,
vol. 50 (1982), pp. 1029-54.
Hansen, Lars Peter and Hodrick, Robert J. "Forward Exchange Rates as
Optimal Predictors of Future Spot Rates: An Econometric
Analysis," Journal
of Political
Economy, vol. 88 (1980), pp. 82953.
Hansen, Lars Peter and Sargent, Thomas J. Rational
Expectations
Econometrics,
With contributions by John Heaton, Albert Marcet, and
William Roberds, Underground Classics in Economics, Boulder and
Oxford: Westview Press, vol. (1991), pp.
Hardouvelis, Gikas A. "The Predictive Power of the Term Structure
during Recent Monetary Regimes," Journal
of Finance,
vol. 43
(1988), pp. 339-356.
Heller, H. Robert. "Implementing Monetary Policy."
Bulletin,
vol. 74 (July, 1988), pp. 419-29.

Federal

Reserve

Hodrick, Robert J. "Dividend Yields and Expected Stock Returns:
Alternative Procedures for Inference and Measurement." National
Bureau of Economic Research Technical Working Paper No. 108 (July,
1991).
Kohn, Donald. "Policy Targets and Operating Procedures in the 1990s."
Federal
Reserve
Bulletin,
vol. 76 (January, 1990), pp. 1-7.
Lumpkin, Stephen A. "Repurchase and Reverse Repurchase Agreements," in
Cook, Timothy Q., and Rowe, Timothy D., eds, Instruments of the
Money Market, Richmond: Federal Reserve Bank of Richmond (1986).
Mankiw, N. Gregory, Miron, Jeffrey A. and Weil, David N. "The
Adjustment of Expectations to a Change in Regime: A Study of the
Founding of the Federal Reserve," American Economic Review,
vol.
77 (1987), pp. 358-374.




-21-

Roberds, Runkle, and Whiteman

Mankiw, N. Gregory and Miron, Jeffrey A. "The Changing Behavior of the
Term Structure of Interest Rates,n Quarterly Journal of Economics,
vol. 101 (1986), pp. 211-28.
McCulloch, J. Houston. "U.S. Government Term Structure Data," in
Friedman, Benjamin, and Hahn, Frank, eds., The Handbook of
Monetary Economics, Amsterdam: North Holland (1990).
Meulendyke, Anne-Marie. U.S. Monetary Policy and Financial
New York: Federal Reserve Bank of New York (1989).

Markets.

Meulendyke, Anne-Marie. "A Review of Federal Reserve Policy Targets and
Operating Guides in Recent Decades•" In Intermediate
Targets and
Indicators
for Monetary Policy:
A Critical
Survey.
New York:
Federal Reserve Bank of New York (1990).
Mishkin, Frederic S.

"The Information in the Term Structure:

Further Results," Journal

of Applied

Econometrics,

Some

vol. 3 (1988),

pp. 307-14.
Poole, William. "Interest Rates and the Conduct of Monetary Policy: A
Comment,** Carnegie-Rochester
Conference Series on Public
Policy,
vol. 34 (1991), pp. 31-41.
Sargent, Thomas J. and Wallace, Neil. "The Real-Bills Doctrine versus
the Quantity Theory: A Reconsideration," Journal of
Political
Economy, vol. 90 (1982), pp. 1212-36.
Shiller, Robert J., Campbell, John Y. and Schoenholtz, Kermit L.
"Forward Rates and Future Policy: Interpreting the Term Structure
of Interest Rates," Brookings Papers on Economic Activity,
vol.
(1983), pp.
Simon, David P. "Expectations and the Treasury Bill-Federal Funds Rate
Spread over Recent Monetary Policy Regimes," Journal of Finance,
vol. 45 (1990), pp. 467-77.
Simon, David P. "Secrecy, Signalling and the Accuracy of Expectations
during the Borrowed Reserves Operating Regime," Journal of Banking
and Finance, vol. 15 (1991), pp. 329-41.




-22-

Roberds, Runkle, and Whiteman

Stigum, Marcia. "The Repo and Reverse Markets,"
Jones-Irwin (1989).




-23-

Homewood: Dow

Roberds, Runkle, and Whiteman

1.

Data Series
Abbreviation

Series

Availability
75: 1: 2 - 91: 7:24

FF30Y
FF60Y
FF90Y
FF180Y

Overnight Effective
Federal Funds Rate
30-Day Fed Funds Rate
60-Day Fed Funds Rate
90-Day Fed Funds Rate
180-Day Fed Funds Rate

79:11:13
79:11:13
79:11:13
79:11:13

-

91:
91:
91:
91:

7:24
7:24
7:24
7:24

RPY
RP30Y
RP60Y
RP90Y
RP180Y

Overnight Repo Rate
30-Day Repo Rate
60-Day Repo Rate
90-Day Repo Rate
180-Day Repo Rate

75: 1: 2
79: 8:27
79: 8:27
79: 8:27
79:11:13

-

91:
91:
91:
91:
91:

7:24
7:24
7:24
7:24
7:24

TBI

One-month (28 Day)*
T-bill Rate
Three-month (91 Day)*
T-bill Rate
Six-month (182 Day)*
T-bill Rate

75: 1: 2 - 91: 7:24

FFEY

TB3
TB6

75: 1: 2 - 91: 7:24
75: 1: 2 - 91: 7:24

•Maturities of T-bills will fluctuate between auctions.
above are for Thursdays.




-24-

Maturities




ro

00"T>-noo"0-»iTJ-ti"0-n"0-Ti
oooo

o o o o p p

-<

Mill
I

I

I

I

I

I

I

I

I

I

•

I

«00>00>00>0>OvO^O>0>0

Srt
^J#
o

7C
C

i/»

O
0)

p
*

•

1
rt

c
II

2.

I A O
on

<i a.
-•» o o v> o>
n i T S~ ~

V)

<•
7*

II

i

i

i

•

i

•

i

i

i

i

i

•

i

•

i

i

i

i

i

•

i

i

i

oo oo oo oo oo oo oo oo oo oo oo oo

i

i

-#.

^J **J "^ "^

tr
p>
3
<Q
(0
o

o>

V

o o o o o o o o o o

o o o o o o o o o o o o

o o o o o o o o o o o o

20l§i§§lill| l~§°io§l|§l§ 1111

o o
o o o o o o o o o
• • o
O -» • 0 0 ~ » 0 - » - » - » I M W

roLLpQ-ip^o'^ww

O O O O O O O O O O O O

o o o o o o o o o o o o

CO
ID
H»
(0
O
ft
ID

oooo

iisiiiiiiiii

oooo

3SSS

o o o o o o o o

o

o o o o . o o o o . o

a

o o o-*

-»
- . _ » _ . ^ r\) ^ J - *
oooootO"_ooo40ooouir\3r\)

^8?

i

i

i

i

•

i

S
« S^OIS.

-»-»-*

-» ro-» K>-» ru UJ r\*

S»

N32.

VI - •

rt

ii 2 , ^

a
3*

oo-»w

O O-^
-» O - » - * - * 1/1 ->|
. . . o o
W9B-»« • O O O O O ^ I r J O O

ooorvi

m^jg^'§8S Sgk^kkk^«l IS^lS^SiSlgS ISgss |

& 2 55 i f
5 w

3

i

S*»«£giS552§3 S ^ k K K ^ S g gSS^gSSSsSS as§§ *1
ON-»C
5*^ o - i

n
a

n

oMl
i

cr
ID

CD

rt
Dl
ft
ID

i

5

3
ft

^&S3SS3S2SaS SttggfiiS&gJLg S g ^ S S ' s i ! ^

O OO • O • O « O O W N
. . O• O• O• • • •
^ 4

x

a

©

8*33fcS3SI5l2giB
x
9

n
H-

o o o o o o
o-»o**o*
. . . . . .
o • • • • •
-^-T* CM\J-» IVJ • U t O O U J O
O O w P> ^ u i o t / t o "" * ~

-H
>v
r>
-H

I

00

i

*

w

c

i

? oo oo oo oo oo oo oo oo oo oo oo

o

c

o
vt

i

•osssss-^
**-•
l<^-<^-<-<

So

rt - I
rt

... O
-» rt O
O » O

II
7%.

O O

«||SIilli|3i

iaiiaiiiiaii SHSHSSSH? nmmm*

a^i ?

D»
3

mrnmrnrnmrnrnmm

rnrnmmrnmmrrirfirn

m rn m m i

<
-1
»

<

rti m ftl r+i m

-<-<-<-<-<

0

IX

to
O O 00 > | ( M f l M * J N - »

oooo-^o^en-r-uiro-*

OOOOOOOOOO
l*OOOOOo'o-»0

O O O O O O O O O O
I M O O O O O O O O O

oo*^>oro**uio-^oo

0J
3

O.
tt

U1*^ U N - *

3
Q
»

O O O O O

fr

vn -!>• w r o - »

*<

O OO OO
• • • . •
tO - » 0 0 ^

oi/iuJO*otnoo-*o^i

ro en en o p
ro oo ro ^ ^

ifl^UJOO^UlUIOO^fO

S O O S N

ro

£ 0 1 / 1 0 0 0*

o i -» « j -» o

•r*

kurt

3 ~

10.281
17.485
11.463
9.725
15.696

rt

o utro o r o
en oofo o* -»

kurt

fK W J - - * 00 *"* - * 00 t> •%!
0*0*^J**00*^"OUIUl

c

00 Qv N O 0*

to 82:10: 6

gouj «-*-» ro
8.072
13.210
25.497
-0.096
2.336

X

s
o o .r* -»ro

91: 7:2

o^enuiujroooo^joo

"

09OOWN

a

rt
O

o
o
<0*^OQ0<r*U«<O<OQ0<0

->»en-grooo

a
a

1: 2 to79:10: 3

C

o
tr

75:

O0B fO^J O f O - J - * 0 - »

* * ^ w * » mro - * o * -» *» to

OCh(hVfltt

skew

7t

"

3

(0
oo oo o
* * - * - » o *»
enroroo p

1.998
1.981
0.562
1.067
2.706

C

CO
*-

o o o o - * to

<OUIV/IQO
•f^OOCKOUl

3

79:10:11

O 0 0 N W ^ - » - » W W O
. . . . . . . . . .

X

3
IP
0

skew

.
"^

-* o o o o

I

?

s

V)
K-

•

<

W M ^ N - *

&
<

Fr
«»
&

dev

->J^ J*00 -»0»O MO O *

-»roc»ro<o *-* uir* o o <*

•<-<-<-<-<

CT
(0

<
to
-i

(0
rt

to
rt
CL

0.644
0.289
0.070
0.498
0.048

_»_»_*cnroo-*->iro-»
">iboV»opouiV*ro"oo»
oen-r-ouopoo^o
oo->i<o-«j-4-r»UJoo<oui

I

O O O O O

m m m m rtl

dev

o o o o o o o o o o
en-»-»ro-»oororoen
ui**Qo-r*<ooooo**Uiro
***»^<ov/iroooo»->jen

O O O O O O O O O O
CL
U l - J IO M _ » - » - * (\J |\> U l Q .

0)
-i

3d

c

P

a
x

Hft

N N - » 0 - » 0 0 0 0 - »

3

rO**UJ-*^
^OOOWOO

O O O O - *

O C O N O S N N W N ^

<0-r*U1UJUJ

<O-V|-NIOOOOQOC*OOOO

O en enc* O
uioo ->jro ro

Ul * * *> .T» Ul

woooto

i
X

3.

2.352
0.365
0.730
0.588
1.592




3

3

-1.501
-0.740
-0.649
-0.720
-3.427

ONN0a0B00 0»00<OO
-»NNNNOON(M-*

F - « 0 9 O N N t t W O W
U1 UJ - * O* O W 00 00 - • -sj

3.934
1.511
2.565
3.001
2.717

r o o o o - » o o o o o
go-*>o-»ro-»roroQo<o
orJwro^iroui^rooJ

—

-* o r o - * o
b'f O N W

-2.342
-1.419
-2.048
-2.028
-2.991

0 0 0 > O O N U I ^ O U 1 - 0
00-»-*ONMW1WOW
en0000OO^O00-»^|

-» o o -* o o o r o -»ro
oooNob»"ouiou»^j

Ulttl

CD

-»oo-»oooro-*ro

I

roro ro

3*

%U i
0009

to

P»
3

.£>

Q)

- I -« - * -H -H

<

UlUl WW W

-I

— W O
w » < pi
8
P( t D H
ft 8 K

&

ooooo^joui.r>wro-*

o o oo->io* ui ^ w ro-*

oooooooooo
oooooooooo

O O O O O O O O O O
O O O O O O O O O O

3
B

-»-^OOOM-»NOO

3

i n .r> w ro

U1-T" W N -*

a

§o£

ooooo
ooooo

o o o o o

a

O O O O O
W O W N ?
o - * oo w

B
5

Ul #"* WfO -»

I

O O O O O
O O O O O

3
V

-*ro -*p-»
w-»
oF-» 3

rt

a
<t
<

oooooooooo

</» .

o o o o o

oooooooooo
o.
o* o o 08 o> -*j ON p* oo * s <r»
^ o # * w 5* o* o i w o r J
<

oipo ^* o *s
w w wroro

o o o o o
ro ro.r*
ro"r* iw r o
ro ui
ro
-» w
o •- * r o

rt

a

a.
<

ooooo

a.

skew

oooooooooo
uio^vnsu*^(>^sv/i
•r* o OB oo w o o ^ i o ro

oooooooooo

2.047
1.023
1.416
1.173
0.119

oooooooooo

8 H i ft»
3
3
ft ft*Q
3*8
O 8 8
3
50 PP 8 3
8 8
8 8 ft
f t ri 3*
< 8
D 8
8
*0
•< * M
— 8 I
P-O
3 8
ft*<
8
3 H
8 I
3 0
O P8 P»
P»

s

o o o o o

WO*>OtOOO-»WWUl

-0.600
-0.330
-0.290
-0.330
-0.290

o o o o o o o o o o
wLnuifo'-»wro'-»wro

rt
O

UlUl

^1

ro

UlUl

rt

O * ^ 0*00 W

oooooooooo
o o o o o o o o o o
-*rororo^»'-*ro-*fofo
vnww<r*~groo-»wro
o o o o o o o o o o

oooooooooo

o o o o o

§ roo#* owoirooo» -roo*

_» o -» - » o
o bo In o po
o o o o o

max

l/1Q\-^lfl-»^O-»000t
O O O O O O O O O O

0.230
0.550
0.580
0.650
0.350

O->J-<<JOOOOOOOOOOOO

O N N O Q0O0 OKQ* OOO OT

€h \jn en 0* ON
en oo -viro ro

w o o o o o

9>

^ r o w oo oo




ro

I
ff

79:10: 3

roo-soow*»o-»

o u i o o ro

rt
O

kurt

a

O

rt
O

17.242
6.486
9.541
10.150
5.159

rr

rt
O

—ft

82:10: 6

ro
•r>

*>* o o - » o - * - * oo r o r o

o o o o o
V»o u t o o

-0.890
-1.400
-1.230
-0.740
-0.600

oooooooooo
ro-»-»ro-»-»-*-ftro-»
oo^jwwoioo-*-^*"
o o o o o o o o o o

rt

c

ro

kurt

-» W W ^ -T* -*J * * 0 #- O
WU1-»-»NO^NO-*

o* * * o - * o o o - » O K o - *
_»ro-»Ot-*-»00U4^4Ot

ro

1.358
2.613
1.174
1.111
2.682

rvi _* ro -» -» o - * o ro w ?c
-» oo o -r* o * * o - * v i - * - i

*-

84: 2: 1

o
o

-» O - * O O O O - » O O </>

• ••

10:11

„
.
£
*^

skew

JC
rt
K

0.289
0.269
0.207
0.064
0.694

w .r* o* ro * * o ro .r*-» o*
OKoi - o - * o ^ i r o * * o o *
u i o i - g . r * - » . r > - > j - * ^ j 3s

8 *J
M 8
P-ft
0 8

a-

8
h

a

8
3

a

09

•<
a

41

pft

8
3

Roberds, Runkle, and Whiteman

5.

VAR Tests of Expectations Model of the Term Structure

Data
Sample

Long
Rate

NO. Of
Lags(L)

Short
Rate

Wald Test of
Expectations
Model
t-X^L)]
(P-value)

75-79

TB3

FFEY

21

76.1
(0.001)

75-79

TB3

FFEY

42

158.0
(0.00)

79-82

TB3

FFEY

21

48.6
(.224)

84-91

TB3

FFEY

21

65.0
(.0129)




-28-

Roberds, Runkle, and Whiteman

Slope Coefficients in Term Structure Regressions! Fed Funds Market
Data Sample: 84: 2: 2 - 91: 7:24

irr
Overnight

n=

30-Day

30-Day

0.71084
(0.87272E-01)
[0.43868715]

60-Day

0.69768
(0.68365E-01)
[0.45445806]

0.59242
(0.98340E-01)
[0.15240927]

90-Day

0.63523
(0.83047E-01)
[0.35558630]

0.39346
(0.14370)
[0.065956]

180-Day

0.53355
(0.13379)
[0.19241131]

0.21209
(0.28217)
[0.01516]

60-Day

0.10545
(0.30458)
[0.00284]

90-Day

-0.14113
(0.60790)
[0.00159]

Data Sample: 79:10:11 - 82:10: 6
m*
Overnight

n=

30-Day

30-Day

0.66258
(0.13481)
[0.18600642]

60-Day

0.80027
(0.24702)
[0.15789838]

0.69623
(0.53894)
[0.02302]

90-Day

0.88427
(0.25799)
[0.15622411]

0.77503
(0.59434)
[0.03277]

180-Day

0.87620
(0.19714)
[0.15993267]

0.98028
(0.24848)
[0.10281]

60-Day

0.97080
(0.33589)
[0.08755]

90-Day

1.3203
(0.26700)
[0.08581]

15. Note: m=Short Rate Maturity, n=Long Rate Maturity. Standard errors of slope coefficients in
parentheses. R2's in brackets.




-29-

Roberds, Runkle, and Whiteman

Slope Coefficients in Term Structure Regressions, Repo Market
Data Sample: 84: 2: 2 - 91: 7:24
m=
n=

Overnight

30-Day

30-Day

0.81293
(0.76210E-01)
[0.47094]

60-Day

0.80306
(0.83843E-01)
[0.51417]

0.63056
(0.10172)
[0.15022]

90-Day

0.74918
(0.10663)
[0.42462]

0.58669
(0.13858)
[0.11086]

180-Day

0.53700
(0.22351)
[0.16325]

-0.31357E-01
(0.34781)
[0.23165E-03]

60-Day

-0.37417
(0.31626)
[0.25596E-01]

90-Day

-1.2195
(0.44449)
[0.11419]

Data Sample: 79:10:11 - 82:10: 6
m*
n=

Overnight

30-Day

30-Day

0.73162
(0.76086E-01)
[0.20980]

60-Day

0.55037
(0.10406)
[0.90085E-01]

0.34580
(0.37397)
[0.79651E-02]

90-Day

0.41506
(0.19693)
[0.44481E-01]

0.33395
(0.48853)
[0.59293E-02]

180-Day

0.15139
(0.29643)
[0.62615E-02]

0.34026
(0.24751)
[0.13233E-01]




-30-

60-Day

0.44142
(0.26959)
[0.18373E-01]

90-Day

0.56562
(0.22936)
[0.16460E--01]

Roberds, Runkle, and Whiteman

Slope Coefficients in Term Structure Regressions, T-Bill Market
Data Sample: 84: 2: 2 - 91: 7:24
m=
n=

Overnight
(SR=Fedfunds)

28-Day

0.37178
(0.11692)
[0.25126]

91-Day

0.67718
(0.13641)
[0.35458]

182-Day

0.47338E-01
(0.66608E-01)
[0.01087173]

91-Day

-0.83553E-02
(0.28326E-01)
[0.00135683]

Data Sample: 79:10:11 - 82:10: 6
m*
n=

Overnight
(SR==Fedf unds)

28-Day

0.54902
(0.84179E-01)
[0.29063]

91-Day

1.2083
(0.24637)
[0.41399]

182-Day

0.57291
(0.17944)
[0.56781]




91-Day

0.18989
(0.11866)
[0.16096]

-31-

Roberds, Runkle, and Whiteman

8.

Continued
Data Sample: 75: 1: 2 - 7 9 : 1 0 : 3
m«

n=

Overnight
(SR=Fedfunds)

28-Day

.22336
(.65927E-01)
[.97948E-01]

91-Day

.39235E-01
(.21011)
[.015899E-02]

182-Day

-.74599E-01
(.38467)
[.26626E-02]




91-Day

.43495
(.37102)
[.049477E-01]

-32-

Roberds, Runkle, and Whiteman

Slope Coefficients in Term Structure Regressions, Fed Funds Market
Data Sample: 84: 2: 2 - 91: 7:24, Settlement Wednesdays

m=
Overnight

n=

30-Day

30-Day

0.85147
(0.36168E-01)
[0.87987]

60-Day

0.84914
(0.51974E-01)
[0.81514]

0.75958
(0.13590)
[0.17065]

90-Day

0.81140
(0.64996E-01)
[0.73914]

0.29532
(0.13990)
[0.32129E-01]

180-Day

0.79142
(0.69310E-01)
[0.57071]

0.15565
(0.18608)
[0.82124E-02]

60-Day

0.13821
(0.20262)
[0.54109E-02]

90-Day

-0.29710
(0.36759)
[0.64982E-02]

Data Sample: 84: 2: 2 - 91: 7:24, Wednesdays before Settlement
m=
n=

Overnight

30-Day

30-Day

0.43032
(0.86797E-•01)
[0.16910]

60-Day

0.43615
(0.11433)
[0.10329]

0.56261
(0.70605E-01)
[0.16686]

90-Day

0.31526
(0.10734)
[0.51279E-•01]

0.41008
(0.93854E-01)
[0.79219E-01]

180-Day

0.17994
(0.12682)
[0.15009E--01]

0.12697
(0.17736)
[0.51560E-02]




-33-

60-Day

0.10924
(0.21018)
[0.31432E-02]

90-Day

-0.16026
(0.45797)
[0.20927E-02

Roberds, Runkle, and Whiteman

Continued
Data Sample: 79:10:11 - 82:10: 6, Wednesdays
m«
n=

Overnight

30-Day

30-Day

0.73740
(0.76320E-01)
[0.39902]

60-Day

0.79166
(0.12191)
[0.25797]

1.1530
(0.40904)
[0.70105E-01]

90-Day

0.86129
(0.13293)
[0.23771]

0.85878
(0.49648)
[0.45068E-01]

180-Day

0.93573
(0.20738)
[0.23866]

0.98905
(0.41132)
[0.10274]




-34-

60-Day

1.0780
(0.41318)
[0.10083]

90-Day

1.4088
(0.53714)
[0.87880E-01]

Roberds, Runkle, and Whiteman

10.

Slope Coefficients in Term Structure Regressions, Repo Market
Data Sample: 84: 2: 2 - 91: 7:24, Settlement Wednesdays

m*
Overnight

n=

30-Day

30-Day

0.88000
(0.41332E-01)
[0.91237]

60-Day

0.89058
(0.59263E-01)
[0.86838]

0.81124
(0.14040)
[0.17430]

90-Day

0.86816
(0.73338E-01)
[0.80642]

0.63977
(0.12307)
[0.11341]

180-Day

0.82090
(0.99945E-01)
[0.59916]

-0.39116E-03
(0.30401)
[0.31330E-07]

60-Day

-0.35666
(0.35431)
[0.22184E-01]

90-Day

-1.1811
(0.42519)
[0.10827]

Data Sample: 84: 2: 2 - 91: 7:24, Wednesdays before Settlement
m=
n=

Overnight

30-Day

30-Day

0.57175
(0.90018E- 01)
[0.22958]

60-Day

0.57329
(0.12672)
[0.16959]

0.94002
(0.10909)
[0.17167]

90-Day

0.52235
(0.11224)
[0.13487]

0.59400
(0.10046)
[0.14026]

180-Day

0.19879
(0.16344)
[0.17115E-•01]

0.56861E-01
(0.24290)
[0.85628E-03]




-35-

60-Day

-0.27002
(0.29575)
[0.13148E-01]

90-Day

-1.1915
(0.42815)
[0.98355E-01

Roberds, Runkle, and Whiteman

10.

Continued
Data Sample: 79:10:11 - 82:10: 6, Wednesdays

m»
n=

Overnight

30-Day

30-Day

0.77928
(0.88908E-01)
[0.27686]

60-Day

0.54894
(0.17192)
[0.99390E-01]

0.36621
(0.46336)
[0.69809E-02]

90-Day

0.48640
(0.24359)
[0.64508E-01]

0.53541
(0.57689)
[0.15607E-01]

180-Day

0.28259
(0.29999)
[0.24219E-01]

0.24637
(0.44599)
[0.73326E-02]




-36-

60-Day

0.32061
(0.45207)
[0.10550E-01]

90-Day

Roberds, Runkle, and Whiteman

11.

Slope Coefficients in Term Structure Regressions, Fed Funds Market
Data Sample: 84: 2 : 2 - 91: 7:24, Day after FOMC meetings

m=
n=

Overnight

30-Day

30-Day

0.74509
(0.83258E-01)
[0.55146]

60-Day

Q.71413
(0.10129)
[0.53943]

0.71192
(0.17200)
[0.28684]

90-Day

0.72288
(0.10569)
[0.46611]

0.41039
(0.16886)
[0.76046E-01]

180-Day

0.70486
(0.14008)
[0.30932]

0.86951E-01
(0.18779)
[0.29287E-02]

60-Day

-0.20294E-01
(0.27085)
[0.10258E-03]

90-Day

-0.31490
(0.45530)
[0.77648E-02]

Data Sample: 79:10:11 - 82:10: 6, Day after FOMC meetings
m=
n=

Overnight

30-Day

30-Day

1.1375
(0.12522)
[0.55483]

60-Day

1.1084
(0.28921)
[0.32927]

0.42279
(0.74265)
[0.73300E-02]

90-Day

1.0367
(0.31461)
[0.27301]

0.60534
(0.89888)
[0.16951E-01]

180-Day

0.94613
(0.26896)
[0.28244]

1.0098
(0.41959)
[0.12071]




-37-

60-Day

1.0853
(0.29268)
[0.18424]

90-Day

1.2601
(0.48919)
[0.10950]

Roberds, Runkle, and Whiteman

12.

Slope Coefficients in Term Structure Regressions, Repo Market
Data Sample: 84: 2: 2 - 91: 7:24, Day after FOMC meetings

m=
n=

Overnight

30-Day

30-Day

0.58366
(0.10163)
[0.32624]

60-Day

0.69122
(0.90822E-01)
[0.47256]

0.94945
(0.15035)
[0.33104]

90-Day

0.69851
(0.10085)
[0.40754]

0.81131
(0.15859)
[0.19485]

180-Day

0.60386
(0.14340)
[0.19960]

0.46057E-01
(0.32509)
[0.47395E-03]

60-Day

-0.44065
(0.38018)
[0.31240E-01]

90-Day

-1.4686
(0.47491)
[0.14990]

Data Sample: 79:10:11 - 82:10: 6, Day after FOMC meetings
m=
n=

Overnight

30-Day

30-Day

1.1657
(0.19487)
[0.56774]

60-Day

0.84924
(0.32370)
[0.24901]

0.31677
(0.72952)
[0.49330E-02]

90-Day

0.75639
(0.38906)
[0.17021]

0.38240
(0.91514)
[0.61153E-02]

180-Day

0.39520
(0.41207)
[0.58096E-01]

0.38272
(0.52590)
[0.15331E-01]




-38-

60-Day

0.35125
(0.62701)
[0.96151E-02]

90-Day

0.66613
(0.36359)
[0.36161E-01]

Roberds, Runkle, and Whiteman

1. Distribution of daily changes in effective Federal Funds rate, by
operating regime.




-39-

Roberds, Runkle, and Whiteman

2. Distribution of daily changes in effective Federal Funds rate
during the Fed Funds targeting period, by day of the settlement
period (settlement day = day 5.)

-0.5

0
hundreds of basis points

0.5

day 5

3. Distribution of daily changes in effective Federal Funds rate
on settlement day during the nonborrowed reserves targeting
period.




-40-




Roberds, Runkle, and Whiteman

4. Distribution of daily changes in effective Federal Funds
rate during the borrowed reserves—lagged accounting regime,
by day of the settlement period (settlement day = day 5.)

-0.5

0
hundreds of basis points
day 5

day 10

5. Distribution of* daily changes in effective Federal Funds
rate during the borrowed reserves—contemporaneous accounting
regime, by day of the settlement period (settlement day = day
10.)
-41-

FOMC OPERATING PROCEDURES AND THE TERM STRUCTURE:
SOME COMMENTS
Glenn D. Rudebusch1
This discussion is divided into two parts. The first describes some
recent empirical results regarding the amount of information in the
yield curve for forecasting future changes in short rates. My goal
is to highlight several of the results of Roberts, Runkle, and
Whiteman (1992) (henceforth RRW) and to compare their research to
work done by previous authors. The second section contains an
interpretation of this entire body of evidence in light of several
characteristics of the Federal Reserve's monetary policy operating
procedure.
FACTS
RRW provide a careful study of the Rational Expectations Hypothesis
(REH) of the term structure using data at the short end of the
maturity spectrum. Their paper contains much fascinating empirical
detail, but this note focuses on a single issue: the predictive
power of the term structure for future movements in interest rates.
The REH of the term structure implies that the current
spread between a long rate and a short rate should predict future
changes in that short rate. I consider two special cases of such
term structure predictions. First, I examine pairs of securities in
which the maturity of the long-term debt instrument is twice that of
the short-term debt instrument. Second, I consider the ability of
the spread between the overnight rate and the one-month rate to
predict future changes in the overnight rate.
Let r(l) be the yield on a one-period bond, and let r(2)
be the yield on a two-period bond. Then, the expectations
hypothesis implies that

1. Glenn D. Rudebusch is on the staff of the Board of Governors of
the Federal Reserve System, Division of Monetary Affairs.
2. In addition, in a separate appendix written after the
conference, I provide a formal model of the propagation of changes in
the federal funds rate out the yield curve in order to address some of
the comments and questions that were generated by the results in RRW
and several of the other conference papers.




Rudebusch

r(2) t « l/2[r(l)t + E t r(l) t+1 3 + X;

(1)

that is, the current two-period yield equals the average of the
actual and expected one-period yields in sequence plus a term
premium T. Assuming rational expectations,
r(l)t+1 = V ( 1 ) t + 1

(2)

+ u

t+l«

where u +- is a forecast error orthogonal to information available
at time t. Substituting (2) into (1) and rearranging provides a
simple testable equation of the REH of the term structure:
l/2[r(l) t+1 - r(l)t] = a + p[r(2) t - r(l)t] + e t + 1 .

(3)

Under the REH null hypothesis, p = 1 and a = -T; that is, after
taking expectations of both sides of (3), one-half the optimal
forecast of the change in the short rate should equal the spread
between the long rate and short rate (minus a term premium). In
addition, the error term is orthogonal to the right-hand side
regressors, so ordinary least squares provides consistent
coefficient estimates.
Studies that have tested the REH using equation 3 have
A

obtained a wide variety of estimates of p. These p's are often
significantly less than one; of particular note, however, is the
dependence of the estimates on the maturity of the debt instruments
being examined. Figure 1 provides estimates of P from eight studies
prior to RRW that use data on yields of U.S. Treasury debt. The
point estimates are arrayed as a function of the short (one-period)
bond maturity, which ranges from two weeks to five years.
Each
observation is numbered according to its source. For example, the
"l"'s in figure 1 are taken from Campbell and Shiller (1991), who
3. This corresponds to equation 4 in RRW with n = 2m.
A

4. For example, the P's shown at the three-month maturity are
obtained from a regression of the change in the yield of the threemonth bill on the yield spread between the six-month and three-month
bills. Although equation (3) is a convenient form for expressing
results across a range of maturities, it does not describe the
frequency of observation. Typically, empirical studies have used
overlapping observations that are more frequent than those separated
by the maturity of the short bond.



-2-

Rudebusch

provide a careful, exhaustive study spanning a large range of
maturities. Based on this collection of point estimates from
previous researchers, the shaded band in figure 1 provides an
A

informal summary of the relationship between the P's and the
maturity of the short bond. Apparently, the forecast power of the
term structure for changes in short rates is quite high for forecast
horizons (i.e., shorter bond maturities) no longer than one month.
As the horizon increases, forecast power initially disappears, as
estimates of p fall essentially to zero over the range from three
months to one year; however, with horizons longer than one year,
forecast power starts to improve. The result is, according to
Campbell and Shiller (1991), a "U-shaped" pattern of coefficients.
The evidence of RRW is consistent with this U-shaped
pattern for the short maturities that they investigate. For
example, in their table 8, based on the spread between 3-month and
6-month Treasury bills, the estimates of P are not significantly
different from zero. Also, based on the spread between the 30-day
and 60-day term federal funds rates, RRW (table 6) report estimates
of p equal to about 0.6 at the one-month horizon. Again, the
predictive content of the yield curve, while substantial at very
short maturities, appears to vanish at a forecast horizon of about
three months.
RRW also analyze the forecasting ability of yield spreads
that involve the overnight federal funds rate. As above, the basic
insight of the REH is that if the yield curve is steeply sloped,
future short rates should on average be above the current short
rate. Let the length of a period be a day, so r(l) is the
overnight federal funds rate and r(30) is the yield on a 30-day
bill; then the expectations hypothesis implies that
(4)

r(30)t

= l/30[r(l)t

+ Et

29
L r(l)

t + i

]

+ T.

Assuming rational expectations, the analog to equation 3 is

5. Or, in their words, consistent with this "hole in the ozone
layer," a metaphor that carries the gratuitous connotation of welfare
loss .




-3-

Rudebusch

(5)

29
29
l/30[ Z r(l).,,] - r(l). - a + p[r(30). - r(l).] + I e..,.
i-0
i-1

Under the REH null hypothesis, p = 1; that is, the deviation
of today's federal funds rate from its expected average level over
the next month should equal the spread between the current 30-day
A

and one-day rates (minus a term premium). RRW find P to be fairly
close to one. In their table 6, for example, using term federal
funds data, they estimate p to be around 0.7. This high predictive
power at short horizons generalizes to spreads between the overnight
rate and the 60-day and 180-day bill yields and is broadly
consistent with earlier work by Simon (1990).
In summary, I characterize the evidence on the forecasting
ability of the yield spread with four propositions:
51. Spreads between the overnight federal funds rate and
one-month, two-month, and three-month yields are very good
predictors of the change from the current daily rate to the average
daily rate that prevails over the relevant time period.
52. Spreads between short-term bills--for example, 30-day
and 60-day Treasury bills--are good predictors of the change in
the short bill yield.
53. Spreads involving longer bill rates, say, the 3-month,
6-month, and 12-month yields, have essentially no predictive content
for future changes in these bill rates.
54. Spreads involving medium and long maturity bonds-specifically, for maturities longer than two years--do appear to
have some predictive content for movements in future interest rates.
INTERPRETATION
Simply put, propositions SI through S4 indicate that the yield curve
is useful for forecasting future interest rates, but only at certain
maturities. Recently, several authors have linked this finding to
the behavior of the Federal Reserve.
Specifically, they have
asserted that the procedure by which the Fed controls the federal

6. RRW provide evidence consistent with the first three of these
propositions.
7. See Mankiw and Miron (1986), Cook and Hahn (1989, 1991), and
Goodfriend (1991) .




-4-

Rudebusch

funds rate is responsible for the varying predictive power of the
yield curve. This section clarifies and extends this argument.
First, it is useful to describe the attributes of the Fed's
operating procedure that are crucial for understanding the above
results regarding the yield curve. Although the Fed's operating
procedure has changed greatly over the past two decades--from direct
federal funds rate targeting in the 1970s to indirect targeting
through reserves in the 1980s --following Cook (1989) and Goodfriend
(1991), I take as given that over this period the Fed has always
taken an active interest in controlling the federal funds rate.
Although the exact mechanism has changed over time, the following
three attributes have characterized the Fed's underlying procedure
for controlling interest rates:
Fl. Transitory deviations from target are allowed on a
daily basis. That is, the federal funds rate is not pegged to a
target on an hourly or even a daily basis. Indeed, the Fed
generally enters the federal funds market only once each day, in the
late morning; thus, the intraday spot federal funds rate can display
wide fluctuations. Furthermore, historically there has often been a
target band for the federal funds rate rather than a single target
value; using a band also allows transitory deviations from a mean
value (usually the target band midpoint).
F2. Targets are adjusted not continuously but in limited
amounts at discrete intervals. Rather than immediately adjusting
the target in response to each new piece of information, the Fed
behaves as if it has a threshold whereby only after sufficient
information has accumulated will a target change be triggered.
Also, targets are usually adjusted by only 25 to 50 basis points at
a time; thus, when new information requires a larger change, the Fed
implements a series of smaller target adjustments that are separated
by several days to a couple of weeks.
.F3. Except when a quick succession of target changes is
needed to make a large adjustment (as noted in F2), the target is
set at a level the Fed expects to be able to maintain.
These three attributes are apparent in figure 2, which shows
federal funds rate targets from an illustrative episode of direct
interest rate targeting. The targets are shown as a solid line; the
actual daily effective federal funds rate appears as a dashed




-5-

Rudebusch

line.
Attribute Fl is apparent in the large but temporary
deviations of the daily rate from the target. F2 is reflected in
the "step-function" appearance of the target. Finally, the
infrequent occurrence of target changes, shown by long steps or
treads of many months duration, provides support for F3.
The three attributes of the Fed's "interest rate smoothing"
operating procedure can explain the term structure results SI
through S3. First, the transitory daily deviations described by Fl,
coupled with the target persistence of F3, imply substantial
predictable variation in the overnight rate. If today's rate is
unusually high, perhaps because the day is a settlement Wednesday
with strong reserve pressures, tomorrow's rate (and future daily
rates) will likely return to the target level. This occurrence
explains why the spread between the 30-day term federal funds rate
(which should be close to the current target rate) and the 1-day
rate (which is transitorily high) moves with the spread between the
average 1-day rate that will prevail over the next month (which is
also close to the target rate) and the current 1-day rate. RRW
provide confirming evidence for this interpretation: Their table 3
indicates that large changes on Wednesdays tend to be reversed on
Thursdays, and their tables 9 and 10 indicate that SI holds best for
9
these volatile settlement Wednesdays.
The predictive information described by S2, which is
available at the very short end of the term structure, can be
explained by the discrete nature of policy changes, attribute F2.
Let us suppose that a significant piece of new information arrives-say, in the form of news about some macroeconomic variable--that
clearly requires a major policy change. If the Fed accomplishes
this policy change by a series of moderate target adjustments
conducted, say, over a three-week span, then the gap between the
time that new information influencing policy was released and the
time that the full policy action is finished generates predictable

8. The target series was obtained by linking the expected federal
funds rates given in the Federal Reserve Bank of New York's weekly
"Report of Open Market Operations and Money Market Conditions" during
the period.
9. Cook and Hahn (1989, footnote 6) also describe similar evidence.




-6-

Rudebusch

changes in interest rates at horizons of less than one month.
The lack of predictive information in the 3-month to 1-year
maturity range of the term structure, noted in S3, reflects
attribute F3. As Mankiw and Miron (1986) argued, if market
participants (rationally) expect the Fed to maintain the current
federal funds rate target, then current spreads will have no
predictive power for actual future changes in interest rates.
Thus, attributes Fl through F3 of the Fed's operating
procedure appear to be responsible for yield spread results SI
through S3. Further support for the reasoning linking F2 with S2
and F3 with S3 is provided by the analysis of Cook and Hahn (1989).
They found that the 3-month, 6-month, and 12-month bill rates all
moved on average about 50 basis points in response to a change of
1 percentage point in the federal funds rate target during 1974-79.
That bill rates move by only about half of the target change
suggests that target changes are forecastable to some extent (as
implied by F2). However, that all three bill rates move by about
the same amount means that new information about the federal funds
target has little effect on the slope of the yield curve from
3 months to 12 months. This finding confirms the notion that the
new federal funds rate target is expected to be maintained over that
period.
Finally, for completeness, let me examine S4, the
proposition that spreads between long rates contain forecasting
power. Since Fama and Bliss (1987), this proposition has been
reduced to the issue of whether interest rates display a slow

10. Cook and Hahn (1990) stress a different, but related, aspect of
F2 as an explanation of S2. They note that the information threshold
required for a target change implies that news that suggests an
imminent policy action is often available to the market. For example,
if weak employment data that suggest the need for a policy easing are
released, the Fed may wait a week for another reading on inflation
before acting. This gap could also generate predictable target
changes over a very short period.
11. Cook and Hahn (1989) show that this same reasoning holds if the
funds target is expected to change in the near future (consistent with
A2) and then to persist at its new level. Strictly speaking, if one
assumes a constant term premium along with the persistence of targets,
A

the spread would be constant and the standard error of |J would be
infinite. However, as shown in Mankiw and Miron (1986), with a timeA

varying term premium, the P is biased toward zero.




-7-

Rudebusch

reversion to mean at long horizons. The subsequent debate is
summarized in Shea (1992). My own view on such issues, expressed in
Rudebusch (1992, 1993), is that conclusions about the stationarity
or nonstationarity of yields are very tenuous given the size of the
available samples. However, although such deep conclusions cannot
be made with any degree of certainty, at a practical level, S4
probably reflects the fact that over the sample period under
consideration, markets have correctly anticipated that the Fed would
be able to restrain inflation to moderate levels at business-cycle
12
frequencies.
Coupled with a stationary real rate, the Fed's
containment of inflation has probably generated the predictive power
contained in long-maturity nominal interest rate spreads.
CONCLUSION
The explanation that RRW propose for linking Fed operating
procedure to the empirical results on the predictive power of the
term structure hinges on the Fed's elimination of seasonal
fluctuations in interest rates, which would be predictable and which
presumably would be reflected in spreads (see Hardouvelis (1988)).
This interpretation appears flawed. The Fed's elimination of
weekly, monthly, quarterly, and half-yearly seasonals would imply
that figure 1 was a step function, displaying no predictive power up
to six months and significant predictive power from twelve months
and beyond. This implication does not accord with the U-shaped
curve from the empirical evidence.
Instead, I have provided an interpretation of the term
structure results that relies on characteristics of the Fed's
interest rate targeting operating procedure. A consideration of the
normative value of this procedure would also be of interest.
Goodfriend (1991) conjectures that the smoothing characteristics of
the operating procedure facilitate the communication of Fed
intentions to financial markets. Further research on this issue
would be illuminating.

12. Again, the evidence in Cook and Hahn (1989) is instructive.
They find that the 20-year bond yield responds by only about 10 basis
points to a 1 percentage point change in the federal funds target.
This is consistent with slow mean reversion at long horizons.




-8-

Rudebusch

REFERENCES
Campbell, John Y. and Shiller, Robert J. "Yield Spreads and
Interest Rate Movements: A Bird's Eye View," Review of
Economic Studies, 58 (1991), 495-514.
Cook, Timothy.

"Determinants of the Federal Funds Rate: 1979-1982,"

Federal Reserve Bank of Richmond Economic Review, (1989), 3-19.
Cook, Timothy and Hahn, Thomas. "The Effect of Changes in the
Federal Funds Rate Target on Market Interest Rates in the
1970s," Journal of Monetary Economics, 24 (1989), 331-351.
Cook, Timothy and Hahn, Thomas. "Interest Rate Expectations and the
Slope of the Money Market Yield Cure," Federal Reserve Bank of
Richmond Economic Review, (1990), 3-26.
Fama, Eugene F.

"The Information in the Term Structure," Journal of

Financial Economics, 13 (1984), 509-528.
Fama, Eugene F. and Bliss, R.R. "The Information in Long-Maturity
Forward Rates," American Economic Review, 77 (1987), 680-692.
Goodfriend, Marvin. "Interest Rates and the Conduct of Monetary
Policy," Carnegie-Rochester Series on Public Policy 34 (Spring,
1991), 7-30.
Hardouvelis, Gikas A. "The Predictive Power of the Term Structure
during Recent Monetary Regimes," Journal of Finance, 43
(1988), 339-356.
Mankiw, N. Gregory and Miron, Jeffrey A. "The Changing Behavior of
the Term Structure of Interest Rates," Quarterly Journal of
Economics, 101 (1986), 211-28.
Mishkin, Frederic S. "The Information in the Term Structure: Some
Further Results," Journal of Applied Econometrics, 3 (1988),
307-14




-9-

Figure 2
Federal Funds Rate

6.5

Target
Actual

6.0

c
5.5
XT

c
a
n
tr
5.0

4.5

4.0

Sep
1976




Oct

Nov

Dec

Jan

Feb

Mar

Apr

May

Jun

Jul
1977

Aug

APPENDIX
PROPAGATION OF FEDERAL FUNDS RATE CHANGES TO LONGER-TERM
RATES UNDER ALTERNATIVE POLICY REGIMES
Glenn D. Rudebusch1
My comments above on Roberds, Runkle, and Whiteman (1992, henceforth
RRW) focused on the relationship between FOMC operating procedures
and the amount of information in the yield curve for forecasting
future changes in short rates. However, several papers at the
conference, including RRW, touched on the general issue of the
propagation of changes in the short rate out the yield curve: most
notably, Gagnon and Tryon (1992), Goodfriend (1992), Hess, Small,
Brayton (1992), and Kasman (1992). Discussions both at the
conference and subsequently tried to discern from this array of
evidence whether adopting an alternative operating procedure that
changed the variability of the funds rate might also affect the
variability of long rates. In this appendix, I develop a model that
links movements in long rates to policy and non-policy federal funds
rate shocks in order to interpret some of the findings in earlier
papers.
In particular, my analysis clarifies the information content
of RRW's evidence on the nature of the linkage between short and
long rates. I show that the degree to which changes in the funds
rate are transmitted to changes in long rates is unrelated to the
predictive power of the yield spread. The model also provides a
clear interpretation of changes in measures of interest rate
volatility and correlation. Such changes are evident in the
narrative history of Goodfriend (1992) as well as in the earlier
evidence of Johnson (1981) .
A MODEL OF SHORT AND LONG RATES
This section describes a simple theoretical structure that
links Federal Reserve actions to movements in the funds rate and in
longer rates. There are two crucial elements of the model: First,

1. Glenn D. Rudebusch is on the staff of the Board of Governors of
the Federal Reserve System, Division of Monetary Affairs.




Rudebusch

from equation 1.

Substituting the first result into the second and
4
taking expectations yields

E r ., *= E v(r + £ u , .+ £ ,,J )
*t t+k
t t i . 1 t+i
t+k
(4)

- rt.

That is, the expected funds rate at any point in the future is equal
to today's target.

This result can be used in equation 3 to solve

for the current long rate
t-1
R. - (l/x)r. + (l/x)£. + (l/x) I r t
x
x
x
i=El x
(5)

+ 9.
x

- r t + (l/x)£t + 6 t .

The current long rate is thus equal to the current target plus a
small fraction l/x of the current transitory funds rate shock plus a
term premium.
To express the long rate in terms of the shocks affecting
the short rate, equation 5 can be rewritten as
R

- r

, + u

+ (l/x)e

+ 9 .

It is the policy shocks (u ) affecting r
R

because the term (l/x)£

unrelated to r .

that will be reflected in

is negligible and the term premium 8

is

Accordingly, the amount of short-rate variation

that is transmitted to the long rate will depend crucially on the
size of G .
u

4. This expression assumes that financial market participants can
discern the current target rate on any given day; that is, they know
e t and u at time t. Thus, in this model, there is no avenue for
misperceptions of monetary policy, arising, say, from the adoption of
new operating procedures, to affect the relationship between the funds
rate and longer rates. A useful extension to the model would be to
re-examine the results assuming market participants had difficulty
distinguishing transitory daily fluctuations in the funds rate from
true policy shifts.




-4-

Rudebusch

INTERPRETING RRW'S HISTORICAL EVIDENCE
The results of RRW indicated a clear difference across the
pre-1979 and the 1979-1982 periods in the ability of the bill-funds
yield spread to predict future changes in the funds rate. Their
historical evidence was obtained by regressing actual changes in
future short rates on the current spread between the funds rate and
longer rates.
In terms of the model outlined above, RRW's key
regression can be obtained by subtracting r from both sides of (3),
which after rearrangement gives
X-l

(6)

(i/x)az V t + i ' c ^ - i ) / * ) ^ "

(R
t

" rt} ' V

Under the assumption of rational expectations,
(7)

Vt+i

= r

t+i

+ v

t+i-

where v + . is a forecast error orthogonal to information at time t.
Then (6) can be rewritten as the regression
X-l
(8)

(1/X)BI

r t + i - ((x-l)/x)rt - P(Rt - rt) + e t .

The dependent variable on the left-hand side requires the
X-l
construction of (1/x) I r++- * which is simply the average of funds
i-1 x x
rates observed after time t, and the use of (7) implies that P
equals one.
X-l
The regression error, e = - 8 - L v ,., is correlated
X

X

i = 1

T-M

with R through the common term premium but is not correlated with
r . Because of the correlation between the regression error and the
X

A

regressor, the OLS estimate of p, denoted p, will be biased downward
A

from one. Still, the size of P has been interpreted, as in RRW, as
measuring the ability of long rates to forecast future movements in
short rates.
A

Given the model for the funds rate, the value of P (in the
population) can be easily determined. By definition,
5. This regression is essentially equation 4 in RRW.




-5-

Rudebusch

t-1
P - Cov[(l/x) I r t + ± - ( ( t - l ) / T ) r t . \

- rt]/Var(Rt

- rt).

The numerator of this term can be rewritten as
T-1 i
C o v [ ( l / x ) E ( I u t + i + e t + i ) - ((t-l)/x)e t . 6 t - ((x-l)/x)e t ]

= ((x-i)/x) 2 o^.
while the denominator can be rewritten as
Var[r t + (l/x)e t + 6 t - r t - e t ] = a2Q - ((x-1)/x) 2 o 2 .
A

Thus, the population value of p is given by
plim p - ((T-l)/t) 2 o|/[oJ - ((T-1)/T)2G2£] .
- 1 / [X(X/(T-1)) 2

(9)
where X s OQ/&1,

- 1],

the ratio of the variance of the term premium to

the variance of the transitory shock.
As can be seen from (9)» the estimate of P varies inversely
A
2
with X.

Thus, increases in O

are reflected in a higher p.

Intuitively, increased transitory deviations of today's funds rate
from the target provide more predictable future variation in the
funds rate because today's deviation will be eliminated, on average,
the next day through reversion to target.

This predictable future

reversion to target is incorporated in the long rate, which boosts
A
2
the value of p. In contrast, increases in a^ simply increase the
noise in the long rate and lower p. Most importantly, however, the
A

crucial feature of (9) to note is that P does not depend on the
variance of the policy shocks (o ) .

Intuitively, this results

because such shocks do not affect either side of the regression (8):
the shocks are reflected completely in both r

and R

and so do not

change the yield spread, and they are permanent and so do not show
up in the difference between future and current funds rates.

Thus,

A

evidence on the size of P places no restriction on the size of
policy shocks to the funds rate and hence has nothing to say about




-6-

Rudebusch

the extent to which the variability of the funds rate is reflected
in long rates.
The actual estimates of p from RRW during the 1975-1979 and
1979-1982 periods (with standard errors in parentheses) are
(Fact 1)

P* - 0.04
(.21)

and

j}** - 1.21 ,
(.25)

where a single asterisk denotes the period before October 1979 and a
double asterisk denotes the 1979-1982 period. The finding that p
A**

is less than P
(10)

X

implies that in terms of the model variances
(that is. C2Q ItT

> X

> ClQ

/<T

).

As we shall see below this evidence places no restriction on the
volatility of long rates or on the correlation of movements in the
long rates with movements in the funds rate.
INTERPRETING JOHNSON'S HISTORICAL EVIDENCE
In this section, I use the model of the funds rate
(equations 1 and 2) and of the long rate (equation 5) to analyze the
historical evidence provided in Johnson (1981) regarding changes in
the behavior of interest rates after 1979. In particular, the next
two subsections focus, in turn, on changes in (1) the volatility of
the funds rate and longer-term rates and (2) the correlations
between movements in the funds rate and in longer rates.

The Volatility

of the Funds Rate and Longer-Term

Rates

First, let us compute the volatility of the funds rate and
of the long rate in terms of the parameters of the model. The
variance of the change in the funds rate, denoted Ar - r - rr_i»
is given by
G

Ar

S

Var

^Art)

= Var(rt - r ^

+ et - e ^ )

- Var(ut + e t - e t . 1 )

6. These estimates are based on the spread between the three-month
Treasury bill and the funds rate and are taken from table 8 in RRW.




-7-

Rudebusch

-<

(ID

+

*>i-

Similarly, the variance of the change in the long rate is given by
°2AR

s

Var(ARt)

= Var[rt - r ^

+ (1/x) (et - e ^ )

+ 9t -

6^]

- G2u + 2(l/x)2G* + 2 G Q .

(12)

Thus, the volatility of the funds rate depends on both the variance
of the permanent policy shock, (T , and the variance of the temporary
shock, G . The volatility of the long rate depends primarily on the
variance of permanent policy shocks and the variance of the term
premium, G Q ; the variance of the transitory shock is negligible
because its weight, 2 (1/x) , is so small.
The historical standard deviations of changes in the funds
rate and in a variety of longer rates during the 1975-1979 and 19791982 periods are shown in table 1.
Clearly, the volatility of
all rates increased in the later period. Label this result as "fact
2"; that is.
(Fact 2)

G ^ < 02A"

and

C ^ < G^* ,

where again a single asterisk denotes the period before October 1979
and two asterisks denote the 1979-1982 period.
The historical evidence on volatility expressed by fact 2
could be reconciled with equations 11 and 12 in a number of
ways. A plausible set of assumptions about the variances of the
model's shocks that could explain fact 2 are:
(13)

c\

< c\

.

(14)

<*<<".

(15)

<TQ < CQ

.

7. This table extends some of the results in tables 4 and 9 of
Johnson (1981) to daily data and a larger post-1979 sample.




-8-

Rudebusch

Inequality 13 implies that the Federal Reserve allowed much larger
temporary deviations from its implicit target rate in the 1979-1982
period than before. Inequality 14 implies that the Federal Reserve
was much more aggressive in moving the target rate in the later
period than before. The larger policy shocks reflected, in part the
new emphasis on movements in Ml underlying Federal Reserve policy
g
actions.
Finally, inequality 15 implies that the variance of
the term premium also increased after 1979, which is consistent with
9
the increase in interest rate risk implicit in fact 1.
Thus, the greater volatility of the funds rate in the 19791982 period reflects both-the looser control by the Federal Reserve
over day-to-day movements in the funds rate (inequality 13) as well
as larger permanent policy shifts (inequality 14). The greater
volatility of longer rates reflects the greater policy shocks as
well as the increase in the volatility of the term premium.
The Correlation
of Movements in the Funds Rate and Longer
Rates
How does the correlation between changes in the funds rate
and the long rate depend on the model parameters? This correlation
10

u

is given by
p a Corr(Ar , AR )
- Cov(Art, ARt)/VVar(ARt)Var(Art)
- (K + 2 / T ) / V ( K

+ 2)(K +

2X

+ 2/x

2}

9

(16)

2 2
here K s a /a , the ratio of the variances of the policy and
transitory shocks, and where, as before, X s C Q / C T . The correlation

8. Both (13) and (14) are consistent with evidence presented in
Balduzzi, Bertola, and Foresi (1992).
9. Inequality 15 is consistent with the evidence in Cook and Hahn
(1990); the various proxies for the (ex ante) term premium that they
display show a clear increase in volatility during the 1979-1982
period.
10. Note that the covariance of the two rate changes is given by




Cov(Art, AR t ) - Et[(Art)(ARt)]
- E t [(u t +e t -e t . 1 )(u t +(l/t)(e t -£ t . 1 )+G t -e t . 1 )]
- c2 +
U

(2/T)O*
£

-9-

Rudebusch

2

p depends positively on K: increases in G increase the covariance
of Ar and AR proportionately more than their variances and hence
2
boost p. while increases in G increase the variances more than the
covariance and hence diminish p. Intuitively, the long and short
rates will be more closely correlated when the shocks that drive
them both (the permanent u policy shocks to r ) are more important
than the shocks that essentially drive only the short rate (the £
shocks of r about r ) . The correlation also depends negatively on
X because increases in the variability of the term premium (relative
to G ) simply add noise to the long rate and leave the short rate
unaffected.
The actual correlations of changes in the funds rate with a
variety of longer rates are shown in table 2.
The correlations
during the 1975-1979 period, p , are shown in the first column; the
correlations during the 1979-1982 period, p , are shown in the
second column. The correlations are all higher in the later period;
that is,
(Fact 3)

p

< p

.

In terms of the model variances, fact 3 implies that
K

< K

(that is, a2 /cr

< G2

/<r

e

u

e

( t h a t i s , GQ ICT

> GQ

u

),

or
X

> X

o r , most l i k e l y ,

/tr

)

b o t h 12

11. This extends some of the results in tables 10 through 13 in
Johnson (1981) to daily data and a larger post-1979 sample.
12. One factor that has been ignored is duration (see Goodfriend,
1992) . The results in table 2 at maturities greater than one year are
obtained from coupon securities, while the model considers only zerocoupon securities. The yield on a coupon security would move a bit
more in response to a transitory shock than the yield on a zero-coupon
security, and the correlation between the funds rate and the bond rate
would be slightly higher.




-10-

Rudebusch

CONCLUSION
The change in Federal Reserve operating procedures in
October 1979 ushered in, as had been expected, an era of increased
funds rate volatility. At the time, many were surprised by how
variable longer rates also became. In assessing whether a change in
operating procedures will increase the variability of long rates,
the key insight from the above model is that one must focus on the
permanent policy shocks. Such a narrow focus is not surprising
because only the long-lived shocks to the short rate will affect
long rates under the rational expectations hypothesis of the term
structure. Accordingly, whether future changes in the procedures
governing the behavior of the funds rate will affect long rates
depends in large part on whether the associated re-specified
reaction function for policy has been linked to variables that are
subject to more permanent shocks. More generally, in a world where
the rational expectations hypothesis of the term structure may not
hold, the variability of long rates depends on what market
participants believe
about changes in the size of the permanent
shocks.




-11-

Rudebusch

REFERENCES
Balduzzi, Pierluigi, Bertola, Giuseppe, and Foresi, 'Silverio. "A
Model of Target Changes and the Term Structure of Interest .
Rates," (1992), manuscript.
Cook, Timothy. "Determinants of the Federal Funds Rate: 1979-1982,"
Federal Reserve Bank of Richmond Economic Review, (1989), 3-19.
Cook, Timothy and Hahn, Thomas. "The Effect of Changes in the
Federal Funds Rate Target on Market Interest Rates in the
1970s," Journal of Monetary Economics, 24 (1989), 331-351.
Cook, Timothy and Hahn, Thomas. "Interest Rate Expectations and the
Slope of the Money Market Yield Cure," Federal Reserve Bank of
Richmond Economic Review, (1990), 3-26.
Gagnon, Joseph E. and Tryon, Ralph W., "Price and Output Stability
Under Alternative Monetary Policy Rules," (1992), manuscript.
Goodfriend, Marvin. "Interest Rates and the Conduct of Monetary
Policy," Carnegie-Rochester Series on Public Policy 34 (Spring,
1991), 7-30.
Goodfriend, Marvin, "Interest Rate Policy and the Inflation Scare
Problem: 1979-1992," (1992), manuscript.
Hess, Gregory D., Small, David H., and Brayton, Flint, "Nominal
Income Targeting with the Monetary Base as Instrument: An
Evaluation of McCallunTs Rule," (1992), manuscript.
Johnson, Dana. "Interest Rate Volatility Under the New Operating
Procedures and the Initial Response in Financial Markets," in
New Monetary Control Procedures, vol. 1, February 1981, Board
of Governors of the Federal Reserve System.
Kasman, Bruce, "A Comparison of Monetary Policy Operating Procedures
in Six Industrial Countries," (1992), manuscript.




-12-

Rudebusch

Roberds. William, Runkle, David, and Whiteman, Charles H. "Another
Hole in the Ozone Layer: Changes in FOMC Operating Procedure
and the Term Structure," (1992), manuscript.




-13-

Rudebusch

Table 1
Standard Deviations of Changes in Federal Funds Rates
and in Rates on Various Treasury Securities
(Daily data; percentage points)
SflflipJ-e period
Type of security
Federal funds
3-month bill
6-month bill
1-year bill
3-year note
5-year note
10-year note
20-year bond

7W9
.32
.09
.07
.08
.06
.05
.04
.03

79-82
.96
.34
.31
.27
.21
.19
.17
.16

The sample period "75-79" includes data from January 6, 1975 through
September 28, 1979; the sample period "79-82" includes data from
October 15, 1979 through October 1, 1982.

Table 2
Correlations of Changes in the Funds Rate with
Changes in Rates on Various Treasury Securities
(Daily data)
Sample period
Type of security
3-month bill
6-month bill
1-year bill
3-year note
5-year note
10-year note
20-year bond

75-79
.07
.05
.10
.06
.05
.04
.02

The sample periods are the same as in table 1.




79-S2
.21
.22
.20
.19
.16
.15
.15




Figure 1
Federal Funds Rate
Target
Actual

1976

1977

A POLICYMAKER'S GUIDE TO INDICATORS OP ECONOMIC ACTIVITY

Charles Evans, Steven Strongin, and Francesca Eugeni

The evaluation of economic indicators has often progressed with an
odd independence for the way in which such indicators are actually
used in practice in the economic policy process.

The search is

often for one "best" indicator, where "best" typically refers to
winning in some narrowly defined contest of general purpose
forecasting ability measured over some pre-selected time-span.
The actuality of the policy process is far richer.

Indicators are

used in a kind of chaotic democracy, each indicator casting a vote
based on its own forecast and the policymakers weighing each vote,
based on their assessment of the current credibility of the
indicator.
This is quite different from the standard academic
formulation of economic policy where a "true" model is developed
and then policy run in a way optimizes the performance of the
model.

Understanding this difference in approach leads to very

different ways of evaluating indicators.

It is not just enough to

produce a "best" model; rather, it is important to understand what
type of information is contained in a given indicator so that its
message can be properly evaluated and also how much weight to give
that message given what else is also known.
Indicators, like people, perform better or worse depending
on the context in which they operate.

Efficient usage requires

matching indicators both with appropriate questions and with other
complementary indicators.

For instance, some indicators do far

better at predicting short-run changes in activity, but do not do
very well at pinning down the level of activity over longer time
spans, while other indicators forecast short-run phenomena poorly,
but do better at predicting average activity over longer time
span.

Also while some indicators have very close substitutes,

such as the twenty or so short-term interest rates sometimes used
in econometric studies, and thus provide little additional
information beyond that already contained in other indicators,
some indicators can provide substantial additional information,
thus providing important confirming or contradicting information.
The policymaker needs to know how to match questions with




Evans, Strongin, and Eugeni
indicators depending on the current policy context.

A swiss army

knife is a fine general purpose tool, but it is hardly a
substitute for a we11-equipped workshop.
This paper seeks to develop and implement a set of
techniques for evaluating indicators of economic activity that
more closely match the actual use of such indicators in the dayto-day policy process.

We see that process as primarily involving

the re-assessment of short- to medium-term economic activity based
on indicator by indicator analysis with the primary decision
matrix being whether it is necessary to ease or tighten policy in
order to realize appropriate levels of economic activity.

We do

not address the longer run issues of assessing appropriate levels
of economic activity or other issues involving inflation or the
value of the dollar nor do we address the question of how best to
implements those decisions.

Evaluating indicators in this context

has four primary parts; ranking candidate indicators,
characterizing the nature of the information in those indicators,
assessing their usefulness in practice and determining what
relative weight should be given to each indicator.

The idea is to

develop the information that a policymaker needs in order to
interpret information as it comes in and to choose which
indicators to watch depending on the questions being asked.
As policymakers typically use indicators one at a time, all
of our analysis will be carried out on a bivariate basis.
Multivariate regression models allow indicators to play off
against one another so that if two indicators hold both common and
independent information better statistical fits can usually be
obtained by fitting one multivariate model rather than mixing 2
bivariate models.

The advantage of using the mixing approach is

that when one of the indicators begins to misbehave, which they
do, you can, at least temporarily, just ignore that indicator.
Second, by only using the primary information over-fitting is less
of a worry.

Third and most important, the mixing approach allows

a much more precise assessment of exactly the type of information
is contained in each indicator and thus allows policymakers to
reoptimize their choice of indicator sets based on the type of




-2-

Evans, Strongin, and Eugeni
question being asked.
Beyond the focus on bivariate models, there are a number of
other differences between our work and normal econometric practice
that are worth noting.

First, as will be shown in the paper

different indicators are useful at different forecast horizons, so
that we will not be suggesting one best model, but rather we will
be suggesting ways of combining indicators depending on the
precise policy question being asked.

Second, along these same

lines as we are more concerned with the interpretation of each of
the individual indicators rather than the construction of a
structural model of the economy, we will pay much more attention
to characterizing the type of information in each individual
indicator than is normally the case.

Also, since the forecasts

derived from the indicators typically get averaged together either
informally in the policymaker's mind or formally in the mixing
models shown in the last section of this paper, we analyze the
degree to which one indicator can be said to have information
which is independent from another.

Policymakers are often faced

with a variety of indicators pointing one way and another group
pointing a different way, in such cases it is not only useful to
know what weight would have produced the best forecast
historically, but the degree to which the indicators are
independent bits of information or the same information being
repeated over and over again in a variety of guises.

Policymakers

quite rightly give greater weight to information which they see as
independent confirmation.

It is useful in this light to more

fully analysis the independence of information in various
indicators.

It is also helpful to know if the indicator in

question usually contains the type of information being sought.

METHODOLOGY
As noted above, the primary focus of this paper is the examination
of various data series as indicators of changes in real economic
activity, which we measure alternately as annualized log change in
real GDP, employment and industrial production.

In most cases

results are supplied for all three measures of economic activity.




-3-

Evans, Strongin, and Eugeni
The major focus will be on the forecasting real GDP, except in the
sections of the paper which focus on issues of timing in which
case employment will be used, since it is available at the monthly
frequency allowing for more precise estimation of the pattern of
impact over time.
Throughout the paper the indicators are used to produce
forecasts of economic activity.

The specific functional form of

the forecasting equation is always the same.

One year of data for

the indicator and one year of lagged economic activity is included
in the regression.

Thus, the exercise is strictly equivalent to a

bivariate VAR with one year of lags, 4 lags for the real GDP
models and 12 lags for the employment and industrial production
models.

The models are estimated in log differences and rates of

change are annualized.

Interest rates and interest rate spreads

are used in their level form.

In many of the tables an additional

forecast is provided with the label "NONE".

In this case, the

forecast is based solely on the past history of economic activity,
a pure auto-regressive model with one year of lagged data.

This

pure auto-regressive forecast is referred to as the no-indicator
forecast.

When the horizon of forecast is varied, we simply

change the dependent variable in the regression rather than
dynamically iterate the one period ahead forecast.

This optimizes

the parameterization for the forecast horizon in question, rather
than multiplicatively combining estimation errors forward.
Symbolically the forecasting equation can be written:
y : . r y t =A(L)Ay M + B(D x.^ + co,

where Yt it the log of economic activity at time t and It is the
indicator at time t, k is the number of periods in the forecast
horizon and A(L) and B(L) are polynomial in the lag operator L of
order one year.
The indicators are split into four groups, which we call
families.

Each family is meant to represent a natural division of

indicators into groups which are likely to share similar
characteristics.




For example, the first family we examine is

-4-

Evans, Strongin, and Eugeni
interest rates, the second is money-based measures, the third is
interest rate spreads and the fourth is composite indicators, such
as the Department of Commerce Leading Indicators and the S&P 500.
The fourth group also contains those series which do not fit
neatly into the overall classification scheme.
The idea is to first examine the indicators within a family,
finding out which indicators within each family produce the best
forecasts and contain the most independent information and then
taking these "best" indicators and examining what is to be gained
by mixing the information from different families.
number of purposes.

This serves a

First, by breaking the large list of

potential indicators into smaller groups it makes each examination
more manageable.

Second, by using natural groupings it allows us

to look at questions such as what is the best interest rate or the
best money measure in a natural way.

Third, one key issue for

indicators is the degree to which they actually contain
independent information.

Focusing on groups which are already

thought to have similar information provides a natural focus to
learn if these preconceptions are accurate or if some of these
groups contain more than one type of information.

Lastly, by

first selecting the best indicators at the family level and then
mixing between families, we can produce a mixed forecast which, as
noted above, closely approximates the way policy forecasting
appears to be done in practice.
Each family of indicators is subjected to the same analysis.
First, each family of indicators is described and a table is
presented which lists the indicators examined and their means,
standard deviations and their correlations with the measures of
economic activity.

Then each of the indicators is subjected to

four evaluations, 1.) Classical goodness of fit rankings, 2.)
Characterization of fit, 3.) Indicators performance in practice
and 4.) Encompassing tests.
The classical goodness of fit rankings are based on simple
full sample regressions estimated on data from the beginning of
1962 through the end of 1991.

The results

two of each family analysis section.




-5-

are presented in table

Table two shows the rankings

Evans, Strongin, and Eugeni
for each indicators in the family based on the regression they
produce.

The idea is that the best indicators are the ones that

produces the best fit where fit is measured by the R* of the
regression or the standard deviation of the residual from the
regression1.

This closely approximates the oldest notions of

evaluating the best indicators of economic activity for policy.
It is also closely linked to the notion of Granger causality,
which measures whether or not the indicators actually helps
forecast economic activity.

The p-value for this test is also

included in the table.
The second evaluation seeks to characterize the type of
information in the indicator.

Typically the question can be

thought of as if the indicator goes up today how does that change
my expectations about economic activity in the future.

This is

analyzed by calculating the dynamic response path of employment
for each of the indicator forecasting equations, which shows how a
one standard deviation2 increase in the indicator changes
expectations about future growth rate of employment for each month
for the next 3 6 months3.

This allows us to characterize the

information in the indicator based on how fast economic activity
responds, how much it responds and how long the change in activity
lasts.

Figure 1 in each family section graphs the dynamic

response path for each indicator in the family, as well as the 2
standard deviation bands on the estimates of the dynamic response
paths to show the amount of uncertainty about the response path.

1. In the appendix tables which include sub-sample results are
also presented.
2.
The standard deviation measure used is the one from a
bivariate VAR for the indicator and the measure of economic
activity, this is used to approximate the average size of movement
in the indicator series.
3 . This is basically the same as an impulse response function
except that the identifying assumption is not derived from a
specific decomposition of the error matrix, but on the assumed path
of the actual series, i.e. the indicator changes given the level of
current activity. This is arithmetically equivalent to an impulse
response function using a Choleski decomposition with the indicator
ordered last.




-6-

<*****,

Evans, Strongin, and Eugeni
Table 3 summarizes this information in terms of the maximum
response for all three of the measures of economic activity,
showing the timing, size and uncertainty of the maximum response
of economic activity for each indicator in the family.
The third evaluation switches the focus to how well the
indicators are likely to work in practice.

To this end, goodness

of fit is reinterpreted in a way closer to the way forecasts are
actually used.

First, table 4 shows the goodness of fit ranking

recalculated for a series of forecast horizons, so that we can get
a better feel for what these indicators are good at.

First, the

single period horizon used in the first evaluation and then a onequarter horizon, a two-quarter horizon and a one-year horizon4.
Table 5 in each section then repeats this analysis using
forecasting equations which do not contain any prior information.
Specifically, the forecasting equations are estimated sequentially
using Kalman filtering techniques using only the sample
information available prior to the period being forecast.

This

provides a more accurate assessment of how an indicator is likely
to perform in practice.

These forecasts are then ranked

by the

mean squared error (MSE) of the forecasts from 1972 onward.
2

R s are no longer well defined.

The

This analysis is followed up by

Figure 2 in each section which graphs the cumulative residuals for
Kalman forecasts from 1972 onward.

This allows us to examine if

these forecasts tend to perform badly during recessions or if
there was some particular point in the past where they did
especially well or poorly.

It also tells us if the forecasts have

tended to miss in some systematic fashion over time.

The

residuals are measures as the actual growth in economic activity
minus the forecasted growth.

Thus, a downward trend in the

cumulative residuals would indicate a prolonged period of over-

4.
It should be noted that these are not iterated VAR
forecasts, rather the forecast parameters are chosen to maximize
performance at the forecast horizon specified, this can either be
thought of as a state space estimation minimize the t+k forecast
variance or as simple OLS with the dependent variable the t+k growth
rate. This avoids any problem that might result from a indicator
that performs poorly at high frequencies having that failure
interfere with longer frequency forecasting.




-7-

Evans, Strongin, and Eugeni
predicting growth in activity.
The fourth evaluation switches the focus to independence of
information.

As noted above one of the most important factors to

understand about indicators is whether of not they contain
independent information relative to some other indicator.

This

allows a policymakers to assess whether a new piece of information
actually contains any additional information or whether it is
simply the same information with a different label.

The way to

evaluate this is through a set of techniques called encompassing
tests.
encompass

In the context of this paper, indicator A is said to
indicator B, if given the forecast implicit based on A

there is no additional information in indicator B.
said to dominate
encompass A.

Indicator A is

indicator B if A encompasses B and B does not

The simplest way to test this is to run a regression

with economic activity as the dependent variable and the forecast
of activity based on indicator A and the forecast of activity
based on indicator B as the independent variables.

Symbolically

this can be written
AGDPr=<|> for(A)t+

(1-$) for(B)t

+t

Where for(A)t and for(B)t are the forecasts of GDPt based on
indicators A and B respectively and <)> is relative weight OLS
assigns to for(A)t and for(B)t.

If <J> is significantly different

from 0 then we can reject that for(A) is encompassed by for(B).
Likewise if l-<(> is significantly different from 0 then we can
reject that for(B) is encompassed by for(A).

If neither is

encompassed then both indicators contain independent information
and a better forecast can be obtained by mixing both sets of
information with the relative weight given by <(>.

If only one is

encompassed, then it is said to be dominated and only the other is
necessary to produce an efficient forecast.

If both are

encompassed then either indicator alone can produce an efficient
forecast, this occurs when there is a very high degree of
collinearity and the standard error of the parameter estimate is
large.




In this case the indicator which has the best historical

-8-

Evans, Strongin, and Eugeni
track record would likely be the superior choice.

The

generalization to longer horizons is straight forward, though the
calculations of the standard errors is more complicated since the
errors are no longer independent.
Table 6 in each family section contains the encompassing
test.

The table is read as follows.

The indicators are listed

both along the top and along the side.

The numbers in the table

refer to the test that the indicator listed along the side is
encompassed by the indicator along the top.

The test statistics

are the significance levels for the test the indicator along the
top does in fact contain all the information in the indicator
along the side.

For the sake of readability values below .05 are

left blank.
The way to interpret these tables is that an indicator whose
row is blank contains information that is independent of every
other indicator in the family.

An indictor whose column is full

of high numbers is said to encompass those indicators.

An

indicator that did both would be said to dominate the family.

In

general, what we will search for is the set of indicators in each
family which contain all the information in the family using as
few indicators as possible.

In general this will mean that the

best variable from the previous tests will be included plus
additional indicators which contain independent information i.e.
the indicators that add the most.

Formally this means that all

indicators that are not encompassed by any other indicators in the
family plus whatever additional indicators are necessary to fully
encompasses or cover

all the other indicators in the family.

This

is analogous to finding a set of minimum sufficient statistics.
The indicators that make it through this process will then
be tested in the mixing model section of the paper in betweenfamily encompassing tests, which examine whether or not there is
independent information between families or not.

Then a set of

"best" indicators will be selected in order to develop a mixing
models of indicators which contain independent information for
each of the forecasting horizons.

These models will contain

estimates of the appropriate relative weights that should be




-9-

Evans, Strongm, and Eugeni
applied to the individual indicator based forecasts.

Completing

the circle of policy forecasts, The mixing model will be timevarying to see if there is any gain from adjusting the weight
applied to these individual forecasts based on recent performance.

INTEREST RATE LEVELS
Table 1.1 lists the nominal interest rates which were selected for
investigation, along with some descriptive statistics.
rates are expressed at annual rates:

All of the

the Federal Funds rate (FF),

3- and 6-month Treasury bill rates (TB03 and TB06), 1-, 3-, 5-,
and 10-year constant maturity Treasury bond rates (CM01, CM03,
CMOS, and CM10), the 3-month Eurodollar rate (EUR03), the 6-month
Commercial Paper rate (CP6), and the BAA bond rate (BAAS).

Each

of these interest rates is negatively correlated with the economic
activity variables.

The interest rates with the largest absolute

correlation with real GDP are the Federal Funds rate, the 3-month
Eurodollar rate, and the 6-month Commercial Paper rate.
Table 1.2 reports statistics for the one-period-ahead
forecasting model.

Notice that all of the interest rates provide

significant predictive power for all three economic activities.
The R2 fall within fairly narrow bands indicating that the
relative rankings are not particularly important--all of these
indicators are useful at the one-month forecast horizon.
Figure 1.1 graphs the response of the employment growth
forecast to a one-standard deviation change in information about
last period's indicator.

As with the F-tests in Table 1.2, the

response paths are virtually identical across the interest rates
considered:

employment growth rises for three or four months and

then falls, eventually asymptoting back to zero from below the
axis.

The confidence bounds on these responses are sufficiently

wide that the initial response could be zero.

For all of the

interest rates, however, there is a point within the first year
that employment growth is significantly negative:

the largest

such effects are for the 6-month Commercial Paper rate and the BAA
bond rate.

For all of the indicators across all of the

activities, the maximum effect is negative and occurs within one




-10-

Evans, Strongin, and Eugeni
year of the impulse.
Tables 1.4 and 1.5 rank the indicator forecasts for m sample and out-of-sample forecasting behavior.

Focusing on the

out-of-sample results first, notice that for industrial production
and employment at the one-month horizon, the no-indicator
forecasts perform better than the interest rate forecasts.

But

for GDP all of the interest rate forecasts outperform the noindicator forecasts at all horizons.

Focusing on GDP, the Federal

Funds rate is ranked first at the four-quarter growth horizon;
but the 3-month Eurodollar rate is best at the one- and twoquarter horizons.

The success of the Eurodollar rate is also

evident for industrial production and employment at all horizons
beyond one-month.

The 6-month Commercial Paper rate improves in

forecasting accuracy as the horizon increases;

this is true for

GDP, industrial production, and employment (placing no worse than
third at the one-year horizon).

In general, the shorter maturity

bills perform better than the longer maturity bonds (3-, 5-, and
10-year Treasuries).
The in-sample results of Table 1.4 indicate that the
Eurodollar rate increases in ranking due in part to its out-ofsample stability.

In the out-of-sample rankings the Eurodollar

rate is first for industrial production (3-, 6-, 12-months),
employment (6- and 12-month), and GDP (one- and two-quarters).

In

6 of these 7 instances, these represent an increase in ranking
from the m-sample results.

In contrast to this stability, the 6-

month Commercial Paper rate does not fare as well.

At the shorter

forecast horizons, it goes from being ranked number 1 or 2 insample to either 6, 9, or 10 out-of-sample.

For the industrial

production and employment, the Federal Funds rate also experiences
a substantial out-of-sample forecast deterioration at the shorter
forecast horizons relative to the in-sample rankings.
The cumulated residuals from the Kalman forecasts in Figure
1.2 show that, overall, the indicators in our interest rate family
consistently underforecasted real GDP between 1974 and 1982.

The

upward trend in the cumulated residuals during this period can be
explained in part by an unprecedented increase in inflation, which




-11-

Evans, Strongm f

and Eugeni

caused interest rates to rise without the normally anticipated
decline in output.

On the other hand, between 1983 and 1989, the

Federal Funds rate, the 6-month Commercial Paper rate, the
Eurodollar rate, and all of the Treasury bill rates performed
well, as shown by the flattening of their cumulated residuals
slopes during this period.

Between 1990 and 1991, however, the

indicators performance deteriorated again, as all of the interest
rates missed the 1990-91 recession and consistently overforecasted
real GDP.
Table 1.6 reports the encompassing results for GDP.
simplest case is for the 4-quarter horizon:

The

the Federal Funds

rate dominates the other interest rates since it is unencompassed
and it encompasses all other interest rates at this horizon.

At

the one- and two-quarter horizons, however, this domination does
not hold;

none of the interest rates are unencompassed at these

horizons.

Since all of the interest rates Granger-cause economic

activity in Table 1.2, it is probably not surprising that each of
the interest rates contains useful forecasting information.

For

example, at the one-quarter horizon the Federal Funds rate, the 3month Eurodollar rate and the 6-month Commercial Paper rate all
can be said to encompass each other, i.e.

if you know one

interest rate based forecast knowing another is not much help.
Since all of these interest rate forecasts are encompassed by at
least one other interest rate forecast, the next criterion for
selection is to determine if any one of the interest rate
forecasts can cover all of the other interest rate forecasts.

In

fact, at the one-quarter horizon, the Federal Funds rate, the 3month Eurodollar rate, and the 6-month Commercial Paper rate all
cover every other interest rate.

The 3-month Eurodollar rate

covers the Federal Funds rate and the 6-month Commercial Paper
rate with higher levels of significance, and since, as noted
above, the 3-month Eurodollar rate was the number one ranked
indicator in the out-of-sample forecasts of GDP at the one-quarter
horizon, the 3-month Eurodollar rate is selected as the best
interest rate level indicator at the one-quarter horizon.




-12-

Similar

Evans, Strongm, and Eugeni

reasoning leads to the selection of the 3-month Eurodollar rate
for the two-quarter horizon.




-13-

TABLE 1.1 - DESCRIPTIVE STATISTICS

QUARTERLY (Jan 62 - Dec 91)

MONTHLY (Jan 62 - Feb 92)

Mean

Std. Dev.

Correlation with
Real GDP

-0.245

7.370

3.304

•0.353

-0.190

-0.222

6.620

2.686

-0.299

2.647

-0.186

-0.219

6.777

2.622

•0.295

7.265

2.872

-0.173

-0.215

7.282

2.849

-0.282

CM03

7.597

2.746

-0.162

-0.226

7.608

2.737

-0.257

CM05

7.736

2.708

-0.161

-0.231

7.744

2.705

-0.251

CM10

7.866

2.674

-0.154

-0.231

7.869

2.678

-0.237

EUR03

8.033

3.282

-0.224

-ff.254

8.055

3.232

-0.352

CP6

7.341

2.879

-0.223

-0.252

7.359

2.844

-0.342

BAA

9.588

3.108

-0.188

-0.286

9.590

3.120

-0.269

Correlation with
Industrial
Employment
Production

Indicator

Mean

Std.Dev.

FF

7.352

3.345

-0.230

TB03

6.605

2.715

TB06

6.761

CM01
,
f"




TABLE 1.2 - CLASSICAL GOODNESS-OF-FIT STATISTICS

MONTHLY (Jan 62 - Feb 92)

MONTHLY (Jan 62 - Feb 92)

INDUSTRIAL PRODUCTION

EMPLOYMENT

Change
Indicator

QUARTERLY (Jan 62 - Dec 91)
GDP

Change

Change
, R2

In R2

SEE

P Value Rank

6

0.338

0.220

3.148

0.0000

3

0.0028

5

0.293

0.176

3.252

0.0001

6

2.291

0.0015

3

0.304

0.186

3.227

0.0000

5

0.054

2.294

0.0020

4

0.309

0.191

3.216

0.0000

4

0.047

2.307

0.0080

7

0.279

0.161

3.285

0.0002

7

0.268

0.150

3.310

0.0003

8

R2

lnR2

SEE

P-Value

Rank

R2

lnR2

SEE

P-Value

Rank

FF

0.259

0.063

8.965

0.0057

10

0.423

0.050

2.303

0.0052

TB03

0.263

0.067

8.940

0.0030

8

0.426

0.053

2.297

TB06

0.271

0.076

8.887

0.0007

5

0.429

0.055

CM01

0.273

0.078

8.875

0.0005

4

0.428

CM03

0.265

0.069

8.930

0 0022

7

0.421

i

—
•»
^

CM05

0.263

0.067

8.941

0.0030

9

0.419

0.046

2.310

0.0109

8

0.253

0.136

3343

0.0009

10

CM10

0.265

0.070

8.925

00020

6

0.417

0.043

2.315

0.0171

10

0.354

0.236

3.110

0.0000

1

EUR03

0.276

0.081

8.859

0.0003

3

0.431

0.057

2.287

0.0010

2

0.348

0.231

3.123

0.0000

2

CP6

0.286

0091

8.797

0.0001

1

0.438

0.065

2.273

0.0002

1

0.258

0.140

3.333

0.0007

9

BAA

0.283

0.087

8.818

0.0001

2

0.419

0.045

2.312

0.0124

9




TABLE 1.3 - MAXIMUM IMPACT OF DYNAMIC MULTIPLIERS

MONTHLY (Jan 62 - Feb 92)

MONTHLY (Jan 62 - Feb 92)

QUARTERLY (Jan 62 - Dec 91)
GDP

EMPLOYMENT

INDUSTRIAL PRODUCTION

Months to
Max

Max Impact

Std. Dev.
at Max

Months to
Max

Max Impact

Std. Dev.
at Max

Quarters to
Max

Max Impact

Std. Dev.
at Max

FF

7

-1.349

0454

10

-0.406

0.129

3

-1.442

0 307

TB03

5

-1.577

0499

9

-0.335

0135

3

-1.250

0 280

TB06

5

-1.610

0.496

12

-0.365

0.130

3

-1.354

0 303

CM01

5

-1.655

0468

12

-0 411

0.123

3

-1.443

0 308

CM03

5

-1.550

0.482

12

-0.414

0.146

3

-1.383

0 326

CM05

5

-1.446

0480

10

-0.431

0141

3

-1.367

0 303

CM10

12

-1.270

0.488

12

-0.382

0.145

3

-1.332

0314

EUR03

5

-1.615

0.475

9

-0.494

0.124

3

-1605

0.290

CP6

5

-1.793

0.484

9

-0.464

0.123

3

-1.502

0.282

BAA

5

-1.973

0486

7

-0 395

0138

3

-1.280

0310

Indicator




TABLE 1.4 • MULTIPERIOD FORECASTS (In-Sample)

INDICATOR

(

1MON
R2 RANK

MONTHLY (Jan 62 - Feb 92)

MONTHLY (Jan 62 - Feb 92)

INDUSTRIAL PRODUCTION

EMPLOYMENT

3MOS
R2 RANK

6MOS
R2 RANK

12MOS
R2

RANK

1MON
R2 RANK

3MOS
R2

RANK

QUARTERLY (Jan 6 2 - Dec 91)
GDP

6MOS
R2 RANK

12MOS
R2 RANK

lOTR
R2 RANK

2QTRS
R2 RANK

4QTRS
*R2 RANK

FF

0258

10

0J351

3

0400

3

0530

2

0423

6

0576

3

0571

3

0561

2

0338

3

0463

3

0 530

1

TB03

0263

8

0333

7

0337

6

0 477

5

0426

5

0564

7

0543

6

0529

4

0293

6

0402

5

0 496

3

TB06

0271

5

0346

5

0353

4

0483

4

0 429

3

0570

5

0 550

4

0528

5

0304

5

0406

4

0 487

5

CM01

0J?73

4

0341

6

0350

5

0 455

6

0 428

4

0567

6

0 547

5

0509

6

0309

4

0397

6

0443

6

CM03

0264

7

0325

8

0329

8

0410

7

0.421

7

0561

8

0540

8

0.486

7

0279

7

0350

7

0377

7

CMOS

0263

9

0318

9

0314

9

0 388

8

0419

8

0560

9

0536

9

0474

8

0268

8

0332

8

0346

8

CM10

0265

6

0307

10

0280

10

0346

10

0417

10

0550

10

0514

10

0442

10

0253

10

0296

10

0307

10

EUR03

0276

3

0371

2

0420

2

0509

3

0431

2

0581

2

0575

2

0546

3

0354

1

0 471

2

0 490

4

CP6

0286

1

0382

1

0438

1

0541

1

0 438

1

0592

1

0592

1

0564

1

0348

2

0 475

1

0516

2

BAA

0283

2

0348

4

0332

7

0361

9

0419

9

0570

4

0543

7

0468

9

0258

9

0329

9

0315

9

NONE

0196

11

0201

11

0115

11

0097

11

0373

11

0.489

11

0414

11

0269

11

0118

11

0123

11

0076

11

>-*

•>v4




TABLE 1 5 - KAl MAN MULT1PEWO0 FORECASTS (Out of Sample)

M O N T H . Y (Jul 73

Feb 92)

INDUSTRIAL PRODUCTION

COP

EMPLOYMENT

CMOS
RMSE RANK

12MOS

1MON

RMSE RANK

RMSE RANK

11

7 141

6

4 778

3

2 761

11

2 162

11

1988

8 601

10

7353

11

4993

6

2707

9

2 129

10

8

8 301

7

7 076

7

4847

4

2644

7

2 074

10 664

7

8 198

5

6898

4

4882

5

2630

5

CMOS

10 577

4

8 154

3

6 897

3

5029

7

2604

CMOS

10 609

5

8229

6

6 973

5

5 131

8

CM10

10629

6

8 349

8

7172

10

5354

EURC3

10483

2

7899

1

6415

1

CPf

11 196

10

8426

9

6 801

BAA

10518

3

8 184

4

7 158

1MON

3MOS

INDICATOR

RMSE RANK

FF

11232

11

8 845

TB03

11 168

9

TB06

10 735

CM01

1
00
1




RMSE RANK

QUARTERLY (Jul 73I - D e c 9 l )

MONTHLY (Jul 73 • Feb 92)

12MOS

1QTR

RMSE RANK

RMSE RANK

2QTRS
RMSE RANK

4QTRS
RMSE RANK

10

1664

2

3793

2

2859

3

2160

1

2031

11

1728

6

3969

9

3075

6

22G0

5

8

1974

9

1702

4

3862

4

3000

5

2251

4

2 042

7

1932

5

1706

5

3826

3

2996

4

2356

6

3

2 011

5

1913

2

1732

7

3876

5

3094

7

2483

7

2599

2

2 010

4

1920

4

1757

8

3936

7

3144

8

2552

8

9

2625

4

2036

6

1969

8

1826

10

3949

8

3249

10

2683

9

4 531

1

2630

6

1993

2

1847

1

1610

1

3622

1

2754

1

2222

3

2

4 724

2

2752

10

2 124

9

1951

6

1686

3

3880

6

2827

2

2 216

2

9

5 495

11

2657

8

1998

3

1915

3

1784

9

4006

10

3197

9

2 725

10

3MOS

CMOS

RMSE RANK

RMSE RANK




o^oddsofioocvjuJs
o o o o o o o o o o

O

ooo>^G5d>o>£»inc40)
o o o o o o o o

II
I I

I

* - CO

I I CO g

J J
I I

(Q

l§§§§S§gsf§
o o o o o o o o o o

I
I

I I I I I - I
I I I I I J I

I
I

o o o

<OCM^-^,»-CM<£>©

o o o o o o o o

r^ i^ co co

co
c

o

W O O B O O "

* - co co co <o co
o> to
8 8 § r^ 8 8 co !2 3 8

CM

oo

0 0 0 0 0 0 0

I

I I

I I
I I

m

1 8
I ™
o

co

i o o o © o © o

c

o i i o o n u i n

© O

u

to

i 8 8 S 2 2 S £ 5$
©

CM h-

288S88 «8£

8 3 8 8 8 3 3 «S8

fc

0 0 0 0 0 0 0

f - w N t f c m

CM

II
l I l 1o 2o2o8o8 ! ^

a
C

0 0 0 0 0 0

c

© ©

o o

I I
I I

I I I
I I I

I
I

I I
I I

I I
I I

I I £
1 I o

I I
I I

Ireco 1
I c
'

I I
I I

r»*

I
I

I
I

I I
I I

I
I

Iret
I c I

I o

00

8 581 1 §

1 1
1 1

1 1 1 « & 1
II
I C O I

ll
CM 1

0 0

! 8$
I

00 o>

o o

CO

3 2 s I
CM ^
^ t«>
O I
O O O

1o
*

8

I
I

10
O

CO CM ™ * - <— CM
O O c O O O

j

«
re33°i£ ,
o> J: ® 4 m ^ l

CM O O 00

! ? ^ S 5 ^ !

I

1 1
l 1

1

0

c

0 0 0 0

, ao tn o> in co
*3Jn8&£

^r co co 10 * -

c

CMinN.CMN.^O>?MCp
^O>(DtO(D^r«-CMC0

0 0 0 0 0 0 0 0 0

© © © © ©

l

j

I I
I I

00
O

635

i 8 j I! c; ^» l§ •

I
I

O

I «£ re

o

O

co

N-

I O O IO

O

CM

coco*-<53co^^
com<o<o<0CMCMin
0 0 0 0 0 0 0 0 0

1 1

s I I s

I
I

1 0

8

re 1 j
<= l I

CO CO CM CM CO

* 5 co 8 8 8 CM
fc

g

© © © © © ©

e o 2 S $ 8 5 8 8 5 S 5
^ a a m m i o m i A r ^ o o
0 0 0 0 0 0 0 0 0

co
n m ^ n i A o O

_

0 0 0 0 0 ^ C ( D <

o o 2 o o ^ a c j p <
u . h h U U U U U J U f f l

U.H-KOOOOUJOCD

-19-

ttt-i-ooooujom

1.1. Dynamic Response of Employment to Interest Rate Levels
Fed Funds (FF)
annualized percent growth rates
050 r




5 year Treasury bond (CM05)
annualized percent growth rates
050 r

-20-

1.2. Interest Rate Levels: Cumulated Kaiman Residuals in Forecasting Real GDP
Fed funds (FF)
cumulated Kaiman residuals
100 r

5 year Treasury bond (CM05)
cumulated Kaiman residuals
75 r

10 year Treasury bond (CM 10)
75 r

6 month Treasury bill (TB06)
100 r

1 year Treasury bond (CM01)
100 r

6 month commercial paper (CP6)
100

3 year Treasury bond (CM03)
100 r




1973

-21-

Evans, Strongm, and Eugeni

THE MONETARY AGGREGATES
Table 2.1 lists the monetary indicators which were selected for
investigation, along with some descriptive statistics.

For this

family of indicators all but one of the variables are expressed as
(log) growth rates:

the monetary base (Board of Governors (MB)

and St. Louis (MBSTL) versions), Ml, M2, M3, L, and long-term debt
of nonfinancial institutions, as well as real Ml and real M2
(deflated by the consumer price index).

The other monetary

indicator is the ratio of nonborrowed reserves (this period) to
total reserves (last period) (NBRX).

Strongin (1991) has found

that this normalized reserve aggregate contains much of the
information about monetary policy actions which Sims (1991)
attributes to innovations in the Federal Funds rate
(orthogonalized relative to output and prices).
Two observations about the descriptive statistics seem to be
in order.

First, these aggregates are plausible choices as

monetary indicators of economic activity.

Focusing on GDP, the

aggregates tend to be correlated with GDP, and the highest
correlations are with the real aggregates Ml and M2.

In fact, it

appears to be roughly the case that as the endogenous component of
the monetary aggregate increases, the contemporaneous correlation
with economic activity increases.

This is loosely the

causation/reverse causation debate--do the larger monetary
aggregates influence activity more than the narrower aggregates,
or are they influenced more?

Second, for most of the aggregates

the standard deviations are about one-half or less than the
average growth rates;

however, for real Ml and M2, the standard

deviations are 2 and 6 times greater than the average growth rate.
In turns out below, that these two aggregates, nominal M2, and the
NBR/TR ratio are the most useful indicators.
Table 2.2 reports statistics for the one-period-ahead
forecasting model, an autoregression of the economic activity
variable with lagged values of the indicator included.

Focusing

on GDP, notice that nominal M2, real Ml, real M2, and the NBR/TR
ratio provide significant predictive power for GDP beyond the




-22-

Evans, Strongin, and Eugeni
information contained in past values of GDP.

These three

indicators consistently provide predictive power for industrial
Production and employment as well.

For GDP the lowest ranking

indicators tend to be nonfinancial debt, the monetary base, and
the broad aggregate L.
Figure 2.1 graphs the response of the employment growth
forecast to a one-standard deviation change in information about
last period's indicator.

For all of the monetary indicators, a

positive impulse eventually leads to a positive growth of
employment.

For most of these indicators, however, the

imprecision of these forecasts is large enough so that the
response is either not statistically significant for most of the
response path (nominal Ml, M3, L and nonfinancial debt) or
entirely insignificant (both monetary bases).
have similar response patterns:
response being a bit earlier.

Real Ml and M2 all

persistent and quick, with the Ml

The responses of nominal M2 and the

NBR/TR ratio are also persistent with a bit more raggedness than
the responses to the real aggregates.
the longest significant response.

The NBR/TR ratio also has

For all of the indicators and

economic activity variables, the maximum one-period impact occurs
within one year (reported in Table 2.3).
Tables 2.4 and 2.5 rank the indicator forecasts for m sample and out-of-sample forecasts at various horizons.

Turning

to Table 2.5 first, notice that for the one-month forecast horizon
for both industrial production and employment, the best forecast
is one without any monetary indicators.

For GDP there are four

indicators which consistently provide additional information for
forecasts:

real M2 (which is always first), the NBR/TR ratio

(always second), nominal M2 and real Ml.

These indicators are

also useful for industrial production and employment for six-month
horizon and beyond.

They are also the highest ranked indicators

in Table 2.4 for the in-sample forecasts.
The monetary aggregates which consistently provide no
additional predictive power beyond the no-indicator model in the
out-of-sample rankings are the two monetary base measures, nominal
Ml, and L.




They also do poorly in the in-sample rankings.
-23-

This

Evans, Strongm, and Eugeni
lack of information is stable across forecast horizons.
The cumulated residuals from the Kalman forecasts shown in
Figure 2.2 provide another perspective of the out-of-sample
performance of our family of money based measures.

In our case,

the best indicator is again real M2 as its cumulated residuals
path clearly stays near zero values, except for isolated periods
of large forecast errors in 1978 and 1981, when real M2
underforecasted economic activity. Real M2's performance was again
noticeably good between 1990 and 1991, when most of the other
money based indicators clearly failed to predict the recession.
The NBR/TR ratio was relatively stable from 1973 to 1981, but has
shown a consistent pattern of overforecasting output growth since
1982.

This deterioration may be due to increasing reluctance on

the part of banks to borrow from the discount window.

The

performance of other monetary aggregates is less reliable and
clearly more volatile than the behavior of real M2 and the NBR/TR
ratio.

For example, the two measures of the monetary base and Ml

consistently underforecasted real GDP between 1974 and 1977, as
shown by their upward sloping paths.

Overall, the path of nominal

aggregates plunged during the credit control program of 1980,
overpredictmg output growth during the mild recession.

From 1983

to 1988, these nominal aggregates performed fairly well,
exhibiting uncharacteristic stability, except for Ml which did
substantially worse between 1983 and 1984.

Finally, between 1990

and 1991, there was a considerable deterioration in the
performance of Ml, L, and the two measures of the montary base, as
they consistently overpredicted economic growth.
Table 2.6 reports the encompassing results for GDP.

For

each of the forecast horizons, we find that real M2 is not
dominated by any of the other forecasts (reading across the real
M2 row, the hypothesis is always rejected at low marginal
significance levels).

None of the other indicator forecasts can

cover the information contained in real M2.

Furthermore, the real

M2 forecasts cover the information contained in all of the other
indicator forecasts (reading down the real M2 column, the
hypothesis that real M2 covers each forecast is not rejected).




-24-

Evans, Strongm, and Eugeni
Therefore, real M2 is a dominant indicator within the class of
monetary indicators selected here for GDP.5




-25-

TABLE 2.1 - DESCRIPTIVE STATISTICS

QUARTERLY (Jan 62 - Dec 91)

MONTHLY (Jan 62 - Feb 92)

Mean

Std. Dev,

Correlation with
Real GDP

-0.058

6 784

2.195

0.034

-0.021

-0.027

6.662

2.282

0.013

5.864

0 005

-0.033

6.055

3.730

0.157

7.750

4.082

0.119

0.013

7.769

3.292

0.236

M3

8.323

4.072

0.113

0.092

8.363

3.520

0.246

L

8.138

3.662

0.167

0.175

8.183

3.057

0.239

DBTNF

8.977

2.752

0.175

0.290

9.017

2.446

0.180

M1R

1.085

7.245

0.063

0:009

0.971

5.143

0.297

M2R

2.675

5837

0.156

0.053

2.685

4.868

0.353

NBRX

0.976

0.027

0.059

-0.026

0.983

0029

0.154

Correlation with
Industrial
Employment
Production

Indicator

Mean

Std.Dev.

MBSTL

6.785

3.617

-0 014

MB

6.710

3.321

M1

6.160

M2

i
i




TABLE 2.2 - CLASSICAL GOODNESS-OF-FrT STATISTICS

QUARTERLY (Jan 62 - Dec 91)

MONTHLY (Jan 62 - Feb 92)

MONTHLY (Jan 62 - Feb 92)

GDP

EMPLOYMENT

INDUSTRIAL PRODUCTION

___

R2

Change
lnR2

SEE

P-Value

Rank

R2

Change
lnR2

SEE

P-Value

Rank

R2

Change
lnR2

SEE

P-Value Rank

MBSTL

0.225

0029

9169

0 3997

8

0 393

0 019

2 363

0.5618

9

0.166

0 049

3 532

01744

7

MB

0222

0027

9182

0 4760

9

0.389

0 015

2 370

0.7445

10

0.145

0027

3 577

0 4734

9

M1

0 221

0026

9188

05172

10

0 401

0 028T

2.346

0.2192

8

0.172

0.055

3519

01284

5

M2

0.252

0057

9 003

0.0144

3

0442

0 069

2.264

0.0001

2

0.219

0101

3419

00084

4

M3

0229

0.033

9144

02755

7

0 412

0 039

2.324

0 0383

5

0.169

0 052

3525

01483

6

L

0.236

0041

9.100

01246

5

0 404

0.030

2.341

0.1486

6

0.164

0046

3538

01993

8

DBTNF

0.231

0036

9.128

0.2092

6

0.401

0.028

2.346

0.2156

7

0.124

0006

3 620

09352

10

M1R

0.244

0048

9 054

0.0477

4

0418

0044

2.314

0.0148

4

0.250

0.132

3351

0 0012

2

M2R

0284

0 089

8808

0.0001

1

0444

0.071

2.260

0.0001

1

0.346

0.228

3128

00000

1

NBRX

0 277

0.081

8854

0.0003

2

0 426

0.053

2.297

0 0028

3

0.249

0.131

3.352

0 0012

3

Indicator

•27-







TABLE 2.3 - MAXIMUM IMPACT OF DYNAMIC MULTIPLIERS

QUARTERLY (Jan 62 - Dec 91)

MONTHLY (Jan 62 - Feb 92)

MONTHLY (Jan 62 - Feb 92)

GDP

EMPLOYMENT

INDUSTRIAL PRODUCTION

Months to
Max

Max Impact

Std. Dev.
at Max

Months to
Max

Max Impact

Std. Dev.
at Max

Quarters to
Max

Max Impact

Std. Dev.
at Max

MBSTL

4

1.054

0489

8

0 221

0148

2

0 787

0 323

MB

10

0.894

0508

5

0192

0.138

2

0 410

0300

Ml

7

1.145

0.515

3

0 370

0.126

2

0 671

0328

M2

7

1.755

0 476

9

0 705

0.137

2

0904

0309

M3

7

1.393

0538

9

0 500

0149

3

0 787

0344

L

7

1655

0507

9

0458

0.143

3

0 739

0333

DBTNF

2

1.214

0447

5

0 332

0.126

4

0113

0 301

MIR

7

1.371

0.513

5

0 485

0.124

2

1011

0 321

M2R

7

1567

0464

5

0 568

0.128

2

1069

0 289

NBRX

12

1.467

0496

8

0 449

0.148

3

1047

0308

Indicator

TABLE Z4 - MULT1PERIOD FORECASTS (In-Sample)

Indicator

l

1MON
R2 RANK

MONTHLY (Jan 62 - Feb 92)

MONTHLY (Jan 62 - Feb 92)

INDUSTRIAL PRODUCTION

EMPLOYMENT

3MOS
R2 RANK

6MOS
R2 RANK

12MOS
R2 RANK

1MON
R2 RANK

3MOS
R2 RANK

QUARTERLY (Jan 62 - Dec 91)
GDP

6MOS
R2 RANK

12MOS
R2 RANK

10TR
R2 RANK

2QTRS
R2 RANK

4QTR$
R2 RANK

MBSTL

0225

8

0230

8

0144

10

0113

10

0393

9

0506

8

0434

8

0274

10

0166

7

0154

8

0102

8

MB

0222

9

0226

9

0145

8

0135

8

0389

10

0499

9

0424

9

0276

9

0145

9

0144

9

0121

5

Ml

0221

10

0259

7

0199

6

0127

9

0401

8

0523

7

0454

7

0288

7

0172

5

0183

7

0096

10

M2

0252

3

0335

3

0333

3

0268

4

0442

2

0581

2

0540

2

0394

4

0219

4

0249

4

0186

4

M3

0229

7

0274

5

0211

5

0165

5

0412

5

0546

5

0487

5

0324

5

0169

6

0189

5

0107

7

L

0236

5

0269

6

0195

7

0146

7

0404

6

0527

6

0461

6

0291

6

0.164

8

0184

6

0097

9

DBTNF

0231

6

0225

10

0144

9

0151

6

0401

7

0496

10

0420

10

0285

8

0124

10

0133

10

0121

6

M1R

0244

4

0320

4

0304

4

0282

3

0418

4

0557

4

0517

4

0 415

3

0250

2

0288

3

0244

3

M2R

0284

1

0413

1

0478

1

0567

1

0 444

1

0609

1

0604

1

0573

1

0346

1

0447

1

0514

1

NBRX

0277

2

0365

2

0386

2

0417

2

0426

3

0562

3

0525

3

0446

2

0249

3

0327

2

0292

2

NONE

0196

11

0201

11

0115

11

0097

11

0373

11

0489

11

0414

11

0269

11

0118

11

0123

11

0076

11

i




TABLE 2 5

KA11AAN MULTIPfniOO FORECASTS (OiH^>l San**>)

MONTHLY (Jul 73 •F«b92)
INDUSTRIAL PRODUCTION
1MON
kxfcatar

RMSE RANK

MBSTL

10406

MB
Ml

10275
10 452

B
5
10
6

M2

10 366

M3

10447

9

3MOS
RMSE RANK
8317
8 142
8 218
7780
8 161

It
5
10
3

12MOS

1MON

RMSE RANK

RMSE RANK

7468

11

5 736

8

6

5 594

7273
6648

8
3

3MOS
RMSE RANK

CMOS
RMSE RANK

12MOS

1QTR

2QTRS

RMSE RANK

RMSE RANK

RMSE RANK

.4QTRS
RMSE RANK

2583

6

2040

to

2058

11

2 012

9

4 108

7

3 474

10

2904

8

7

2553

5

2011

6

2023

7

1986

7

4 114

8

3426

7

2840

7

5 760

9

2587

8

2022

7

2030

8

2023

10

4 149

10

3455

9

2992

11

5 425

4

2585

7

1931

3

1896

3

1894

3

3944

3

3252

3

2809

4

8

4 073

5

3394

6

2948

10

11

4 136

9

3432

8

2926

9

1955

6

4 242

11

3 495

11

2820

6

4 097

6

3285

4

2 775

3

3674

1

2844

1

2219

1

2

3003

2

2550

2

5

2006

6

1997

2028

8

^050

10

2043

2034

9

2041

11

2064

11

2004

5

1905

4

3

1838

1

1731

1

1558

•

2

1903

2

1628

2

1711

8

7308

9

5 843

10

2610

9

7382

10

5843

11

2630

10

5 498

6

2535

4

10405

7

8 174

9

DBTNF

10112

4

8 157

6

7 270

7

MIR

10630

11

8 159

7

6952

4

5309

3

2671

M2R

10111

3

7368

2

5902

1

4229

1

2483

NBRX

9947

2

7362

1

6096

2

4594

2

2477

L

GDP

EMPLOYMENT

CMOS
RMSE RANK

7244

QUARTERLY (Jul 73 - &o c 9 l )

MONTH Y (Jul 73 r < * » 2 )

1998

9

2

3799

I

o
I




3358

§3
X

>

' - O O O O O O O O O

^8

i "» S trt

8 S$2> 8 8 3 © .

I ^- c\» r ^

0) ^

^

s

» S <r>

o o o

© oo

0 0 *•"* "m CO f*» ^~

o o o o o o o o o o

I

t ^ o n o t n w
o o o o o o o o

co CM oo oo r»» CM o>

i

©

o o o o o o o o

§

c

S S 5 > 0 ^ r ^ Q O |

m

c o ^ o ^ f ^ c n c o c o

™
c

M N - O f l O ^ J O N

CO

o o o o o o o o

co
c

SK8gRf2

O l C O C O ' V O O l O D O O n

o o o o o o o o

c

o o o o o o o

R t £ 8

o> o> oo (C co o> ^» o *"*

o o o o o o o o o o

P ^ o o o o 5 o > o o S o 3 J?^r
c

o o o o o o o o

o

©

( 0 CO CO CO 'Cf T
CD Q N W N N
2> CD — CO CM K

o o o o o o o

fc

~- o o o o o o

!

lie

I

! !

i

i

c

o o o

C

' » »

1 1 rc i

o
o

©

c

l i^ 1 l

I I
I I I
1 t (0 1 I I

l £ l l

N
-. 8
-J *S
<c co m

c

t-s

^

83 i ! S«8 ! i !

©

o

o

©

fc

©

.

I l/)CVI
I
I N W ffl I

' co <£> ™ '

O) .

CM

O ,

,

,

s

5! 3 28 ! ! !

o

o

o *- o

«

°1
CO tO if)

i

i *

i

o

o

o

o oo

© ©

^ ^ **"•
CO CD ~ •- w tc n o in o> i
00 <D <£> ) : tD OJ <£)

^•^ *•»» c *_. ^ s —^ •

1 « / W

tf) 00 O) I

i £ 2 I ; 88

o o o

Ul i

? 8 • ! 2 2 $ !I
o o

I

c

S 2

c

o oo

§2

'SS!

.S88

IOMO)
*vi « s * v # « »/•» rvi

o oo

o oo

8
o

5 <^ CO ' 3
CM 85 £^ I
co *- ^fc I -v8«! o i
o o
o o o

!S
o

E

8-

c

i

8

i

I

I

c

i

i

I
I

u->
I £ I I I
I o • • I

I
I

l
I

l 8 I
I "* I

s

8§R i
o o o i

o
a
o

»a
r- o

o o

8

S 2 5 i

'-'-«> i

o oo

I
I

o

l i
l i

re co 5 I vn CM 8
:: co <o i ** \n *-

.2
5

51




z
x
P cc cc cc

en
CD CO ^

2 2 2

eo

CD * - C M CO

w

? _
cc cc cc

CD CD * - CM o

CD ^

- _

2 2 2 2 2 - J Q 2 2 Z

2 -J 0 2 2 Z

-31-

S2 ~
O

CD i - N

P
P ccc
c ccc
c cic
n

CD <- CM GO

2 2 2 2 2-J Q 2 2

Z

o
z

2.1. Dynamic Response of Employment to Money Based Measures
St. Louis monetary base (MBSTL)
annualized percent growth rates
050 r




10

Nominal L (L)
annualized percent growth rates
0 75 r

15
20
months

15
20
months

-32-

25

2.2. Money Based Measures: Cumulated Kalman Residuals in Forecasting Real GDP
St. Louts monetary base (MBSTL)

Nominal L (L)

cumulated Kalman residuals
75 i -

cumulated Kalman residuals
75

50 h

50

25

25

•25

-25
» i

•50

i

i

i

i

i

i

i

i

i

i

i

i

i

i

i « i

FRB monetary base (MB)
75

Nominal nonfinancial debt (DBTNF)
75

r

50 J-

50

25 h

25

0
-25 h

v v"
t

i

i

i

i

r
i

I 1 I I I I I I I I I I I I I I I 1 I

•50

f

v\/-~~v

j k/^ X/

i

•50

i

t

t i

i

i

i

i

i

i f

Nominal M1 (Ml)
75 r

L„l

I I I I

I I I

I

1 I I 1 1 1 I 1 I

RealMI (M1R)
75

r

50 f25 U

-25
•50

Kr'Vv^v-v
•

i

i

i

i

i

t

i

i

i

i

i

i

i

i

i

i i

i

Nominal M2 (M2)

Real M2 (M2R)

50

100 r-




V ^ / v
•25 '

'

'

'

'

'

'

» '

*

'

^KT
'

'

'

'

'

'

I

I I
"85

I

I I
"88

'

*

'

NBR/TR ratio (NBRX)
25

1973

76

79

r

1973

-33-

•J I I
76

I

I
79

'82

'91

Evans, Strongin, and Eugeni

INTEREST RATE SPREADS
Recent research on financial market indicators of economic
activity have brought renewed attention to interest rates spreads.
Laurent (1988), Bernanke (1990),
Friedman-Kuttner (1992),

Estrella and Hardouvelis (1991),

Kashyup-Stein-Wilcox (1991), and Stock-

Watson (1989) have suggested and tested various interest rate
spreads as predictors of economic activity with significant
success.

The idea behind most of these spreads is that the

difference in yields between two different debt instruments
provides information beyond that in the level of interest rates.
The two primary types of interest rates spreads that have been
used are risk-spreads which measure the difference in yield
between a private debt instrument and the yield on a government
bond of equivalent maturity and term-spreads which measure the
difference in yield of government debt instruments of different
maturities.
Typically, the motivation for the risk spreads is that the
risk in the private debt instrument is a measure of the market's
assessment of the near term risk in the relevant business
environment and that high risk implies a tough time for business
ahead.

Friedman-Kuttner have argued that this interpretation is

probably flawed since the spreads are typically too large to be
explained by any reasonable estimate of the risk inherent in the
private debt instruments and suggest that liquidity considerations
play a significant role in the pricing of public-private spreads.
Following their lead, we also will refer to these spreads as
public-private spreads.
The term-spreads seek to measure the relative availability
of credit through time.

The convention is that the shorter

maturity yield is subtracted from the longer.

Thus, a positive

spread would indicate that short term funding is available at a
lower rate than longer term funding.

The normal interpretation is

that if short-term funds are especially cheap relative to longterm funds this will encourage borrowing and economic activity.
An alternative explanation is that the higher long-term yields are




-34-

Evans, Strongm, and Eugeni
signaling expectations of higher future credit demand resulting
from increased economic activity.

A third interpretation is that

by taking the difference between a short- and long-term interest
rate you are correcting the shorter term rate for changes in
inflationary expectations and taxes, leaving a better measure of
short-run credit conditions.

In any case, all of these term-

spread measures have the counter-intuitive implication that a rise
in long-term interest rates is good for the near-term outlook of
the economy. Estrella and Hardouvelis (1991) and Strongin (1990)
attempt to reconcile the term-spread results with current theory
with limited success.
We test 3 public-private spreads and 5 term-spreads5.

The

specific measures we use are the TED or Eurodollar spread which is
the 3-month Eurodollar rate minus the 3-month Treasury bill rate.
The Commercial Paper spread which is the 6-month Commercial Paper
spread minus the 6-month Treasury bill rate, and the Baa spread
which the Baa yield minus the 10-year Treasury bond rate6.

The

five term-spreads contain three spreads based on the Federal Funds
Rate, a short, medium and long spread -- the short spread is the
3-month bill rate minus the Federal Funds rate -- the medium
spread is the 12-month bill rate minus the Federal Funds rate -the long spread is the 10-year bond rate minus the Federal Funds
rate.

There are two intermediate spreads as well the 12-month/3-

month spread and the 10-year/l-year spread.
Table 3.1 shows that as expected the public-private spreads
all show a strong negative correlation with economic activity and
the term-spreads all show a positive correlation with activity:
the shorter the term-spread, the higher the correlation.
Table 3.2 indicates that based on classical measures of fit

5. These are the only commonly used spreads available for the
entire data sample. Other spreads are examined in the appendix for
shorter samples, but the results are no different and the publicprivate spreads presented either dominate or at least equal any of
those presented in the appendix.
6. The 10-year rate is used because the 7-year which might be
preferred is not available for a sufficient time span. The appendix
also includes the Baa/7-year spread and spread's using the AA bond
yield.




-35-

Evans, Strongin, and Eugeni
all of the spreads do fairly well in explaining all three measures
of economic activity.

The R2s for industrial production range

from .236 to .339; the range for employment growth is .416 to
.459; and the range for GDP is from .234 to .339.

With the

exception of the 12-month/3-month term-spread, every spread
Granger causes activity at a high level of significance.

The only

exception is the 12-month/3-month spread which fails to Granger
cause industrial production.

The public-private spreads do a

better job of predicting employment and industrial production with
the Commercial Paper spread and the Baa spread ranking 1 and 2.
For GDP the results are more mixed with the Commercial Paper
spread and 12-month/Federal fund spread coming in 2nd.
The dynamic response path graphs in Figure 3.1 shows
substantial difference in the dynamic response of employment
growth by type of spread.

The response of employment to an

increase in the Baa spread shows a quickly rise, peaking at only 3
months.

The response then dies just as quickly.

The response

paths for the two shorter-term public-private spreads the
Commercial Paper spread and the Eurodollar spread build rapidly
then plateau for a number of months and then die quickly.

The

term-spread response paths, with exception of the 12-month/3-month
spread, all build slowly, peak and then slowly die out.

Only in

the case of the Baa spread is there a well-defined peak in the
response path, all of the other spreads show extended period of
impact.

This would suggest that the strength of the Baa spread

will be in very short horizon forecasts, the strength of the
Commercial Paper and Eurodollar spreads will be at short and
middle horizons, while the strength of the term-spreads will be in
longer term forecasts.

Table 3.3 suggests similar conclusions

with the Baa spread showing the quickest, largest and most tightly
estimated peak for employment and industrial production.

The

longer horizon GDP results show the impact of the Baa spread
falling off considerably though it is still very quick.
Tables 3.4 and 3.5 strongly re-enforce these conclusion and
provide some startling evidence on the effect of forecast horizons
on indicator performance.




First, in table 4 it is clear that the
-36-

Evans, Strongm, and Eugem
performance of the Baa spread falls off dramatically as the
forecast horizon is increased.

Ranking 2nd for industrial

production and employment at the one-month horizon, the rank drops
to 6th and 7th for industrial production and employment
respectively for the three-month horizon and is dead last by six
months for all measures of activity.

The Commercial Paper spread,

on the other hand, does very well ranking 1st until the one-year
horizon in both employment and industrial production, when it is
superseded by a number of term-spreads.

In forecasting GDP the

Commercial Paper spread still does very well at the one-quarter
horizon, but fades quickly falling to 4th at the six-month horizon
and 6th at the one-year horizon.

The Federal Funds rate based

spreads do very well as the forecasting horizon lengthens.
Starting out in the middle to back of the pack at the shortest
horizons they rise to dominate the top of the ranking at the oneyear horizon with the 12-month/Federal Funds spread rising to 1st
for all three measure of activity.

The intermediate spreads

rarely do well.
Table 3.5, showing the out-of-sample results shows a very
similar story in terms of rankings.

The interesting additional

fact is how well the spread models stand up to the no-indicator
model.

At every horizon except one-month the spread models

strongly outperform the no-indicator model, though at the onemonth horizon the no-indicator model does outperforms all of the
spread models except the Baa spread, which is only good at short
horizons.

Clearly the forecast horizon is extremely important to

the evaluation of interest rate spread models.
The cumulated residuals from the Kalman forecasts in Figure
3.2 show some striking similarities in the overall forecasting
performance of our family of interest rate spreads.

Except for

the 3-, 6-, and 12-month/Federal Funds rate spreads, all of our
spreads tend to overforecast real GDP, as shown by their
consistently negative residuals.

While the 3-, 6-, and 12-

month/Federal Funds rate spreads performed fairly well from 1973
to 1980, they clearly failed during the last three recessions.

In

fact, they all underforecasted economic activity between 1980 and




-37-

Evans, Strongin, and Eugeni
1982, and then overpredicted real GDP between 1990 and 1991.
Between 1982 and 1989, their path was conspicuously flat.

This

suggests that these spreads do well in forecasting
normal

periods of economic activity, but periodically fail in

predicting recessions.

Although the 5-year/ and 10-year/Federal

Funds rate spreads follow a similar pattern between 1973 and 1981,
after 1982 their cumulated residuals path never stabilized but
plunged to persistently negative values.

Our intermediate term

spreads (12-month/3-month and 10-year/1-year spreads) failed
during all of the recessions

in our sample period (including the

1973-1975 recession), and developed a consistently negative bias
after 1982, as they clearly overpredicted real GDP.

All of the

private/public spreads followed the same general pattern of
mediocre performance from 1973 to 1981, and persistent
overprediction of economic activity thereafter.

In general, we

conclude that, although a persistent bias in forecasting exists in
all of the interest rate spreads we investigated, some of them
did fairly well during most of our sample period,

but failed

during periods of large scale financial restructuring.
The encompassing tests in Table 3.6 are exactly what would
be expected given the previous results.

To fully encompass all of

the information in the interest rate spreads it is usually
necessary to include both a public-private spread and a termspread.

Also not surprisingly, the Commercial Paper spread and

the 12-month/Federal Funds rate spreads dominate their respective
groupings at the one- and two-quarter horizons.

It is interesting

to note that the Stock-Watson leading indicator index, which was
designed to fit data at the six-month horizon, chose the
Commercial Paper spread and the 10-year/l-year spread.

For our

sample period, the 12-month/Federal Funds spread narrowly
dominates the 10-year/l-year spread.

At the 4-quarter horizon the

public-private spread no longer contains additional information
beyond that contained in the 12-month/Federal Funds spread.

The

12-month/Federal Funds spread, however, does not dominate since it
fails to cover (only) the 10-year/l-year spread.

We selected the

10-year/Federal Funds spread since it covers more spreads than the




-38-

Evans, Strongm, and Eugeni
10-year/l-year spread, covers the 10-year/l-year spread, and
performs better out-of-sample.

The selection of two term spreads

is consistent with the previously noted results that the publicprivate spread do not contain as much long run information as the
term-spreads.

It is interesting to note that examination of the

entire encompassing results indicate that the separation between
the public-private spreads and the term-spreads is not very clear.
At some horizons some term-spread encompass some public-private
spreads while at other horizons the results reverse.

This would

seem to indicate that there are common multiple driving forces in
the determination of these spreads and that those driver factors
associated with longer horizon economic activity predominate in
the term-spreads with the common factors that drive short-run
performance dominate the public-private spreads.




-39-

TABLE 3.1 - DESCRIPTIVE STATISTICS

QUARTERLY (Jan 62 - Dec 91)

MONTHLY (Jan 62 - Feb 92)

Mean

Std. Dev.

Correlation with
Real GDP

0.252

-0.750

0.807

0449

0.292

0.255

-0 593

0.886

0.442

1.149

0 291

0.254

-0.580

1.082

0.425

0.384

1586

0.210

0.123

0.373

1511

0.321

CM10FF

0.514

1.791

0 200

0.114

0.499

1.713

0.309

TB12TB3

0.170

0.468

0.177

0.158

0.170

0.434

0225

CM10CM1

0.601

1.036

0083

0.001

0.588

0.997

0.170

EUR0TB3

1.428

0.931

-0.235

-0.248

1.434

0.885

-0378

CP6TB6

0579

0.489

-0.305

-0.297

0.582

0.461

-0.431

BAACM10

1.722

0 698

-0 248

-0.390

1.720

0.690

-0.297

Correlation with
Employment
Industrial
Production

Indicator

Mean

Sid. Dev.

TB3FF

-0747

0.864

0.291

TB6FF

-0 591

0.948

TB12FF

-0.577

CM05FF

i

.**
o
i




TABLE 3.2 - CLASSICAL GOODNESS-OF-Ftf STATISTICS

MONTHLY (Jan 62 - Feb 92)

MONTHLY (Jan 62 - Feb 92)
INDUSTRIAL PRODUCTION

GDP

EMPLOYMENT

R2

Change
lnR2

SEE

P-Value Rank

3

0.327

0.209

3.174

0.0000

3

00012

5

0.321

0204

3.187

0.0000

4

2 291

0 0016

6

0.330

0.212

3.166

0.0000

2

0.042

2.318

0.0224

10

0.302

0.185

3.231

0.0000

6

0.416

0 043

2.316

0 0186

9

0.309

0191

3.216

0.0000

5

10

0 424

0.050

2.302

00047

7

0.238

0.120

3.377

0.0026

9

0.0207

9

0.417

0 044

2.315

0.0163

8

0.284

0.166

3.273

0.0001

8

8 870

0.0004

6

0.431

8.058

2.286

0.0009

4

0294

0.177

3.250

0.0001

7

0.144

8.462

0.0000

1

0 459

0.086

2.229

0,0000

1

0.339

0.221

3.145

0.0000

1

0.108

8.691

0.0000

2

0.437

0.064

2.274

0.0002

2

0234

0.116

3.386 0.0033

10

R2

Change
lnR2

SEE

3

0 435

0.062

2.278

0 0004

0.0002

4

0.430

0.057

2.289

8.866

0.0004

5

0.429

0 055

0.061

8.981

0.0084

7

0 415

0.254

0.059

8.992

0.0111

8

TB12TB3

0.236

0.041

9.099

0.1224

CM10CM1

0250'

0.054

9.018

EUROTB3

0.274

0 079

CP6TB6

0.340

BAACM10

0.303

R2

Change
lnR2

SEE

TB3FF

0 291

0 095

8 769

0.0000

TB6FF

0.280

0.085

8.834

TB12FF

0.275

0.079

CM05FF

0.256

CM10FF

Indicator




QUARTERLY (Jan 62 - Dec 91)

P-Value Rank

P-Value Rank

TABLE 3.3 - MAXIMUM IMPACT OF DYNAMIC MULTIPLIERS

MONTHLY (Jan 62 - Feb 92)

QUARTERLY (Jan 62 - Dec 91)

MONTHLY (Jan 62 - Feb 92)

GDP

EMPLOYMENT

INDUSTRIAL PRODUCTION

Months to
Max

Max Impact

Std. Dev.
at Max

Months to
Max

Max Impact

Std. Dev.
at Max

Quarters to
Max

Max Impact

Std. Dev.
at Max

TB3FF

7

1591

0483

6

0.500

0.127

3

1.408

0.310

TB6FF

7

1.731

0468

6

0466

0.128

3

1.283

0 277

TB12FF

7

1.446

0466

6

0.412

0.125

3

1.259

0.285

CM05FF

7

1.091

0478

15

0.355

0.087

3

1.195

0.294

CM10FF

7

1.022

0.497

9

0.370

0.134

3

1.243

0.293

TB12TB3

14

1.375

0382

14

0534

0.111

5

0879

0 282

CM10CM1

5

1.471

0486

9-

0.342

0.134

3

1.243

0.296

EUROTB3

7

-2.083

0515

12

-0.553

0.146

3

-1.493

0.338

CPCTB6

9

-2.476

0480

8

-0.711

0.136

3

-1.449

0.312

BAACM10

3

-2.645

0473

3

-0.519

0.121

2

-0 833

0.312

Indicator

i

£




'

TABLE 3.4 - MULT1PERI0D FORECASTS (In-Sampla)

Indicator

1MON
R2 RANK

MONTHLY (Jan 62 - Fab 92)

MONTHLY (Jan 62 - Fab 92)

INDUSTRIAL PRODUCTION

EMPLOYMENT

3MOS
R2 RANK

6MOS
R2 RANK

12MOS
R2 RANK

1MON
R2 RANK

3MOS
R2 RANK

QUARTERLY (Jan 62 - Dae 91)
GDP

6MOS
R2 RANK

12MOS
R2 RANK

10TR
R2 RANK

2QTRS
R2 RANK

4QTRS?
R2 RANK

TB3FF

0291

3

0402

2

0477

2

0524

3

0435

3

0591

3

0585

4

0549

6

0327

3

0446

3

0437

5

TB6FF

0280

4

0382

3

0456

3

0547

2

0430

5

0590

4

0595

3

0506

2

0321

4

0 459

2

0490

4

0275

5

0369

S

0438

4

0555

1

0429

6

0593

2

0604

2

0526

1

0330

2

0 470

1

0518

1

0334

7

0381

7

0488

4

0415

10

0569

7

0569

6

0585

4

0302

6

0 428

6

0498

2

6

0484

5

0.416

9

0572

5

0574

5

0567

3

0309

5

0 435

4

0491

3

TB12FF
CM05FF

0256

7

CM10FF

0254

8

0337

6

0385

TB12TB3

0236

10

0286

10

0269

9

0358

8

0424

7

0567

8

0563

7

0572

5

0238

9

0333

9

0383

7

CM10CM1

0250

9

0305

9

0296

8

0361

7

0417

8

0552

10

0526

9

0 489

8

0284

8

0374

7

0396

6

EUROTB3

0274

6

0380

4

0413

5

0323

9

0431

4

0571

6

0542

8

0431

9

0294

7

0364

8

0230

9

CP6TB6

0340

1

0500

1

0559

1

0426

6

0459

1

0J634

1

0623

1

0506

7

0339

1

0 429

5

0289

8

BAACM10

0303

2

0324

8

0218

10

0168

10

0437

2

0553

9

0461

10

0293

10

0234

10

0.175

10

0138

10

NONE

0196

11

0201

11

0115

11

0097

11

0373

11

0.489

11

0414

11

0269

11

0118

11

0.123

11

0076

11

-P*
I




TABLE 35 - KALMAN MULTIPERIOO FORECASTS (Out of Sample)

MONTHLY (Jul 73 -Feb 92)

QUARTERLY (Jul 73 • Dec 91)

MONTHLY (Jul 73. Feb 92)

GDP

INDUSTRIAL PRODUCTION

Indicator

1MON
RMSE RANK

• MOS
RMSE RANK

12MOS
RMSE RANK

1MON
RMSE RANK

TB3FF

10161

5

7 421

3

5 763

2

4209

3

2581

5

TBSFF

10371

6

7768

4

6075

4

4081

1

2 624

7

TB12FF

10 582

8

8 164

8

6539

5

4 207

2

2 658

9

CMD5FF

10931

9

8386

9

6 742

7

4 517

4

2 708

10

CM10FF

10970

10

8 390

10

6656

6

4 542

5

2 740

11

TB12TB3

11054

11

8539

11

7169

11

4906

7

2653

8

CM10CM1

10 572

7

8 149

7

6891

8

5053

9

2622

6

EUROTB3

9898

3

7289

2

5937

3

5048

8

2 512

3

CPCTB*

9930

4

6703

BAACM10

9858

1

7 785

3MOS
RMSE RANK

3MOS

$MOS

12MOS

1QTR

2QTRS

4QTRS

RMSE RANK

RMSE RANK

RMSE RANK

RMSE RANK

RMSE RANK

RMSE RANK

1937

4

1785

3

1599

6

3609

1

2674

1

2253

5

I960

6

1775

2

1492

3

3691

3

2691

2

2081

2

1986

7

1791

4

1435

1

3 753

6

2 754

3

2015

1

2055

9

1866

8

1504

5

3 745

5

2811

6

2111

3

2063

11

1853

6

1495

4

3763

7

2785

5

2161

4

2030

8

1866

7

1479

2

4 197

11

3187

9

2 370

6

2055

10

1937

9

1674

7

3857

8

2970

8

2389

7

9

3698

4

2886

7

2 721

8

2

Z760

4

1926
1

4922

1

4806

6

2543

7025

9

5 575

11

2446

1834

5

1779

1632

1

1710

8

3656

2 744

9

1913

2

1972

11

1998

11

3963

9

3485

11

2846

11

1948

5

1953

10

1913

10

4 015

10

3 358

10

2819

10

1 790
5

3

4
1

1

I
J>
I




2463

2

TABUE 3.6 - MULUPERIOD ENCOMPASSING TESTS (Sample Period: Jan 62 - Dec 91)
Probability Value for Null Hypothesis: X is Encompassed by Y
GDP 1 qtr
Y

TB3FF

TB6FF

TB12FF

CM05FF

na
0462
0105

0 185
na
0 227
0 109
0062
0 333

0167
0 999
na
0389
0 231
0 797
0 106

—
—
—

CM10FF

TB12TB3

CM10CM1

EUROTB3

CP6TB6

BAACM10

Maximum
P Value

X __
TB3FF
TB6FF
TB12FF
CM05FF
CM10FF
TB12TB3
CM10CM1
EUROTB3
CP6TB6
BAACM10

—
—
0093

—
0125
0066

—

—
—
—

—
—

na
0215
0 430
0 454

—
—

—
—
—

—
—

0 540
na
0413
0 699

0098
0077
0115
na

__
—
—

—
—

--

0 186
na
0 104

0 185
0999
0 227
0 540
0 231
0 797
0699
0 186
0066
0 104

0 222
na
0936

0569
0798
0 155
0665
0 271
0 755
0 876
0 222
0 093
0 991

0560
na
0 575

0548
0197
0 027
0 176
0 720
0 576
0593
0989
0883
0 973

0 053

—

—
0055

—

na
0 072

0 156
0 109

—

0053

—

GDP: 2qtrs

i

.£*

en
i

TB3FF
TB6FF
TB12FF
CM05FF
CM10FF
TB12TB3
CM10CM1
EUROTB3
CP6TB6
BAACM10

na
0070

—
—
—
—
—
0214
0093
0545

0 569
na
0155
0092
0055
0256

0 467
0 798
na
0337
0206
0 755
0126

na
0 271
0 370
0 876

0 665
na
0 353
0 710

—
—
na

—

0071
na
na

0 459

0 580

0908

0991

0436

0935

0 807

GDP 4qtrs
TB3FF
TB6FF
TB12FF
CM05FF
CM10FF
TB12TB3
CM10CM1
EUROTB3
CP6TB6
BAACM10

na

—
—
—
—
—
—
0752
0883
0500

0 322
na

0142

0989
0 870
0 392

0 548
0197
na
0 176
0144
0 576
0062
0979
0840
0442

NOTE* Values less than or equal to 0 05 are marked with a dash




0131
0056

0 092

na
0 720
0333
0 593
0899
0 774
0863

0 170
na
0261
0 230
0965
0 783
0930

—

—__
—

__
—
na

—
0094

—
0111

na
0428
0115
0973

0 569

3.1. Dynamic Response of Employment to Interest Rate Spreads
1 year T bill less 3 month T bill (TB12TB3)
annualized percent growth rata*
0 75 r

3 month T bill less fed funds (TB3FF)
annualized parcant growth rates

060

r

050

h

0 25

000

11 11 1 1 1 1 1 1

-0.20

1111111

•

' '

*

' ' '

' '

' '

*

» •

-0 25
10 year T bond less 1 year T bond ( C M 1 0 C M 1 )
0 72 r

6 month T bill less fed funds (TB6FF)
060 r

*

*

•

12 month T bill less fed funds (TB12FF)
080

3 month eurodollar less 3 month T bill ( E U R O T B 3 )
0.25 r

5 year T bond less fed funds ( C M 0 5 F F )
0 72 r

6 month commerciai paper less 6 month T bill ( C P 6 T B 6 )
0 25 r

10 year T bond less fed funds ( C M 10FF)
060 r

BAA corporate bond less 10 year T bond ( B A A C M 1 0 )
044 r




i 11 i i i i i i 11 i i
35

-46-

3.2. Interest Rate Spreads: Cumulated Kalman Residuals in Forecasting Real GDP
3 month T bill less fed funds (TB3FF)
cumulated Kalman residuals
100 p

12 month T bill less 3 month T bill (TB12TB3)
cumulated Kalman rtftiduais
75 r

75 k

50 ^

50 U

25 L

25 h

•25
-25

t -i i

i

i

i

t i -t

i

i

i

i

i

i

i

-50

6 month T bill less fed funds (TB6FF)
100 r

I I I I I I t I t I I 1 I I I 1 » f I

10 year T bond less 1 year T bond (CM10CM1)
25

75 k
50 h
25
0

^ V W

•25

I I I I I

I I I I I I

I

111

12 month T bill less fed funds (TB12FF)

.100 ' * ' ' ' '
3 month eurodollar less 3 month T bill ( E U R O T B 3 )
25

100 p
75 I50 U
25 h

^^\A/'
-25

t

,1 I I

l

I I I 1 I 1 1 I I I I

I 1 I

5 year T bond less fed funds (CM05FF)
25 r

6 mo. commercial paper less 6 mo. T bill ( C P 6 T B 6 )
25 r

10 year T bond less fed funds ( C M 10FF)
25 r

BAA corporate bond less 10 year T bond (B A A C M 10)
25 r




0
-25
-50
-75
.100 ' ' '
1973

-47-

' * '
76

' '
79

'

'

'
82

'

' '
*85

'

'
'8

•91

Evans, Strongin, and Eugeni

COMPOSITE INDICATORS
The composite indicator family consists of the NBER experimental
leading indicator series (XLI) and the NBER experimental nonfinancial recession index (XRI2) (which measures the probability
of a recession), the Department of Commerce leading indicators
(lead), the National Association of Purchasing Managers Index
(PMI), the change in the S&P 500 (S&P), changes in sensitive
materials prices (SMPS), and the Kashyap-Stein-Wilcox "mix"
(KSWMIX), which is the ratio of bank lending to the sum of bank
lending and commercial paper lending (see Kashyap et al. (1991)).
It should be noted that the NBER experimental index includes the
10-year/l-year interest rate spread and the Commercial Paper
spread and that the Department of Commerce leading indicator index
includes real M2, which have been used in previous sections.

All

three leading indicator composites are designed to predict
economic activity at a six-month horizon, though the optimization
for the Department of Commerce index is not as specific as either
of the NBER indices.
Table 4.1 shows that most of these series have the expected
correlation with contemporaneous economic activity, except for the
change in the S&P 500 which has small negative correlations with
growth in industrial production and employment and only a small
positive correlation with growth in real GNP.
is positively correlated with real GDP:

The KSWMIX variable

one interpretation of

this correlation is that increased (decreased) bank lending is
associated with expansions (contractions).
Table 4.2 shows that all of these series perform very well
in classical regression analysis.

They all produce high R2s.

The

2

R s for industrial production range from .289 to .391; the range
for employment growth is .434 to .527; and the range for GDP is
from .205 to .455.

Further, each of these indicators Granger

causes activity at a high level of significance.

In terms of

ranking, the Department of Commerce leading indicators and NBER
experimental index are 1st and 2nd for all of the measures of
economic activity,




with the Department of Commerce leading
-48-

Evans, Strongm, and Eugeni
indicators coming in 1st for industrial production and employment
and the NBER experimental index coming in 1st for real GNP.

The

change in S&P 500 comes in last in every category and the change
in sensitive materials prices comes in next to last in every
category.
The dynamic response path graphs in Figure 4.1 show somewhat
similar patterns.7

For all three leading indicators series

—

the NBER leading indicator, the NBER nonfinancial recession index
and the Department of Commerce's leading indicators -- employment
growth shows a rapid rise peaking at 5 months.

From that peak all

three graphs exhibit significantly different behavior.

The NBER

leading indicator graph plateaus for 4-5 months and then drops off
before the end of the year.

Employment's response to changes in

the NBER nonfinancial recession index drops off steadily from the
peak while the response to changes in the Department of Commerce's
response path is in-between with a high initial peak followed by a
stable period then a steady decline.
The response of employment to the changes in the purchasing
managers index and the change in sensitive material prices both
show dramatic jumps in forecasted employment growth peaking at 3
and 2 months, respectively.

Employment growth then falls steadily

in the Purchasing Managers Index graph while it plateaus in the
sensitive material prices graph.

The S&P 500 graph is similar,

showing a leap up followed by a steady decline, except it has a
small initial drop in the first month.

It is interesting to note

that all of these dynamic response paths are barely significant at
the one year mark, despite showing fairly precisely estimated
effects earlier.

As a group these series seem to hold a lot of

information about short-run changes in economic activity, with
most of that information centered at the 3-9 month horizon.
Tables 4.4 and 4.5 which examine the forecasting ability of
these indicators at different forecast horizons, in-sample and
out-of-sample, show very stable rankings as the forecast horizon

7. The dynamic response of employment to the KSWMIX variable
is not reported since it is only available on a quarterly basis.




-49-

Evans, Strongin, and Eugeni
changes.

The leading indicators series do best, posting very

similar performances.

The other series do not do as well, though

the Purchasing Manager's Index does well at the one-month horizon
for industrial production.

Placing 3rd in-sample and 2na out-of-

sample for the one-month horizon then falling off at longer
horizons.

Among the leading indicator series, the Department of

Commerce series does best at horizons of less than six months,
while the NBER index ranks first for horizons of 6 months and
longer.

For GDP, the NBER index always does better with the

differential in performance increasing with horizon.

The KSWMIX

variable does reasonably well in-sample, but out-of-sample it
performs worse than "NONE," the no-indicator forecast.

The one

anomaly in the tables is that the change in sensitive material
prices does very well out-of-sample for GDP at the 4 quarter
horizon, actually outperforming all of the other indicators except
the NBER leading indicator series.
The cumulated Kalman residuals in Figure 4.2 show some
striking similarities and some differences in actual performance
across these indicators.

Except for KSWMIX, all of our composite

indicators have overforecasted real GDP over time, as their
cumulated residuals

are consistently negative.

This bias is

clearly evident during recessions and becomes more dramatic after
1980.

After 1982, while the negative bias is exacerbated in the

NBER leading indicator and S&P 500, the path becomes somewhat more
stable for most of our indicators.

The NBER nonfinancial

recession index is our best performer during this period, which is
not surprising since the index was originally developed in
response to the failure of the NBER leading indicator index to
forecast the 1990-1991 recession.
The encompassing results in Table 4.6 show that for horizons
of two- and four-quarters the NBER index dominates this entire
family of indicators, with the possible exception of the KSWMIX.
At the one-quarter horizon both the Department of Commerce and
NBER nonfinancial recession indices are not encompassed by any of
the other forecasts.

These results are not surprising in light

of the ranking discussed earlier and the fact that the NBER




-50-




Evans, Strongin, and Eugeni
leading indicator index was designed to provide a "best" forecast
of economic activity at a six-month horizon, using virtually all
of the macroeconomic data available.

At the one- and two-quarter

horizons, the KSWMIX is encompassed by the NBER index at the 5%
significance level, but not the 10% level.

We chose not to

include the KSWMIX in the survivor list of indicators due to its
poor out-of-sample performance in Table 4.5.

-51-

TABLE 4.1 - DESCRIPTIVE STATISTICS

MONTHLY (Jan 63 - Feb 92)

QUARTERLY (Jan 63 - Dec 91)

Correlation with
Industrial
Employment
Production

Mean

Std. Dev.

Correlation with
Real GDP

0.429

3.039

4.067

0.547

-0 556

-0.523

0.156

0.131

-0.649

11.148

0.454

0.249

2.993

8.832

0.600

53.380

7.668

0.524

0.681

53.400

7.473

0.632

S&P

6.463

42.410

-0.028

-0.050

6.451

24.588

0.185

SMPS

0.319

0.930

0.325

0.444

0.322

0.912

0.278

0.925

0.040

0.316

Indicator

Mean

Std. Dev.

XLI

3.070

4.162

0.439

XRI2

0.157

0.139

LEAD

2.990

PMI

i

en
i

KSWMIX




TABLE 4.2 - CLASSICAL GOODNESS-OF-FIT STATISTICS

MONTHLY (Jan 63 - Feb 92)

MONTHLY (Jan 63 - Feb 92)

QUARTERLY (Jan 63 - Dec 91)

EMPLOYMENT

INDUSTRIAL PRODUCTION

GDP

R2

Change
lnR2

SEE

P-Value

Rank

R2

Change
lnR2

SEE

P-Value

Rank

R2

Change
lnR2

SEE

P-Value

XLI

0.369

0.165

8.353

0.0000

2

0.484

0.104

2.198

0.0000

2

0.455

0.338

2893

0.0000

1

XRI2

0.338

0.134

8.555

0.0000

4

0.483

0.102

2.200

0.0000

3

0.385

0.268

3.073

0.0000

3

LEAD

0.391

0.187

8.204

0.0000

1

0.527

0.146

2.104

0.0000

1

0.405

0.288

3.022

0.0000

2

PMI

0.355

0.152

8.439

0.0000

3

0.463

0.083

2.241

0.0000

4

0.265

0.148

3.359

0.0005

4

S&P

0.289

0.085

8.864

0.0002

6

0.434

0.054

2.301

0.0029

6

0.205

0.089

3.493

0.0222

7

SMPS

0.300

0.096

8.795

0.0000

5

0.436

0.056

2.297

0.0019

5

0.232

0.115

3.433

0.0045

6

0.243

0.126

3.410

0.0023

5

Indicator

Rank

i

en
OJ
i

KSWMIX




_

_

_

„

.

TABLE 4.3 - MAXIMUM IMPACT OF DYNAMIC MULTIPLIERS

MONTHLY (Jan 63 - Feb 92)

MONTHLY (Jan 63 - Feb 92)

QUARTERLY (Jan 63 - Dec 91)

EMPLOYMENT

INDUSTRIAL PRODUCTION

GDP

Months to
Max

Max Impact

Std. Dev.
at Max

Months to
Max

Max Impact

Std. Dev.
at Max

Quarters to
Max

Max Impact

Std. Dev.
at Max

XLI

6

2.441

0.455

5

0.650

0.124

3

1.731

0.310

XRI2

3

-2.668

0.446

5

-0.811

0.137

2

-1.829

0.309

LEAD

5

2.475

0.453

5

0.766

0.132

2

1.851

0.274

PMI

2

2.655

0.454

3

0.617

0.124

2

1.223

0.314

S&P

5

2.310

0.487

7

0.596

0.145

2

0.935

0.322

SMPS

2

1.801

0.468

2

0.401

0.120

7

-0.679

0.246

KSWMIX

.

.

_

_

_

2

1.021

0.302

Indicator

i

en
i




„

TABLE 4.4 - MULTIPERIOD FORECASTS (In-Sample)

Indicator

1MON
R2 RANK

MONTHLY (Jan 6 3 - F e b 92)

MONTHLY (Jan 63 -Feb 92)

INDUSTRIAL PRODUCTION

EMPLOYMENT

3MOS
R2 RANK

6MOS
R2 RANK

12MOS
R2 RANK

1MON
R2 RANK

3MOS
R2 RANK

QUARTERLY (Jan 63 -Dec 91)
GDP

6MOS
R2 RANK

12MOS
R2 RANK

1QTR
R2 RANK

2QTRS
R2 RANK

4 QTR^
R2 RANK

XU

0369

2

0 555

2

0638

1

0 510

1

0 484

2

0675

2

0694

1

0608

1

0 455

1

0568

1

0401

1

XRI2

0 337

4

0419

3

0364

3

0 255

6

0 484

3

0646

3

0584

3

0417

3

0 382

3

0 316

3

0168

6

LEAD

0 391

1

0.569

1

0606

2

0460

2

0.527

1

0 705

1

0656

2

0500

2

0 405

2

0341

2

0 247

2

PMI

0 355

3

0389

4

0356

4

0 255

5

0463

4

0 580

4

0 483

5

0 325

6

0.265

4

0203

7

0173

5

0434

6

0 572

5

0 530

4

0.381

4

0.205

7

0.216

5

0152

7

0 436

5

0 547

6

0479

6

0 345

5

0.232

6

0206

6

0229

3

.

0 243

5

0 249

4

0193

4

S&P

0.289

6

0.364

5

0349

5

0 269

4

SMPS

0.300

5

0.346

6

0327

6

0334

3

.

.

KSWMIX •

NONE

I

en
en
i




.

0204

.

7

0204

7

0116

-

7

0088

.

7

-

0381

7

0493

-

0417

0282

0.117

0117

0072

TABLE 4.5 - KALMAN MULTIPERIOD FORECASTS (Out-of-Sample)

MONTHLY (Jul 73 - Feb 92)

MONTHLY (Jul 73 - Feb 92)

INDUSTRIAL PRODUCTION

EMPLOYMENT
6MOS
RMSE RANK

1 2 MO S
RMSE RANK

4QTtis

3MOS
RMSE RANK

XU

9441

3

6 293

2

4586

1

4226

1

2 405

3

1662

2

1481

1

1473

1

3 246

1

2 376

1

2 392

1

XRI2

9589

4

7 353

4

6604

6

5444

6

2 439

4

1801

3

1839

3

1897

4

3 427

3

3 026

3

2 758

5

LEAD

9057

1

6081

1

4899

2

4 449

2

2290

1

1614

1

1632

2

1717

2

3 307

2

3 024

2

2669

3

PMI

9172

2

7.163

3

6156

3

5101

4

2 402

2

1864

4

1935

6

1978

7

3838

4

3 319

6

2 736

4

S&P

9921

7

7.442

6

6 287

4

5124

5

2 522

7

1921

5

1884

4

1882

3

3964

6

3 253

4

2758

6

SMPS

9685

5

7 391

5

6 291

5

4 779

3

2 495

6

1944

6

1926

5

1915

5

3914

5

3 306

5

2612

2

.

.

.

4 078

8

3 377

8

2846

8

6

4 052




9894

6

7945

1MON
RMSE RANK

3MOS
RMSE RANK

1MON
RMSE RANK

NONE

12MOS
RMSE RANK

GDP

Indicator

KSWMIX

6MOS
RMSE RANK

QUARTERLY (Jul 73 - Dec 91)

.

7

7125

-

7

5 575

7

2 467

1953

.

-

1943

1QTR
RMSE RANK

2QTRS
RMSE RANK

3369

RMSE 9ANK

2 799

h

o

8088R88

E -S
•R >

~

CO *•»

§ r<*8 8co8h»

d d d d d o d

o o o o o o o

d o o o d d d

2*
l
I

3|R

l
t

! ! ! o8

I

I

CO

28

£j

CT
CM

I

I

I
|

i 8 &

i

£
g

CM <

i $2 5 i

I coqwn J ^
o o o o

o

I I « I
I I C |

* -

i :z

i

i

in . o
i 8 « 2
i

i

CM « j

1 1 1 ^
II
I d

i i
i i

CM CO

CO

! SS i

i

;:

co

8SR8
I

I

d d d d

.
I
I

. cd *- <*> cQ <£
I J; CO CO N O
I c O O O O

8 , H S S
I © ' c d © o" d

•o

2
CO

CO * - tf>
c

o o o '

3 ! 8 <M 8 !

l

fc

I

d o d o

o
o
73

813

X

1

| o* o o o

o "- o> N "«r r*-

_, o> p co co N. <o

S5S8838

£ 8 3 8 5 ? ; !£

o o o o o o

CO

d o d d d d

s
>




X

XU
XRI2
LEAD
PMI
S&P
SMPS
KSWMIX

TABLE 4.6-MULT1P
Probability Value for

CD

!-*£!

£2 < — Q. 0. ;
D oc UJ S •» Z <
x x 3 CL co co :

-57-

-o.2f

Z • * Z CO
a . co co *

4.1. Dynamic Response of Employment to Composite Indicators
National Purchasing Managers' Index (PMI)
annualized percent growth rates
1.05 r

NBER Experimental Leading Index (XLI)
annualized percent growth rates
1.00 r

0.70 h

0.35

000

•0.35
SAP 500 Stock Index (S&P)
1.00 r

NBER Nonfinancia] Recession index (XRI2)

0.25

-1.25

II II I M l II II I III II III III I

III M M I

DOA Composite Index of Leading Indicators (LEAD)
1.20

-0.50
Change in sensitive materials prices (SMPS)
0.75 r

0.80

040

0.00

-0 40




-58-




4 . 2 . Composite Indicators: Cumulated Kaiman Residuals in Forecasting R e a l G D P
NBER Experimental Leading Index (XLI)

S A P 5 0 0 Stock Index (S&P)
cumulated Kaiman residuals
25 r

cumulated Kaiman residuals
25 r

\ / &

•25
•50
-75

t..<

•100

•100

i

•

i

i

i

i •• i

i

i

i

i

i

i

NBER Nonfinancial Recession Index (XRI2)

Change in sensitive materials prices (SMPS)

50

r

50

25

h

i

t

r

-25

* ' '

•4—L.

I

I I

I I I. I

DOC Composite Index of Leading Indicators (LEAD)

Bank lending/(bank lending + CP) ratio (KSWMIX)

50

75

r

50 h
25

-25

National Purchasing Managers' Index (PMI)
50 r

N^

.50 | I 1
1973

-59-

I I
76

|

j t
79

|

| |
'82

I

I

I
85

I

I I
'88

1 1 I
'91




Evans, Strongin, and Eugeni

MIXING MODELS FOR REAL GDP
This section analyzes those indicators drawn from the previous
sections that contain independent information and did well in the
out-of-sample Kalman rankings.

The indicators are subjected to

another round of encompassing tests and rankings.

Finally the

usefulness of these final indicators are assessed in the context
of a time-varying forecast-mixing model.
Table 5.1 presents the Kalman forecast RMSE for the one-,
two-, and four-quarter horizon forecasts of real GDP.

For the

one-quarter horizon the best indicators are the NBER composite
indicators (XLI and XRI2), and the Department of Commerce Leading
Indicators Index (LEAD).

The spreads and real M2 do the worst at

this short horizon, but all of the remaining indicators do
contribute information beyond the own past history of GDP (NONE).
At the two-quarter horizon, the best indicator is the NBER leading
indicator index with the 12-month/Federal Funds rate spread coming
in a distant second:

the NBER leading indicator index is 14% more

accurate than the 12-month/Federal Funds rate spread.

This is not

surprising since the NBER leading indicator index was constructed
by Stock and Watson to produce the "best" forecast of the growth
in economic activity over the six-month horizon considered here.
Turning to the four-quarter horizon, it seems surprising that the
NBER leading indicator index comes in last after the 12month/Federal Funds rate spread, the Federal Funds rate, the 10year/Federal Funds spread, and real M2.

This demonstrates again

that the choice of economic indicators depends critically upon the
horizon being forecast--

at the four-quarter growth horizon, a

different collection of interest rate spreads than the ones
selected by Stock and Watson are useful.
New encompassing results are displayed in Table 5.2.

At

this point, the purpose of these tests is to narrow the list of
indicators in a structured manner.

However, a rigid adherence to

a statistical significance level is not maintained if an indicator
is relatively useful and of independent interest.

At the one-

quarter horizon, the composite indicators the NBER leading
-60-




Evans, Strongm, and Eugeni
indicator index, the NBER nonfinancial recession index, and the
Department of Commerce leading indicators are each undominated and
together sufficient.

The two-quarter horizon is more interesting.

Three indicators are clearly necessary.

The NBER leading

indicator index is undominated, and the 12-month/Federal Funds
rate spread is undominated at the 10% level.

The 3-month

Eurodollar rate is not covered by these two indicators, and it is
not dominated at the 11% significance level.
included in this final cut for two reasons:

Real M2 is also
it is only covered by

the NBER leading indicator index at the 14% significance level and
it is of inherent interest as the best monetary aggregate
considered here.

Finally, notice that the 6-month Commercial

paper spread (CP6TB6) did not make the final list at the twoquarter forecast horizon, but it is a component of the NBER
leading indicator index.
At the four-quarter horizon, three indicators are
undominated:

the Federal Funds rate, real M2, and the 12-

month/Federal Funds rate spread.

The NBER leading indicator index

does not contain independent information beyond these indicators.
The 10-year/Federal Funds rate spread is included in the final
list for three reasons:

it is undominated at the 15% significance

level, it covers the NBER leading indicators index better than the
shorter end of the term structure (12-month/Federal Funds rate
spread), and it is interesting to include a long spread at this
horizon since Stock and Watson found a long spread useful at the
two-quarter horizon.
The next step is to combine these forecasts into a
forecasting model (for each horizon) which allows the weights on
the indicators to vary over time depending upon their recent
performance.

Essentially we would like the model to take the

following form:
Ft = <(>u for(A)t^2c

£or(B)t++u

for(C)

t

where for(A) represents a forecast based upon indicator A and Ft
is the combined forecast.
and sum to one:

The weights <|>lt should be non-negative

in this case, the indicator's weight is a direct

-61-




Evans, Strongin, and Eugeni
measure of its importance for the forecast.

When the

weights

vary over time according to their forecast accuracy, the time path
of the weights provide a direct measure of the indicators'
reliability over time.
way.

Let

eat2

We implement this model in the following

be the sum of (recent) squared forecast errors based

upon indicator i's model.

In this paper, we take "recent" to be

one year of known forecast errors (4 quarters).
the average of the £lt2s at time t and m
<3vgt(£lt2) over time.

Let avgt(elt2) be

is the average of £lt2 -

Then <(>lt is defined to be:

<(>iC = a, - p, (e?t - avg t (e 2 t ) - u,) ,

a, , p, > 0

where the parameters a and p can be estimated by a linear
regression model if the non-negativity constraints are ignored, or
nonlinear methods if the constraints are imposed.8
avgt(elt2) - m

Since £lt2 -

is mean zero by construction, the time-variation due

to the P's nets out to zero over time.

Consequently, the a

estimates represent the average weight associated with each
indicator forecast.

However, over short periods of time when an

indicator's forecast misbehaves, its errors elt2 will be larger
than the average errors;

this will lead to the indicator's

forecast receiving a temporarily smaller weight.
Table 5.3 displays the estimated a weights for these models.
The one-quarter results indicate that the NBER leading indicator
index is the most reliable, having an average weight of .533 in
the combined forecast.

The other indices (NBER Experimental

Recession Index and the BEA Leading Indicators Index) received
about equal shares of the remaining weight.
are estimated to be zero;

The P's in this case

that is, there is no significant

contribution to the forecast accuracy by allowing the weights to
vary over time.
The two-quarter results are more interesting.

As was

8. The results in Table 5.3 were obtained by imposing
nonnegativity constraints.
Initially, each of the P's
constrained to be positive.
If the initial estimate was on
boundary (zero), its corresponding time-varying component
deleted from the estimation.
The a's were constrained to
positive and sum to one.
-62-

the
was
the
was
be




Evans, Strongin, and Eugeni
expected from the encompassing results, the NBER leading indicator
index receives the bulk of the weight in the final forecast (61%).
This agrees with the analysis of Stock and Watson who constructed
the NBER leading indicator index explicitly for its ability to
forecast at this two-quarter ahead horizon.

We. do find that real

M2 receives a substantial weight (19%), while the 12-month/Federal
Funds rate spread is at 10% and the 3-month Eurodollar rate is 9%.
Figure 5.1 graphs the time path of the <|> weights for these four
indicators, as well as the two-quarter GDP forecast and actual.
Notice first that the NBER leading indicator index forecasts have
been quite reliable, only once dropping below a 50% weight in the
combined forecast.

Real M2, however, has varied dramatically in

its usefulness, going negative on two occasions:
immediately following the 1981-82 recession.

in 1976 and

During that

recession, real M2 did not forecast negative growth at any time
(although it did in the 1980 recession), whereas the 3-month
Eurodollar rate, the 12-month/Federal Funds rate spread, and the
NBER leading indicators index did forecast negative growth during
some portion of this recession.9

This poor performance is

captured in the time-varying model by decreasing the weight on the
real M2 forecast temporarily until it begins to improve.

On the

other hand, during the most recent recession real M2 has gone
above a 50% weight (keep in mind that the average weight for real
M2 is .19). During this time, real M2 has grown only slowly and
this lead to a forecast of slow growth during 1991 (see Figure
5.1).

At this same time, the 3-month Eurodollar rate, the 12-

month/Federal Funds rate spread, and the NBER leading indicators
index signalled substantially higher growth than was realized.
Each of these indicators is currently receiving less than its
average weight.

Consequently, the time-varying mixing model finds

that real M2 has been an unusually useful indicator during the

9. It is useful to remember that the primary components of the
NBER leading indicators index are the 6-month Commercial paper
spread and the 10-year/l-year spread.
So it should not be
surprising that the NBER leading indicator index misbehaved during
this period when the 3-month Eurodollar rate and the 12month/Federal Funds rate spreads also misbehaved.
-63-




Evans, Strongin, and Eugeni
recent recession, despite its generally erratic performance at
this horizon versus its relative failure at the twelve month
horizon.
By contrast the four-quarter horizon results appear to be a
picture of stability.

Real M2 and the 12-month/Federal Funds rate

spread receive the largest unconditional weights, 41% and 37%
respectively.

The Federal Funds rate and the 10-year/Federal

Funds rate spread receive considerably less (around 10% each).
The graphs of the time-varying weights indicate that, at this
horizon, real M2 and the 12-month/Federal Funds rate spread have
been reasonably reliable indicators, always staying near their
unconditional weight.

On the other hand, the 10-year/Federal

Funds spread has been extremely unreliable, going to zero or
negative in 1987-88 and during the recent recession.
The contrast between the dominance of the NBER leading
indicator index at the six-month forecast horizon versus its lack
of independent information at the twelve-month horizon
demonstrates strongly the need for a different set of indicators
for each forecast horizon.

The usefulness of the 12-month/Federal

Funds rate spread and real M2 for forecasting real GDP at the
twelve-month horizon indicates that a different index would be
constructed if this forecast horizon was the relevant objective.
A note on standard errors is in order.

Examination of Table 5.3

indicates that the standard errors associated with the parameters
of these mixing models are fairly large.

This is not surprising

in light of the high degree of collinearity that would be expected
of a set of reasonably successful forecasts.

In fact, it is

typically the case that only the strongest indicator at a given
horizon is statistically significant.

All this is saying is that

the relative weights among successful indicators is subject to
substantial uncertainty and that the marginal information after
the first one or two indicators is quickly dropping toward 0.
Nevertheless the point estimates and time paths of these relative
weights provide a useful bench-mark, even though the precision
they are estimated with would not change strongly held prior
beliefs.
-64-




TABLE 5.1 - KALMAN RESIDUALS FOR SURVIVING INDICATORS

Quarterly (Jul 73 • Dec 91)
Real GDP

1Qtr
RMSE Rank

EUR03

3.622

4

2.754

3

n.a.

n.a.

n.a.

n.a.

n.a.

n.a.

2.160

2

M2R

3.674

6

2.844

5

2219

4

CP6TB6

3.656

5

2.760

4

n.a.

n.a.

TB12FF

3.753

7

2.751

2

2.002

1

CM10FF

n.a.

n.a.

n.a.

n.a.

2.161

3

XLI

3246

1

2.376

1

2.392

5

XRI2

3.427

3

n.a.

n.a.

n.a.

n.a.

LEAD

3.307

2

n.a.

n.a.

n.a.

n.a.

NONE

4.052

8

3.369

FF

2Qtrs
RMSE Rank

4Gtrs
RMSE Rank

Indicator

n.a.: The indicator is not an initial survivor at this forecast horizon.

-65-

2.799

-99-

m

r x x o j O S j m
S
B
* ^ ^ a
a

EUR03
FF
M2R
CP6TB6
TB12FF
CM10FF
XU
XRI2
LEAD

3

°

if
l
l

l
1

0.270

0791
0.102

s

l
1

5
°
^33
3*i

8

9«g
o o

1

o .
»i

i ° 2
• i 32

|S

r x x 0 j 0 2 3 f n

g

P P
i

o .

n > i sso1;

i

o

o

s

o o

i

I f»

ro s

o

i

I
I

s

8

3

~
ps

I I

I
I

II

o o
I <D <D I
00 U

I

I

3

o
P Q
O

3




0.105
0959
0960

8.
s-

I

o o

I

:s

88

8

I
I

£

O
O

l

I

I
I

I I
I

|

IS P l
oo

I

I o> I

1 —- I

s
3!
O O

i
i

»

I
I

I
I

I
I

I
I

I
I

I
I

i
i

i

CT> I

I

I

i

i

I
I

I
I

I
I

I
I

O O O O O O O O O
( o o r o - o o D O O O

8S85S = S 3 S 8

I

3
01

I
I

1
I

I
I

37

I
I

o

1 1 1 oS p 1
1 1 1o
^ 1i
, i1 8

11
I I

3

p 1
I

I
I

1
1

1 1 1 1
1 1 1 1

I
I

I
I

I
I

I
I

I
I

o o o o o o o o o
s u o b i b u ^ o } - *

23!

I I I CO I I I
I I I
I I I

» • 8 8 8 8fc3

3

I
I

o
g i l l

I I

_ o o o o o o
I | -; oo .*. ro -^ -* -*
1

23pSS28'

i
i

p

I
I

o o _ o o o o
s b> ^ w o w - I

88^

O
O

I

I I I I Q> I

»
I
I

3

I I

o
o I
5B l

I
I

p o p p3 o
£ 8 en £ P 8
o © s s
o

I
I

1p
t

I
i

|
l

| I 1
l II

1 1
11

o
i i i 8 i i
I

II

en I I

o o o o o o o o o
o o o bo •** ro '-+ <o —•

•of
< 21

M
• 3




TABLE 5.3 - RELATIVE WEIGHTS IN MIXING REGRESSIONS

Real GDP

Indicator

EUR03
FF

1Qtr
*

•

CP6TB6

*

TB12FF

*

CM10FF

n.a.

XRI2
LEAD

0.093
(0260)
n.a.

n.a.

M2R

XLI

2Qtrs

0.187
(0227)
*
.0.103
(0238)
n.a.
0.617
(0.197)
n.a.

0533
(0.174)
0214
(0.155)
0253
(0.206)

n.a.

4Qtrs

n.a.
0.105
(0209)
0.414
(0.178)
n.a.
0.368
(0259)
0.114
(0212)
•
n.a.
n.a.

NOTES:
- Numbers in parenthesis are standard errors.
- n.a.: The indicator is not an initial survivor at this forecast horizon.
- f ) : The indicator is encompassed by other indicators at this horizon.

-67-




5.1. Mixing Results
2 Quarter Ahead Forecast vs. Actual
Real M2 (M2R)
annualized growth rates
9.0

3 month eurodollar (EUR03)
annualized growth rates
90 r

-6.0
12 month T bill less fed funds (TB12FF)
90

NBER Experimental Leading Index (XLI)
90 r

Forecast Reliability Weight
Real M2 (M2R)
weight
0.9 r

3 month eurodollar (EUR03)
weight

06

00

f I I I I I f I I t I I I I I I I f »

-0 3

12 month T bill less fed funds (TB12FF)
09 r

NBER Experimental Leading Index (XLI)
09 r

0.3 h

0.3

0.0

0.0
»

1973

i

i

i

76

i

i

f

79

i

r

i

'82

i

i

i

'85

i

i

i

i

*

i

'91

-0 3

-68-

1", I I I I I f I ,1 I I I I I I I I f I
1973

76

79

'82

'85

'91

-69-

88,
S8,
28.
6L
91.
Z161
16.
T T
roI 1 I I I I I 1 I I 1 I I 1 I I




16.
88.
98.
29,
I I 1 I I I 1 I 1 I \

6L
9L
tl6i
toI J I I » I 1 1

co-

20-

co

00

J 90
( d d 2 l 8 l ) spuni p«| ssa| ||iq j . MIUOUJ zv

-J 90
I f 1 I 1 1 I I I 1 t 1 ] I I

J 90
mala**
(dd) spuni pej

-i 90
(ddOlWO) spun* P«J ssoj puoq i JeeA o i
WfrdAA Ai!|iqBi|eu lSBoaioj
16.

mr

88.
98.
28.
6L
9L CZ61
09I I I I I 1 I I 1 I I I I » I

oeoo
oe

J
06
(ddZlRL) spunj P9* » » l MM 1 MJUOUJ g i

06
(H2VM) 2IN l ^ d

idomo
J 0*6
sew MIMOJO p»zi|tnuue
|en)3v SA jspoejo-j peeqy Jejjeno fr

Joe
wt*j IOMOJO p«zi|*nuue
(dd) spunj p*d

siinsey 6u;xi^ 'zs




Evans, Strongin, and Eugeni

CONCLUSION
Four things become clear as the preceding analysis developed.
First, the forecast horizon is an essential aspect of choosing and
evaluating indicators.

Second, substantial information resides in

the term and public-private spreads and that both of these
seemingly very different types of spreads seem to include
significant common as well as distinct information sets.

Third,

while composite indicators may be extremely useful they are only
as good as their design allows.

The Stock-Watson NBER leading

indicator series does very well at precisely what it was designed
for, forecasting economic activity at a six-month horizon.

Its

usefulness beyond this is far more limited than prior analysis
would have suggested.

The analysis is also suggestive that the

type of general purpose target variable that the old monetary
targeting literature sought, probably does not exist at least in
terms of real economic activity.

Policymakers will continue to

need to mix information according to their current focus.

Mixing

models of the sort used in this paper are meant to be preliminary
work in this regard.

The early results are intriguing.

-70-




Evans, Strongin, and Eugeni

References

Bernanke, Ben S., "On the predictive power of interest rates and
interest and interest rate spreads," New England
Economic
Review, November-December, 1990, pp. 51-68.
Chong, Y. and D. Hendry, "Econometric evaluation of linear
macroeconomic models," Review of Economic Studies,
53,
1986, pp. 671-690.
Estrella, A. and G. Hardouvelis, "The term structure as a
predictor of real economic activity," Journal of
Finance,
46, 1991, pp. 555-576.
Friedman, B. and K. Kuttner, "Why does the paper-bill spread
predict real economic activity?" forthcoming in James H.
Stock and Mark W. Watson eds., New Research in Business
Cycles, Indicators and Forecasting, University of Chicago
Press and the NBER, 1992.
Kashyap, A., J. Stein, and D. Wilcox, "Monetary policy and credit
conditions: evidence from the composition of external
finance," Federal Reserve Board, Working Paper No. 154,
1991.
Laurent, Robert D., "An interest rate-based indicator of monetary
policy," Economic Perspectives,
Federal Reserve Bank of
Chicago, January/February, 1988, pp. 3-14.
National Bureau of Economic Research, Press Release, January 30,
1991.
Sims, Christopher A., "Interpreting the macroeconomic time series
facts: the effects of monetary policy," manuscript, 1991.
Stock, J. and M. Watson, "Interpreting the evidence on moneyincome causality," Journal of Econometrics,
Vol. 40, 1989a,
pp. 161-182.
Stock, J. and M. Watson, "New indexes of coincident and leading
economic indicators," in NBER Macroeconomics
Annual, edited
by O. Blanchard and S. Fischer, the MIT Press, 1989b, pp.
351-409.
Strongin, Steven, "Macroeconomic models and the term structure of
interest rates," Federal Reserve Bank of Chicago, Working
Paper No. 90-14, 1990.
Strongin, Steven, "The identification of monetary policy
disturbances: explaining the liquidity puzzle," Federal
Reserve Bank of Chicago, Working Paper No. 91-24, 1991.

-71-

DISCUSSION OF
A POLICYMAKER'S GUIDE TO INDICATORS
OF ECONOMIC ACTIVITY

Richard W. Kopcke1

This paper examines various indicators, seeking those that are
correlated most highly with the future course of economic activity.
First, the indicators are arranged into "natural groups.H

Second, the

paper selects the most promising indicators from each group.

Third, the

forecasts of the selected indicators are combined to produce mixed
forecasts.
The paper includes many, but not all, of the popular
indicators.

Given that it confines itself to indicators of real

economic activity, perhaps the paper should drop nominal Ml and nominal
M2 (which apparently perform poorly) from its list to make room for
model forecasts, real interest rates, stock prices, consumer confidence,
and other indicators mentioned so much in the press.
Although the paper's strategy avoids using explicit economic
models, in my opinion, it does not escape the consequences of
measurement without theory.

On the most elementary level, the paper's

horse races should include the forecasts produced by economic models.
The mean squared errors of the indicators appears to be high compared to
those of the forecasting services surveyed by Stephen McNees.

On a

deeper level, the paper's strategy seems to require or presume implicit
models which remain, undiscussed, behind the findings.
What determines the natural groups of indicators?

Interest

rates are gathered into one such group, presumably because they are all
called interest rates.

But they do not seem to be a natural group.

The

federal funds rate, for example, principally reflects monetary policy.
The Baa rate reflects general economic conditions.

Perhaps the federal

Richard W. Kopcke is Vice President and Economist of the Federal Reserve
Bank of Boston.




Richard W. Kopcke

funds rate, reserves, etc. constitute a more natural group, while the
Baa rate, real M2, etc. constitute another.
Although the battery of tests performed on the indicators in
each group are reasonably thorough, they are not entirely convincing
without the benefit of the analysis that accompanies models.

More

importantly, it is not clear how the results of these tests are useful
to policymakers.
Historical correlations reflect some mix of fiscal and monetary
policies here and abroad as well as some mix of changing aggregate
supply and demand.

As these mixes vary in the future, these

correlations will likely change.

Specifically, if the average

historical mix should not prevail in this recovery, the indicators may
yield poor forecasts of the course of economic activity in the next few
years.
Historical correlations between indicators and economic
activity may not be a good guide in the future if we know, for example,
that:

(i) fiscal policy will be unusually restrictive for a recovery in

coming years, (ii) the growth rate of the labor force will be less than
one-half that prevailing since World War II, (iii) changing demographics
will reduce the potential magnitude of a housing boom, (iv) the GDP gap
differs from that at the inception of the average recovery, or (v) our
economy is now more open to foreign trade than it had been in previous
recoveries.

Indeed, in the four-quarter forecasts (Chart 5.2, upper

graphs), the indicators, too often, are negatively correlated with the
course of economic activity during the current business cycle.
According to these indicators, the average business cycle is a poor




guide to this cycle.
In the 1940s Haavelmo and Duesenberry explained that the
correlations among state variables (which include both indicators and
economic activity) could not be interpreted outside a model.

Because

these correlations are unstable when economic conditions change, the
remedy requires the modelling of economic behavior, which entails
descriptions of how these correlations are likely to change.

Whatever

the weakness of these models,"however competently they describe the way
businesses, consumers, and governments make decisions, these models
provide a structure needed for private or public policy analysis.
Correlation coefficients are functions of partial correlation
-2-

Richard W. Kopcke

coefficients that might be more stable; nonlinearities are allowed.




If

the Federal Reserve should change its operating procedures (perhaps
following some of these indicators), we cannot anticipate how the
correlations among the federal funds rate, real M2, and economic
activity will change without a model.
To illustrate further the difficulties that interpreting the
correlations between indicators and economic activity pose for
policymakers, consider the federal funds rate (Chart 5.2, upper leftmost
graph) . The correlation between the federal funds rate and activity may
be relatively low for three reasons:

(i) monetary policy has worked

well as a shock absorber, offsetting potential disruptions, smoothing
the ride;

(ii) monetary policy has not reacted to short-run economic

conditions; or (iii) monetary policy has been "out of phase" with the
business cycle.

The correlations of the indicators with activity, by

themselves, do not tell us whether operating procedures should change,
or how they should change.
Setting aside the problems of structural changes, without a
model the correlations among state variables remain dubious guides.

The

paper's bivariate horse races, for example, do not necessarily select
the best indicators.

Bivariate correlations do not predict the order in

which variables are added to or removed from step-wise regressions, and
the results of Granger tests depend on the variables included in the
regression.

Therefore, an indicator which is deemed the best single

candidate in its group may be inferior to another member of its group
when more than one indicator (drawing from any group) is to be
considered at a time.

These problems might diminish if a model were

used to form natural groups from the start, but if we ultimately are to
consider multivariate forecasts, we ought to begin with multivariate
techniques.
In forming multivariate forecasts, the information in each
indicator should not be represented simply by its forecast from its
bivariate regression with economic activity.

These first-stage

regressions restrict the information provided by each indicator, so the
multivariate regression cannot make full use of the correlations among
indicators to describe economic activity.

Constraining the weights of

the forecasts to be positive or to sum to one in the multivariate

-3-

Richard W. Kopcke

regression also prevents the full consideration of all the information
in the indicators.

No explicit model dictates these restrictions.

Indicators may be valuable to bond traders and others who want
instant forecasts, who want inexpensive forecasts, who have little
interest in describing the workings of the economy behind the forecasts,
or who do not require the most accurate forecast, either because they
only need a rough projection or because they make new forecasts very
frequently.

For the purposes of making policy, however, indicators are

not so attractive.

Suppose real M2 and the slope of the yield curve

foretell unacceptable growth of GDP.
give policymakers?

What guidance do these indicators

Should policy change?

If so, how much?

policy influence M2, the slope of the yield curve, and GDP?

How does
I am

reminded of the comment that we must control GDP in order to control M2.
A dilemma also would confront policymakers when, as is often the case,
the indicator that forecasts one horizon best seems too far out of line
with the indicator for a slightly shorter horizon.

Because the paper

concludes that there is no indicator for all seasons, policymakers need
a model or metaindicator to interpret the signals.
For want of a model, indicators also seem to be poor guides for
policymakers, because they provide no framework for setting either the
objectives or the instruments of policy.

For example, indicators do not

show what paths for GDP are feasible or which paths are consistent with
goals for inflation.

Indicators, without a model, do not suggest how

policymakers should react to economic conditions either to achieve a
dynamically stable course for policy or to avoid increasing the
volatility of GDP and prices.
In order to integrate consistently forecasts with policy, we
build models; yet, indicators retain some allure.
in indicators remains for one of two reasons.

Perhaps the interest

First, although our

models are not producing forecasts that are clearly inferior, we may not
take proper advantage of these models for analyzing the consequences of
policy.

Second, the goals of policy may not be specified sufficiently

clearly (perhaps for want of agreement) in order to use the models as a
guide.

In this second case, indicators appear to be useful surrogates;

they fail to stir passions while bridging the potentially disparate
beliefs of policymakers.




-4-




Richard W. Kopcke

REFERENCES

Haavelmo, T. "The Probability Approach in Economics," Econometrica,

vol.

12, Special Supplement, July 1944, esp. pp. 12-39.

Duesenberry, J. S. "Income-Consumption Relations and Their

Implications," Income, Employment, and Public

Policy,

Essays in

Honor of Alvin Hansen, (W. W. Norton, 1948), pp. 54-81, reprinted
in M. G. Mueller, ed., Readings

in Macroeconomics

and Winston, 1966 and 1971), pp. 61-76.

-5-

(Holt, Rinehart,




DISCOUNT WINDOW BORROWING AND LIQUIDITY

W. J. Coleman. C. Gilles, and P. Labadie1

Three features seem centra] to understanding the relationship between U.S.
monetary policy and the comovements of open market operations, monetary
aggregates, and interest rates. First, shocks to bank reserves affect interest
rates in ways that axe not tightly linked to the Fisherian fundamentals (expected inflation, marginal rate of substitution, and marginal productivity of
capital). Second, banks often respond to reserve shocks by adjusting their
borrowing at the Federal Reserve's discount window. Third, the Federal Reserve often conducts open market operations to smooth interest rates that
would otherwise react to private-sector demand shocks. In this paper, we
study a stochastic general equilibrium model that incorporates these features
in an effort to understand important empirical regularities involving monetary
aggregates and interest rates.
The empirical regularities we have in mind are those documented in the
vast literature aimed at uncovering a negative correlation between short-term
interest rates and exogenous policy shocks to nominal monetary aggregates, a
relationship often referred to as the liquidity effect. Cagan (1972) and Cagan
and Gandolfi (1969), among many others, have reported finding negative correlations between Ml itself and various short-term interest rates. Subsequent
studies have reported similar correlations with innovations in Ml backed out
using a Choleski decomposition of the residuals in a vector autoregression (for
a variety of orderings). More recently, however, Leeper and Gordon (forthcoming) have made a strong case that these innovations probably do not represent
exogenous monetary policy shocks, as the money supply may be endogenously
1

Board of Governors, Federal Reserve System. We gratefully acknowledge helpful discussions with Jim Clouse and Josh Feinman.




Coleman, Gilles, and Labadie
determined in ways that are not captured by any Choleski decomposition. To
support their claim, they noted that the statistical properties of these innovations are sensitive to the other endogenous variables included in the VAR,
the sample period, and the measure of money selected for analysis. Some researchers, for example Bernanke and Blinder (1990) and Sims (forthcoming),
have responded to such criticism by assuming that innovations to interest rates
reflect policy shocks, to which the supply of money responds endogenously. For
our purpose, however, this strategy does not resolve the central question: if
there exists a liquidity effect, then why are these interest rate innovations not
robustly negatively correlated with monetary aggregates (an observation also
made by Leeper and Gordon)?
Christiano and Eichenbaum (1991) and Strongin (1991) have tried to obtain robust negative correlations by using nonborrowed reserves as the measure
of money. This approach contrasts with that of Leeper and Gordon, who experimented with monetary aggregates that are at least as broad as the monetary
base. Christiano and Eichenbaum's rationale for using nonborrowed reserves
is based on the widely held perception that the Fed controls this aggregate.
For this reason they associated policy shocks with innovations to nonborrowed
reserves, which they then showed to be negatively correlated with the federal
funds rate. In fact, using nonborrowed reserves as the measure of money, they
found evidence of a negative correlation regardless of whether money innovations or interest rates innovations were identified as the policy shocks, and
they showed that these correlations are remarkably robust to the sample time
period. To explain why the innovations to broader monetary aggregates do not
exhibit a similar correlation, they noted that these aggregates are largely endogenously determined by the banking system. For example, they argued that
total reserves may be inelastic in the short run, and therefore not correlated
with interest rates at all. In this example, policy shocks to nonborrowed reserves do not affect total reserves immediately. Strongin refined this argument;
he argued that innovations to nonborrowed reserves that are not reflected in
shocks to total reserves should be identified as the policy shocks. He asserted,
in essence, that shocks to required reserves lead to an adjustment in both
- 2 -




Coleman, Gilles. and Labadie
nonborrowed and total reserves, whereas open market operations lead to an
adjustment in only nonborrowed reserves.
We develop a model that is rich enough to address the empirical issues
presented above. To do this, we introduce a banking system, reserve requirements, and a discount window into a model of liquidity based on the works
of Grossman and Weiss (1983), Rotemberg (1984), Lucas (1990) and Fuerst
(1992). In these models, and here, the term liquidity effect refers not merely
to a negative correlation between monetary policy shocks and interest rates
but more generally to any non-Fisherian effect on interest rates. Interest rates
deviate from their Fisherian fundamentals because of shocks to the demand for
bank deposits from businesses to finance new investment projects and perhaps
also because of monetary policy shocks. In our model, the interest rate is also
the cost (both pecuniary and nonpecuniary) of borrowing reserves from the
discount window, so that over time there is a well defined relationship between
borrowed reserves and the interest rate. Monetary policy designed to smooth
interest rates then leads to rather complicated mutual dependencies among
open market operations, both broad and narrow monetary aggregates, and
interest rates; in particular, monetary policy can lead to positive correlations
between broad monetary aggregates and interest rates in spite of the liquidity
effect. When policy shocks are correctly identified, however, the model suggests that broad monetary aggregates are negatively correlated with interest
rates, showing evidence of the liquidity effect. Furthermore, the model always
generates a negative correlation between nonborrowed reserves and short-term
interest rates, regardless of what the policy shocks are and how they are identified. Such a result is due to the way the discount window is operated. In
light of this model, one interpretation of Christiano-Eichenbaum and Strongin's results is that they identified the discount window policy. Since this
policy implies a negative correlation between nonborrowed reserves and interest rates whether or not the model incorporates a liquidity effect, their results
shed little light on the presence of such an effect.

- 3 -




Coleman. Gilles, and Labadie
THE MODEL
DESCRIPTION.

To get an overview of the model, consider the following accounting of the
assets and liabilities of banks. Their liabilities comprise demand deposits of
firms and households as well as savings deposits of households. Their assets
are made up of reserves and a portfolio of government securities and loans
to firms. Banks are required to hold as reserves a fraction of their demand
deposits;'to avoid a deficiency, they can borrow reserves at the discount window. Borrowed reserves incur pecuniary and nonpecuniary costs. To start
building a model around this balance sheet, think of households as dividing
their deposits between demand deposits, which can be used to buy goods, and
savings deposits, which cannot. Assume that this division is made before the
value of the open maxket operation is known, resulting in a liquidity effect as
described by Lucas (1990) and Fuerst (1992). Also assume, as Fuerst (1992)
did, that firms must finance their purchases of investment goods with demand
deposits, so that these deposits represent intermediated capital, as in Freeman
and Huffman (1991).
To view the model in more detail, consider a representative household
that ranks stochastic consumption and leisure streams {ct,lt}

according to

the utility function

Lt=0 \t=0

/

where /3{ is the date-i realization of the random discount factor; /3*+i is unknown at the beginning of period t but is revealed later during that period.
The household begins period t with money balances Mt in an interest-bearing
savings account. It immediately transfers amount Zt to a checking account
which bears no interest but can be used during the period to finance consumption ct; only one transfer during the period is allowed. The household must
choose Zt before it knows the realization of any of the current shocks, or prices
for that matter. Its purchases of goods are subject to the finance constraint
Ptct < Zt.
-4-




Coleman, Gilles. and Labadie
At the end of the period, Mt — Zt remains in the household's savings account
and Zt — PfCt in its checking account.
The household derives income from several sources. It provides labor to
the firm, working a fraction of time equal to 1 — it at wage rate Wt] it earns
interest at rate r\ on the amount Mt — Zt in its savings account; it collects a
transfer Xt from the government; finally, as owner of both the firm and the
bank, it collects Il( and II*, the period's proceeds from the sale of output net
of all costs and bank profit respectively. The household receives its income,
including income from labor performed during the period, at the beginning of
the next period, when it is directly deposited into the savings account. With
unspent checking account balances being transferred back into the savings
account, the law of motion for Mt is
Mt+i = Zt - Ptct + (Mt - Zt)(l + r{) + Wi(l - It) + Xt+

Ii{ + II*.

The firm, the second agent in the economy, combines, capital and labor
inputs to produce a homogeneous product sold to buyers of consumption and
capital goods. The production function is
Vt =

F(kt,nt,0t),

where yt is the output, kt and nt are the inputs of capital and labor, and 9%
is a technological shock. The firm owns the capital stock kt and hires labor at
rate Wt] it makes wage payments at the beginning of the next period using the
receipts from the sale of output. The firm must also acquire investment goods
it; it purchases these goods from other firms in the goods market but cannot use
its sales receipts for this purpose. Instead, it finances investment by borrowing
Bt from a bank, which charges interest at rate r*. The bank provides this
financing by crediting the amount to the firm's checking account, increasing
the balance from its starting level of zero. The firm's finance constraint is
Bt > Ptit.
At the end of the period, the firm has spent Ptit on investment goods and
deposits its current sales receipts, PtVt, leaving Bt -f Pt(yt — U) in its checking
-5-




Coleman, Gilles, and Labadie
account. At the beginning of the next period, the firm repays its bank loan
and transfers wages into the worker's savings account. The amount left in the
firm's account, 11^, is paid to the firm's owner as dividend:
n / = Ptyt - Wtm - Ptit -

rtBt.

The stock of capital depreciates at the constant rate 6. so that its law of motion
obeys
fct+i = (1 -6)kt

+ it.

The firm makes all its decisions (namely, J3t, it, and rtt) with full knowledge
of the current shocks and prices.
The bank, the third agent in the economy, starts period t with liabilities
equal to Mt (the household's savings account) and holds an equal amount of
vault cash as an offsetting asset (we write "vault cash" for definiteness; Mt
could also be thought of as an account at the central bank). The household
immediately transfers Zt from its savings to its checking account, without
affecting the bank's total liabilities or assets. The bank pays interest r\ on
Mt — Zt, the amount left in the savings account, but pays no interest on
checking deposits.

By lending Bt to the firm, an amount that is credited

to the firm's checking account, the bank increases both its liabilities and its
assets from Mt to Mt -r Bt. To buy government bonds and to honor checks
written to finance purchases of consumption and investment goods, the bank
depletes its holding of vault cash, Mt] but it replenishes this cash position by
the amount of the checks that firms receive for selling their output, checks
that they deposit in their account. The amount of vault cash that the bank
holds at the end of the period counts as reserves. Note that for an individual
competitive bank, the loan of Bt to a firm drains reserves (when the firm
spends the proceeds) just as much as if the bank had spent an equal amount
to purchase government securities; therefore, at the same rate of interest, the
bank is indifferent between the two types of lending. For the banking system
as a whole, however, loans to firms involve no net loss of reserves, but merely
a transfer from the borrower's bank to the bank of the producer of investment
goods.
-6-




Coleman. Gilles, and Labadie
Reserves. VJ, pay no interest and are subject to a reserve requirement, a
fixed fraction p of the amount of checking deposits on the books of the bank
at the end of the period:
(1)

Vt > p x [(Zt - Pta) + (Bt - Ptit -t-

Pm)].

If the bank cannot satisfy the reserve requirement with the amount of vault
cash it has at the end of the period (after checks have cleared), it can borrow
the shortfall from the government at the discount window. Therefore, the
following accounting identity must hold
(2)

Mt r D t = qtGt + Pt(it T*-

yt) -f Vu

where G% is the number of one-period pure discount government bonds the
bank acquires, at a unit cost of qt = 1/(1 + rt), and D% is the amount it borrows at the discount window. Government bonds, private loans, and discount
window borrowing carry the same rate of interest rt. The bank's objective is
to maximize its period profit, which is given by
(3)

n j = Tt(Bt + qtGt - Dt) - r\(Mt -

Zt\

The government, the fourth agent in the economy, sells one-period bonds
in the securities market and redeems them at the beginning of the following
period, operates the discount window, and makes transfers to the household's
bank account. During period i, the government announces the open market
operation Gt and the amount of transfers Xt after the household chooses Zt
but before any other decision by any agent has to be made. All money flowing
between the government and the private sector, as well as within the banking industry, takes the form of fiat money. The bank starts period t with an
amount of fiat money (which it calls vault cash) equal to Mt. Nonborrowed
reserves Vt — Dt is the amount left in vault cash after the purchase of government bonds and check clearing but before borrowing at the discount window;
in equilibrium, Vt — Dt = Mt — qtGt as can be seen from eq. (2).
Let Ht denote the outstanding supply of fiat money at the beginning of
period t (Mt is best thought of as the demand for fiat money, so that in
- 7 -




Coleman. Gilles, and Labadie
equilibrium Ht = Mt). The law of motion for Ht, which can also be thought
of as the government budget constraint, is as follows:
i?t+1 = Ht T Tt{qtGt — Dt) — Xt.
Think of government policy as a rule that generates the values of Gt and
Xt and that also sets the rate of interest at the discount window.

Assume

that the government lends reserves at the discount window according to an
upward-sloping function if> : [0, oo) —• [0, oo) that relates the rate of interest
it charges to the fraction of total reserves that it lends. Banks cannot lend
at the discount window, so that when the equilibrium rate of interest is lower
than the minimum rate at which the government is willing to lend, V>(0), there
is no discount window activity:
rt = i)(Dt/Vt)
r

t < V^O)

whenever Dt > 0;
whenever Dt = 0.

The argument of if) ought to be the amount supplied at the window, which in
equilibrium turns out to be equal to Dt, the amount demanded.

Incorporating

this equilibrium relationship directly simplifies the notation, but keep in mind
that banks take as given all interest rates, including the rate they face at the
discount window (which is equal to the rate on government securities).
When the Federal Reserve lends at the discount window, the borrowing
bank pays the discount rate plus a nonpecuniary cost; at the margin, this
sum must equal the cost of borrowing from other banks, which is the federal
funds rate. The marginal nonpecuniary cost is thus captured by the difference
between the federal funds rate and the discount rate, called the spread. Historically, the policy of the Federal Reserve seems to have been to supply funds
at the discount window at an increasing nonpecuniary cost (spread), which is
precisely what the function tp assumes. This type of discount-window policy
has been documented in the empirical literature, and is commonly modeled in
the theoretical literature. 2 Chart 1, which graphs the monthly time series for
2

See for example Polakoff(1960), Goldfeld and Kane (1966), and more recently Goodfriend (1983), Dutkowsky (1984), and Waller (1990). In particular, Fig. 1, p. 346 in Goodfriend depicts an assumed ip function that is strikingly similar to the function that would
best fit the scatter plot of our Chart 2.
-8-




Coleman, Gilles. and Labadie
the federal funds rate and the nonborrowed reserve ratio (the mirror image of
the borrowed reserve ratio), reveals the basis for the findings of the empirical
studies. On closer inspection, a picture of the function ib emerges in a scatter
plot of the borrowed reserve ratio against the spread, shown in Chart 2. Since
this picture suggests that the Federal Reserve is ready to lend its first dollar
at a zero spread, the value of t^(0) corresponds to the discount rate. With this
interpretation of ^(0), the model simply assumes a constant discount rate.
A word about terminology is in order. Vt is total reserves in the banking
system; Dt is borrowed reserves; the difference Vt — Dt is nonborrowed reserves;
and required reserves is p x [Zt + Bt + Pt(yt — it — ct)]. Besides total reserves,
it is possible to identify the analogues of several monetary aggregates. M% (or
Ht) corresponds to the monetary base, MO; the analogue of Ml is the sum of
all reservable accounts, Zt + B%\ the total libilities of the banking sector at
the end of the period, Mt + B^ correspond to M2 (strictly speaking, Ml and
M2 both should include Pt{yt — ct — U) as well, but this is equal to zero in
equilibrium); finally, the difference between M2 and MO, which is Bt, is inside
money.
It is now useful to summarize the timing of information and decisions. During period i, the realizations of four random variables shock the economy—the
technological shock 0t> the preference shock /3t+i, the open market operation
Gt, and the government transfer Xt.

At the beginning of the period, the

household must decide how much to put into its checking account, not knowing the current realization of 0t, /3t+i, Gt, or Xt, and therefore not knowing
what interest rates, prices, output, or consumption will be. After it makes
this decision, all four shocks are revealed and prices are set. On the basis of
these shocks and these prices, the household decides how much to consume
and how much to work; the firm decides how much to borrow, how much to
invest, and how much labor to hire; and the bank decides how much to lend
to the firm and to the government. Then trading takes place and checks clear.
The bank monitors its reserve position and borrows at the discount window
to cover any reserve deficiency (the bank can be thought of as borrowing at
the same time it invests in government bonds or lends to firms, because it
- 9 -




Coleman, Gilles. and Labadie
has the same information when it engages in any of these activities). At the
start of next period, the firm pays its wage bill, repays its bank loan, and
pays out its earnings to its shareholder; the government makes transfers to the
household's savings account and redeems the bonds that the bank holds; the
bank pays interest on its savings account, settles its discount window debt,
and pays out its earnings. These activities determine the new initial balance
in the household's savings account. Then a new cycle starts.
The activities of the four agents that have been described above must, of
course, satisfy the following standard market-clearing conditions.
yt

=

a + it

goods market;

nt

=

1 — it

labor market;

Ht

=

Mt

money market.

The economy is competitive, and agents have rational expectations. An
equilibrium is a set of state-contingent prices and interest rates such that
markets clear when all agents solve their optimization problems, treating prices
as given. In the next subsection, we are more explicit about what this means.
THE MODEL AS A RECURSIVE SYSTEM

The household solves a dynamic program, which is recursive under standard
assumptions about preferences, technology, and the stochastic environment.
ASSUMPTION 1.
tiate,

The period utility function U is twice continuously

strictly increasing in both arguments, and strictly

ASSUMPTION 2.

concave.

The production function F has the form F(k, n, 0) = 9f{k, n),

where f is twice continuously
guments,

differen-

differentiate,

concave, and homogeneous

strictly increasing in both ar-

of degree one.

(Stochastic

constant

returns to scale.)
ASSUMPTION

3.

The preference shocks {/3f} and the technological shocks {6t}

are generated by independent

first-order Markov processes.

The support of

&t is contained in (0,1) and that of &t is contained in (0, oo).
Monetary policy consists of a rule that dictates the value of open market
operations, the size of government transfers, and the level of the discount rate;
-10-




Coleman. Gilles. and Labadie
these instruments are not completely independent of each other. The operation
of the discount window is modeled through a fixed function w that relates the
discount rate to borrowed reserves. Think of the government as announcing
this function and keeping it fixed in all periods, leaving the discount rate itself
endogenousiy determined by the demand for borrowed reserves. Given the
function V>, the values of Gt and Xt in period t are implied by the choices of the
ratios gt = Gt/Ht and 7* = iift+i/^t- To induce stationarity and recursivity,
choose (gujt)
ASSUMPTION

4.

order Markov

as the policy variables and make the following assumption.
The monetary policy shocks { # , 7*} are generated by a firstprocess.

Starting with the optimization problem faced by the bank simplifies both
the notation and the analysis. The bank maximizes its period profit, given
in (3), by choosing an optimal portfolio (Sf,Gt,i?t, Vi), subject to the legal
reserve constraint (1), and the accounting identity (2). Clearly, optimization
requires that V% = p[Zt + Bt + Pt(yt — it — ct)] (no excess reserves) if r* > 0. A
zero-profit condition, the result of perfect competition and constant returns to
scale in the banking industry, implies that r\ = [{Mt + Bt — Vt)/(Mt — Zt)] x rt;
this condition in turn yields r\ = r t [l + (l — p)(Zt + Bt)/(Mt — Zt% which holds
whether or not r* > 0. To obtain the last expression, recall the market-clearing
condition yt = ct-r itSince the firm and the bank belong to the household, it is possible to
integrate the problems faced by the firm, the bank, and the household. Because money growth induces a trend in nominal variables, stationarity of the
equilibrium requires that nominal variables—denoted by uppercase letters—
be divided by the supply of fiat money. The new variables are denoted by
the corresponding lowercase letters; thus, nit = Mt/Ht,

zt = Zt/Ht,

and so

forth. Under assumptions 3 and 4, the evolution of the shocks is determined at
the beginning of period t by the vector (/3t,0t-i> 5t-ij7t-i)> which consists of
the latest known realizations of the shocks. The state of the economy at that
time can then be expressed as st = («t?/3t, 0t-i?5t-i)7i-i)> where Kt is the
aggregate per capita stock of capital (as opposed to fct, which is the individual firm's holding). In equilibrium, of course, individual decisions determine
- 11-




Coleman. Gilles, and Labadie
aggregate outcomes, so that K% = kf. A solution is a set of functions p, w.
and r such that pt = p(st,st+i), wt = tu(st,.st+i), and rt = r(s t ,«st+i) yield
the equilibrium values of the normalized price level, the normalized wage rate,
and the rate of interest on date t (again, pt = Pt/Ht and wt = Wt/Ht).

Since

qt = 1/(1 +Tt)} the equilibrium function r determines a function q satisfying
9t = g ( j t i * t + i ) .

Given such pricing functions, let J(m, &, s) denote the value of the optimal
discounted stream of utility for a household starting a given period with money
balances m, while the firm owns capital stock k and the economy is in state
s = («,/?, 0,(/, 7). The household first chooses z, which is the transfer from
its savings to its checking account, expressed as a fraction of the outstanding
supply of fiat money. Then (/9^0^if^7 , ) are revealed (a prime denotes the
realization of a variable that was unknown at the beginning of the period),
and these shocks determine the current price, wage rate, and rate of interest,
as well as the next-period state s'.

To determine s1, the household must

know how the evolution of the aggregate capital stock depends on the state
of the economy. In equilibrium, of course, this law of motion follows from the
individual optimal decisions. On the basis of an assumed law of motion for
K and of p(s,s ; ), w(s1s')}

and r(s,s'),

the*household makes its consumption

and leisure decisions and the firm makes its labor and investment decisions.
What these optimal decisions are can be studied by considering the Bellman
equation characterizing J, the value function.

J(m,k,s)

= max-Ej

max {C/(c,£)' + / 3 J ( m U V ) }

subject to

(4)

2>P<:;
n* = pO1 f(k1 n) — (1 -J- r)pi — wn;
Jfe' = ( l - * ) J b + i;
w(l - I) + (m - z)(l + rb) -f x1 + 7Tf -»- (z - pc)

,
m

=

-12-




Coleman. Gilles, and Labadie
the last constraint on the problem is the law of motion for K. Here p, w. and r
are short for p($, s'), w(s, s1), and r(s, s'), and E9 is the expectation conditional
on 5. Using the results of the bank's optimization problem, the market-ciearing
condition 6' f{k,n)

= c + i, and the firm's optimization condition b = pi. we

have r > 0, v > p(z + 6), and v = p(z + b) if r > 0.
OPTIMIZATION AND EQUILIBRIUM CONDITIONS.

The Bellman equation for J includes two maximization operators; the first
refers to the choice of z, which is conditional only on s, and the second refers
to the choice of (c, £, n, i) which is conditional on both s and sf. Corresponding
to the latter choice, we have the following four first-order conditions:

(<0
u>(j,5')

u>(a,s') = p ( * , a ' y / 2 ( * , n ) ;

(»)

(0

y

J 2 ( m \ Jb',,') = p(*,,')[! + r(s,

M')]Jl{m''?'J);

where A is the Kuhn-Tucker multiplier associated with the finance constraint
(4), so that A(z — pc) = 0. Indexes to the functions U and J denote partial
derivatives; therefore, U\, for example, is the partial derivative of U with
respect to its first argument, consumption.
The first-order condition associated with the choice of z is
(*)

E.

Ui(c,£)
3

S

IP( > ')\

= E. 0[l + rh(s,*')]

Ji(m',k',s')

r

To solve the dynamic programming problem, we need the following envelope
conditions, which give the marginal values of money and capital:

(m) / l ( m , 4 , ( ) . & [ ^ ] .
(*)

J 8 (m, * , . ) = E. [(U,(c,t)

- pX) ( « 7 i ( * , n ) + (1 + r)(l - * ) ) ] ;

where p is short for p(s, s1), and similarly for w and r.
-13-




Coleman, Gilles, and Labadie
Finally, an equilibrium in this economy is a set of functions

w(s,s!),

p(s,$'), and r(s, s1) [or equivalently 9(3, s1)] and a law of motion for the aggregate capital stock K such that the associated solution of the dynamic programming problem—that is, values for (z, A, c, /, n, 2, i/, d) that solve the first-order
and envelope conditions—satisfies the following equilibrium conditions:

c + i = tC/(i, n);
l - * = n;
qg + v - (f = m;
rn = 1;
fc' = « ' ;
r6 =

m —2

xr;

d x r = (fx ip(d/v).
The last equation states that, when the monetary authorities lend at the discount window (d > 0), they do so in accordance with their supply behavior,
so that r = ^{d/v).

In the third equilibrium condition, qg1 + v — d = m, v is

equal to p(z + pi) unless r = 0, in which case v can exceed required reserves.
SOLVING THE MODEL

Consider initially a slightly simplified version of the model in which labor is
inelastically supplied (I = 0) and money supply'is constant (7 = 1). To solve
this simplified model, first reduce the system of equations that determines the
equilibrium to only three equations in the three unknown functions c, z, and
(a transformation of) J\.
To simplify the notation, define £(/9,.s') = /3Ji(l,/c',y). 3 Then the firstorder condition (c) becomes

Ui(c) = (\ + t)PRecall that K is one of the arguments of 5, so that the function £ is well defined.
-14-




Coleman, Gilles. and Labadie
Here and below £ stands for f(/3,s'); accordingly £' below stands for £(/?', s").
Using this equation and the constraint z > pc, which holds with equality
whenever A > 0, isolate p as
(5)

p = nun

^

{

}

Substitute this equation in £ = (3E5i [U^c^/p1], which follows from the definition of £ and the envelope condition (m), to obtain

( = 0E, max

(6)

{***.<}

this equation is the first of the set of three to be solved (£ now replaces J\).
The second equation follows from substituting the expression (5) for p into the
first-order condition (z), obtaining

(7)

£.|ma*{^,*}]=2<;.l(l + r )t}.
h

The last equation in the system follows from the first-order equation (i) and
the envelope condition (Jk):

mm

{**«}
l£l

(8)

z'e, Ux{c')
\ {9"h{k') + (1 + r')(l - 6))
= 0qEs nun < —
<O}CJ

To write (6) - (8) solely in terms of c, z, and £, express r and r in terms of
these functions as follows:

T = i>(dlv);
and
(9)

r> =

where
-15-




Coleman, Gilles, and Labadie
d = qg ~ v - 1;

v = p(z + 6);
6 =

m

| _ _ _ |

m

;

and finally,

i = 0'/(Jfe)-c.
These equations hold provided d > 0 and r > 0; if d = 0, then r < ^(0), while
if r = 0, then v > p{z + 6). Rather than solving this model explicitly, which
can be done numerically using the methodology presented by Coleman (1992),
we devise an example which admits a closed-form solution.

This example

highlights all the features of the model that are useful in interpreting the
empirical regularities mentioned earlier.
AN EXAMPLE

To develop an intuitive understanding of the model, it is instructive to consider
a parametrization that allows a closed-form solution. Suppose that (a) utility
is logarithmic; (b) production satisfies f(k) = fca, for 0 < a < 1; (c) capital
depreciates completely over each period; and (d) the technological shocks 5,
the policy shocks g, and the preference shocks 0 are all iid (although not
necessarily independent of each other). Now, conjecture that no excess cash
is ever held in the goods market and that z is constant at z.
;

circumstances, 6 = zi/c, i = fc , and equations (6)-(8) simplify to

z
(10)

(11)

1 = E, 'P(l+rb)

" = 0'qEsl

)

'6"a(k')a-1'
J

c
where the interest rate r satisfies

<*>

-*r»''
—1 6 -

Under these




Coleman. Gilles, and Labadie
and rb is given by (9). Further conjecture that the consumption function can
be written as
Q =

TTzrt—r$ k ,

1 + Wq)
for some function h. Note that because k!/c = h(/3'g), the function h can be
thought of as the investment to consumption ratio. Since h depends only on
flq and since q = 1/(1 + r), (12) determines r a s a function of 5, /3', and g1.
Write this function, which implies that r and q are iid and independent of s,
as r = RJ^z^ff^g1) and correspondingly q = Q{z^P\g9)\

now substitute these

equations into (9), and the resulting equation into (10), to obtain

! = £ , tt[i

+

[i^->n*««*<i'-™t)m!,M)

This equation has the important implication that z does not depend on 5,
because s enters only through the conditional expectation, and /?' and g1 are
iid. This observation verifies the conjecture z(a) = z. Tofindfc,substitute the
conjecture about the consumption function into (11) and simplify to obtain
h(l3,q) = a(3'q(l + Esl[h(/3"q')}).
Using the fact that 0' and q are iid (because q — Q(z,0',g'),

and (@,g) is iid),

this equation implies
H{l3q)

-l-Ela0<qy

where E[. ] is the unconditional expectation, taken over the constant distribution of (y9',g). It is then straightforward to verify that the finance constraint
in the goods market is always binding; therefore, all the initial conjectures
were correct.
This example leads to a sharp characterization of the response of monetary
aggregates and the interest rate to supply and demand shocks.
equilibrium value of k'/c =fc,rewrite (12) as

(13)

U^^f^Sl-^d-ElaP'q))
q

^\

pz(l+a0'q-E[a(3'q})
-17-

Using the




Coleman, Gilles, and Labadie
Consider first the effect of technological shocks, &. .Such shocks do not
affect r, as (13) makes clear, and thus they do not affect any of the monetary aggregates. They have real effects, of course, since they affect output,
consumption, and investment. But they fail to move nominal interest rates
(although real rates certainly do) because the demand for consumption and
investment goods shift proportionately. This feature is due to the choice of
utility and production functions, and is not a general feature of the model.
It indicates, however, that in the general case productivity shocks can affect
interest rates and monetary aggregates in either direction. Before turning to
the effect of other shocks, it is helpful to list the relevant equations. The first
is (13), which determines the correlation between each shock and the nominal
rate of interest. The others are:
(14)

total reserves:

v = pz[l + h(/3'q)]]

(15)

nonborrowed reserves:

(16)

borrowed reserves:

(17)

Ml:

z + b=z[l

(18)

M2:

1 + 6 = 1 + 2&(0'g).

t; — d = 1 — qg1]
d = v x ^-1(r);
+ h(0'q)]]

To isolate the effect of policy shocks, assume first that there are no other
shocks (a similar procedure will uncover the effect of preference shocks). Note
that the left side of (13) is decreasing in g, while the right side is increasing both
in q and in g1 (recall that T/J is increasing); therefore g' and q vary inversely. For
the same reason, but considering the right side as a function of q and qg\ q and
qgf vary inversely also. Hence, gf, r, and qg1 all move in the same direction.
In view of (15), then, policy shocks induce a negative correlation between the
nominal rate of interest r and nonborrowed reserves v — d. They also induce
a negative correlation between r and v, total reserves, as (14) reveals since h
increases in q. The correlation between r and v can be entirely attributed to
the variance of inside money, z/i(/3'g); this variance also induces a negative
correlation between r and the broader monetary aggregates Ml and M2, as
shown by (17) and (18). From (16), it is clear that the ratio of borrowed to
total reserves is positively correlated with the interest rate, a relation which
-18-




Coleman, Gilles, and Labadie
has nothing to do with the source of the shock but is due exclusively to the
form of ^, that is, to the operation of the discount window. If total reserves
did not respond to the policy shock (an assumption which is sometimes made
in empirical work), the form of ifr alone would induce a positive correlation
between the interest rate and borrowed reserves.
Suppose now that shocks to /3 are the only shocks in the system. The
left side of (13) is decreasing in g, while the right side is increasing in q and
decreasing in /3'g; therefore, q and /3'q (and therefore q and /3' also) move in
opposite directions, while {31 and /3'q move in the same direction. Equations
(14)—(18) then show that preference shocks induce a positive correlation between the interest rate and any of the reserve or monetary aggregates (total,
nonborrowed, and borrowed reserves; inside money, Ml, and M2).
It is now possible to use the example to study more complicated policies.
Suppose that in response to positive preference shocks that would otherwise
increase interest rates, the government chooses its open market operation to
keep the rate constant, which corresponds to a small realization of g1 (in this
case, /3 and g are still iid, but not independent of each other). With the interest
rate constant, /?' high and gf low, all the reserve and monetary aggregates are
high (but the borrowed reserve ratio is constant). If the policy response only
partially offsets the preference shock, all reserve and monetary aggregates may
still rise, while the rate of interest rises also. In that case, despite the presence
of a liquidity effect in the model, open market operations could be seen as
"inducing" a positive correlation between interest rates and various monetary
aggregates (and nonborrowed reserves as well).
CONCLUSION: INTERPRETING THE EMPIRICAL LITERATURE

As mentioned in the introduction, the empirical literature directed to measuring the effect of monetary policy shocks on interest rates is replete with
seemingly conflicting results. The model provides a framework for thinking
about these results and for interpreting the literature; the example brings out
the important features of the model. First, the model highlights the role of inside money creation as an avenue for total reserves to respond to open market
-19-




Coleman. Gilles, and Labadie
operations. In this sense, the model fails to support Strongin's identifying restrictions that total reserves do not respond to open market operations within
a month or a quarter. Second, the model suggests that the operation of the
discount window, summarized by a fixed and positively sloped supply function,
can alone generate a negative correlation between nonborrowed reserves and
the federal funds rate. Such a correlation has been documented by Christiano
and Eichenbaum (1991). While they identified policy shocks as innovations
to nonborrowed reserves, the model suggests an alternative explanation that
has nothing to do with policy shocks. Third, although the model is designed
to have a liquidity effect, a policy of interest-rate smoothing hinders efforts to
detect its presence. This could explain the difficulties econometricians have
had in measuring this effect. To identify policy shocks, it is not sufficient to
identify a variable (such as nonborrowed reserves) that is under the control of
the Fed, since the Fed may use its instrument to achieve particular objectives.
In this sense, the model points to the familiar need, and provides a framework
for, identifying demand and supply shocks to estimate a liquidity effect.

-20-




Coleman, Gilles, and Labadie
REFERENCES

Bernanke. B., and A. Blinder. "The Federal Funds Rate and the Channels of
Monetary Transmission," Working Paper No. 3487. New York: National
Bureau of Economic Research, October 1990.
Cagan, P. The Channels of Monetary Effects on Interest Rates. New York:
National Bureau of Economic Research, 1972.
Cagan, P., and A. Gandolfi. "The Lag in Monetary Policy as Implied by the
Time Pattern of Monetary effects on Interest Rates," American Economic Review, vol. 59 (Papers and Proceedings, 1969), 277-84.
Christiano, L. J., and M. Eichenbaum. "Identification and the Liquidity effect
of a Monetary Policy Shock." Unpublished manuscript, Federal Reserve
Bank of Minneapolis, November 1991.
Coleman, W. J. "Solving Nonlinear Dynamic Models on Parallel Computers,"
Institute for Empirical Economics working paper. Federal Reserve Bank
of Minneapolis, 1992.
Dutkowsky, D. "The Demand for Borrowed Reserves: A Switching Regression
Model," Journal of Finance, vol. 39 (1984), 407-24.
Freeman, S., and G. W. Huffman. "Inside Money, Output, and Causality,"
International Economic Review, vol. 32 (1991), 645-67.
Fuerst, T. S. "Liquidity, Loanable Funds, and Real Activity," Journal of Monetary Economics, vol. 29 (1992), 3-24.
Goldfeld, S. M., and E. J. Kane. "The Determinants of Member-Bank Borrowing: An Econometric Study," vol. 21 (1966), 499-514.
Goodfriend, M. "Discount Window Borrowing, Monetary Policy, and the PostOctober 6, 1979 Federal Reserve Operating Procedure," Journal of
Monetary Economics, vol. 12 (1983), 343-56.
Grossman, S. J., and L. Weiss. "A Transaction-based Model of the Monetary Transmission Mechanism, " American Economic Review, vol. 73
(1983), 871-80.
King, R., and C. Plosser. "Money, Credit, and Prices in a Real Business Cycle,"
American Economic Review, vol. 74 (1984), 363-80.
-21-




Coleman. Gilles, and Labadie
Leeper, E. M., and D. B. Gordon. "In Search of the Liquidity Effect," Journal
of Monetary Economics (forthcoming).
Lucas, R. E., Jr. "Liquidity and Interest Rates," Journal of Economic Theory,
vol. 50 (1990), 237-64.
Polakoff, M. E. "Reluctance Elasticity, Least Cost, and Member-Bank Borrowing: A Suggested Integration," Journal of Finance, vol. 15 (1960),
1-18.
Rotemberg, J. J. "A Monetary Equilibrium Model with Transaction Costs,"
Journal of Political Economy, vol. 92 (1984), 40-58.
Sims, C. A. "Interpreting the Macroeconomic Time Series Facts: The Effects
of Monetary Policy," European Economic Review (forthcoming).
Strongin, S. "The Identification of Monetary Disturbances: Explaining the
Liquidity Puzzle." Unpublished manuscript, Federal Reserve Bank of
Chicago, December 1991.
Waller, C. J. "Administering the Window: A Game-Theoretic Model of DiscountWindow Borrowing," Journal of Monetary Economics, vol. 25 (1990),
273-87.

-22-

Chart 1. Federal Funds Rate and Nonborrowed Reserves Ratio
Monthly, January 1961 - July 1992

percent

1960




1965

1970

1975

1980

1985

1990

i

Chart 2. The Psi Function; 1961 (1)-1992(7).
Ratio of Borrowed To Total Reserves
U.1U

•

0.08

—

•

0.06

•

•

•

•

•

•

•

•

0.04

*•

• • • :
• • •
• •
•

• § • •

•

•

••

•

•

•

•

• • •••

•

••

•
•

•

•
/

••

•

• •
•• •
•
•-

• •
•

•

1

•
•

•

1

•
•

•

•

0.02

•
%

0.00

hi

•

.

:

*• .wA * * i * '„•.••••
1

•• ••
••

• .
l_

-2




finrftarl fFed Funds - Discount Ratol

L_

l_J




Comments on "Discount Window Borrowing and Liquidity"
by Coleman, Gilles, and Labadie
Michael Dotsey
I have been asked to discuss "Discount Window Borrowing and
Liquidity" which I view as very interesting but preliminary work
toward examining "liquidity effects" in a framework that
incorporates a fairly (primitive) reserves market. I use the term
primitive with regard to the reserves market since no interesting
dynamic behavior is present in this market. Viewing work on BRd,
especially that of Goodfriend (1983) this is a shortcoming that I
hope will be addressed by later generations of the model. The
paper, however, is very rigorous and state of the art on other
dimensions and the authors deserve a lot of credit for moving the
liquidity effects literature in this direction.
The empirical motivation for the paper can be traced to
work by Christiano and Eichenbaum (1992) and especially to that of
Strongin (1991). Strongin's work is fairly persuasive and
indicates that in order for any model to replicate data on
liquidity type effects reserve market behavior is likely to be a
crucial ingredient. This is because the liquidity effect only
shows up in NBR's or to be more accurate, in the part of NBR that
represents independent monetary policy. This paper's novel
inclusion of reserve market behavior represents a commendable
extension of this basic line of research.1
In reading this paper, I found that it raised at least as
many questions as it answered. Much of my confusion is not the

1. One thing I would like to see done in these estimations is
removing settlement day data. This data could potentially contaminate
the results. Suppose for instance the Fed misforecasts float or
treasury balances believing there will be more of these funds available
than are actually there. NBR will be low on the settlement day and the
funds rate will be high, perhaps by a substantial amount. Two such
occurrences in a month (at least 25% probability) could make monthly
average NBR a little low and monthly average rF a little high. While I
doubt this is the reason for Strongin's results it would be nice to
purge the data of what is merely an interbank friction.




Michael Dotsey
result nor the fault of this paper in particular, but rather comes
from a lack of understanding and perhaps misgivings of this
literature in general.

In my comments I will discuss some of

these misgivings and, hopefully, my comments will lead to some
discussion from the rest of the audience.
The paper extends a branch of research that is attempting
to understand the effect of monetary policy on interest rates and
real activity.

In particular these papers7 search for a mechanism

that will explain (1) how contractionary monetary policy raises
short-term interest rates and (2) how it causes declines in
economic activity.

This literature received its impetus from

Lucas's (1990) influential paper.

A common feature of most of

this literature involves cash-in-advance constraints that
constrain the amount of money available for use in a loan or
securities market, however, no two papers seem to use the same
exact specification.
Lucas's original setup and CGL (1991) envision bond
traders as only having limited funds and, therefore, open market
operations affect the price of bonds .and thus interest rates.

The

appeal of Lucas's setup is that it eliminates the differential
wealth effect of open market operations that were present in
earlier literature (eg Grossman and Weiss and Rotemberg).

Fuerst

(1991) extends Lucas's setup to a production economy that places a
CIA constraint on both investment and labor expenditures.

Unlike

households' portfolio decisions, production decisions are made
after the stochastic state of the economy is known.

Since

individuals must choose the portion of their portfolio to lend to
firms via intermediaries prior to observing the monetary transfer
or the market clearing interest rate, the monetary transfer can
affect the tightness or looseness of the loan market.

Hence

liquidity effects that have real consequences result from monetary
policy.

Christiano (1991) subjects the Fuerst model and an

alternative version of that model in which investment decisions

- 2 -




Michael Dotsey
are also made prior to the realization of shocks to a statistical
comparison with a RBC model that contains a standard CIA
constraint.

For reasonable parameter specifications the Fuerst •

model can not produce a liquidity effect that dominates
anticipated inflation effects on the nominal interest rate while
the sluggish capital model can produce a dominant liquidity
effect.

Both these models produce too much variability in

consumption and the counterfactual result that consumption and
prices move in opposite directions.

They also produce very low

interest elasticities of money demand and monetary policy has very
little effect on variations in output.

Furthermore, anticipated

inflation has much too large an effect on labor, consumption, and
output.

To remedy this last result, Christiano and Eichenbaun

(1992) relax the CIA constraint on investment.

They also split

the period into two parts allowing firms to adjust their hiring
decision after observing open market operations while initial
hiring and investment decisions are made prior to observing open
market operations.

They do this with the hope of magnifying the

response of employment and output to liquidity effects.

In CGL's

current paper firms face a CIA constraint on investment but can
pay workers out of end of period revenues.

Also, monetary

transfers are made directly to consumers after their portfolio
decision has been made.

Thus these transfers do not affect the

funds available in the credit market and, therefore, do not give
rise to a "liquidity effect."

Because there is a CIA constraint

on capital, monetary policy can have inflation tax effects as
well.

As their work progresses separating liquidity effects from

inflation tax effects will be important.
Not all of these scenarios can be correct.
constraints placed where they are?

These assumptions of infinite

transactions costs are not innocuous.
in these models.

Why are CIA

They are the driving force

It seems that rather than trying to incorporate

a realistic financial structure into a dynamic macro model and

- 3 -




Michael Dotsey
then testing the model, investigators are trying to find a
mathematical structure that produces the correlations they desire.
Apart from Christiano (1991) very little effort is made to see if
these models are an improvement on basic RBC models or even if
they produce counterfactural predictions along other dimensions.
Since other classes of models can produce negative correlations
between NBR and the funds rate, examining how CIA models fit the
data along other dimensions will be important if the CIA approach
is to gain widespread acceptance.
For example a model like that in Goodfriend's (1987) paper
can potentially produce correlations of the type this literature
is seeking. In that model, which has no rigidities, purposeful
behavior by the Fed can set up negative correlations between the
funds rate and NBR. If the Fed wishes to reduce inflation, it can
do so by reducing the future money supply and in particular future
NBR. Due to anticipated inflation effects, the nominal interest
rate would fall increasing the demand for money and total
reserves. If the Fed wishes to reduce price level surprises it
can supply the necessary NBR to prevent price level movements.
Thus this policy sets up the requisite negative correlation. If
that was all that was going on one would expect this negative
correlation to carry over to broader aggregates. However, M2-M1
components of M2 which involve a large savings motive should be
positively correlated with the real rate of interest and movements
in BR, which are highly variable and positively correlated with
the funds rate, could cause TR to be positively correlated on net
as well.
Alternatively say the Fed is following an exogenous upward
movement in the real rate of interest in an attempt to target
inflation. If the own rate on money balances is sticky then money
(Ml) and hence total reserves will decline along their demand
curve. (Also, M2 could be rising with the real rates.) This
would set up a negative relationship between NBR and the funds

- 4 -




Michael Dotsey
rate.

As rm adjusted, total reserve demand would increase as

would NBR as the Fed defended the new higher funds rate.

If the

Fed did not react instantaneously or vigorously enough to the
increased reserve

demand the funds rate could rise further and

then fall as nonborrowed reserves were pumped into the system
reinforcing the initial negative correlation.

Also, sticky price

models may be able to generate some of the correlations displayed
in the data as well.
Also, the question of what constitutes a period is
somewhat fuzzy in this literature.

Is it a day or perhaps a week?

Most people make some form of cash management decision weekly and
I can not think of any time where a shortage of cash has affected
my real consumption for more than a day or two.

Perhaps I'm

taking the CIA constraint too literally, but if the period is
rather short, as I believe it is, then the propagation mechanisms
needed to match the data would seem incredible by RBC model
standards.
I have strayed a little far afield so let me return to
this paper more specifically.

My primary confusion is linking the

author's major contribution which shows how different measures of
money can have different correlations with interest rates with the
motivation for their paper which appears to be the results found
in Strongin.

In this paper money

(1)

M t ^ - Mt + rt(Gt-Dt) + xt.

The xt portion of measured money provides no liquidity effects.
The 6t portion, that is open market operations has the standard
liquidity effects since it influences the portion of firm
borrowing that must be financed by discount window loans.

The

equilibrium condition that is being used is
(2)

NBRt = Vt - Dt =

Mt - Gt

where Vt * 0(M t +B t ).

An increase in 6t (an open market sale)

requires more discount window borrowing and an increase in
interest rates since r =0(D/TR) is increasing.

- 5 -

Using Mt+1 can




Michael Dotsey
contaminate regression results since it rises by r t (G t -D t ), which
will in general be positive in this model and no liquidity effect
will be present.

Furthermore, growth in money via transfers wiVI

further bias econometric results.
For econometric purposes I see no useful way of isolating
any aggregate to uncover liquidity effects-

Xt type disturbances,

in reality, involve transfers from the Treasury.

These involve a

reduction in Treasury accounts at the Fed and an increase in NBR.
What the model here indicates is that one wants to examine only
changes in reserves

that involve changes in the public's asset

positions and that exclude any interest or lump sum payments.
While these decompositional problems are important for
this model and may in fact be important more generally, they seem
to have little to do with Strongin's empirical strategy nor do
they affect interpretations in other models.

Strongin tries to

separate "pure" supply movements in NBR from those engendered by
policy responses to changes in TR. Whether his identification
procedure is a good one or not could be debated, but he is not
concerned with measurement or decompositional problems in various
reserve measures.
The decompositional problem arises in CGL because of their
modeling of xt as having no liquidity effects.

In Fuerst or

Christiano and Eichenbaun, there is only xt and it enters the
model in a way that produces liquidity effects.

That is NBR

supply disturbances that are not responses to TR shocks produce
liquidity effects.

It seems that Strongin's methodology is more

closely aligned with these models.
Whether decompositional problems are important or not, I
don't know.

They arise in this model by a specification that at

this point seems somewhat arbitrary.

It is no more arbitrary than

any other specification in the literature, but that does not make
it convincing.

I believe the author's need to make a convincing

argument as to why some forms of morjey creation are more likely to

- 6-




Michael Dotsey
involve liquidity effects than others if their message is to carry
weight. After all, in this model one could easily reverse the
roles of Xt and Gt or make them complimentary.
The discussion on page 11 regarding the estimation of rp is
also a little confusing. With

(3)

n-1 -± -±£
v

V

they claim that 0 can be estimated no matter what the shock. But
is that relevant? We would like to know how j> is influenced
contingent on different shocks. Here a positive V shock induced
by a shift in the demand for loans causes 0 to rise and n to
fall, while a decline in NBR due to an open market sale (G up)
also causes n to fall and i> to rise. It is only the latter effect
that one has in mind when discussing liquidity effects, so perhaps
the ratio is not the correct variable to focus on. Rather, in
this model it should be the relationship between the level of NBR
and the funds rate. Also in estimating 0, one would expect shifts
in the function over time since administration of the discount
window has changed over time. For example, I believe window
administration was more lax when the Fed faced a membership
problem.
I would also downplay somewhat figure one. The interest
rate of consequence is the spread between the funds rate and the
discount rate. When one looks at this graph the correlations seem
at least as pronounced. But has anything but a borrowed reserve
demand function been uncovered?
Finally, the discussion concerning adjustably pegging the
interest rate based solely on technological disturbances raises
questions concerning the nominal determinacy of the model (see
McCallum (1981, 1986)).
Overall, I thought this paper was interesting and
represents a nice attempt to start thinking about how behavior in

- 7 -




Michael Dotsey
the market for reserves influences the correlations we observe
between various monetary measures and the funds rate. Given my
qualms concerning this methodology's ability to explain anything
at business cycle frequencies, I would suggest directing the model
in an alternative direction. Perhaps this framework could be used
to help explain short-term term structure movements in interest
rates and examine the so-called "ozone hole." This line of
inquiry would be interesting since it could integrate reserve
market behavior and a tight specification of policy in a fully
developed general equilibrium model.

- 8 -

Michael Dotsey
REFERENCES
Christiano, Lawrence J., "Modeling the Liquidity Effect of a Money
Shock," Federal Reserve Bank of Minneapolis Quarterly Review,
Winter 1991. 15(1). 3-34.
Christiano, L.J., and M. Eichenbaum, (1992) "Liquidity Effects,
Monetary Policy and the Business Cycle," unpublished ms..
Federal Reserve Bank of Minneapolis, Northwestern University,
NBER and Federal Reserve Bank of Chicago, July 1992.
Fuerst, T.S. "Liquidity, Loanable Funds, and Real Activity,"
Journal of Monetary Economics, vol. 29 (1992), 3-24.
Goodfriend, Marvin. "Discount Window Borrowing, Monetary Policy,
and the Post-October 6, 1979 Federal Reserve Operating
Procedure," Journal of Monetary Economics, vol. 12 (1983),
343-56.
, "Interest Rate Smoothing and Price Level TrendStationarity, " Journal of Monetary Economics, March 1987,
19, 335-48.
Grossman, S. J., and L. Weiss. "A Transact ion-based Model of the
Monetary Transmission Mechanism," American Economic Review,
vol. 73 (1983), 871-80.
Lucas, R.E., Jr. "Liquidity and Interest Rates," Journal
Economic Theory, vol. 50 (1990), 237-64.

of

McCallum, Bennett, T. (1981) "Price Level Determinancy with an
Interest Rate Policy Rule and Rational Expectations,"
Journal of Monetary Economics, vol. 8 (November).
, (1986) " Some Issues Concerning Interest Rate
Pegging, Price Level Determinacy, and the Real Bills
Doctrine," Journal of Monetary Economics vol. 17 (January).
Rotemberg, J.J. "A Monetary Equilibrium Model with Transaction
Costs," Journal of Political Economy, vol. 92 (1984), 40-58.
Strongin, S. "The Identification of Monetary Disturbances:
Explaining the Liquidity Puzzle." Unpublished manuscript,
Federal Reserve Bank of Chicago, December 1991.




Credit Conditions and External Finance:
Interpreting the Behavior of Financial Flows and Interest Rate Spreads
Kenneth N.Kuttner1

Aflurryof recent macroeconomic research has drawn attention to the relationship between
monetary policy, credit conditions, and the markets for short-term debt Two recent papers have
focused onfirms'substitution between bank and non-bank externalfinancein particular, proposing
macroeconomic indicators based onfinancialmarket activity. Kashyap, Stein, and Wilcox (1992)
employ quantity data directly, arguing that the share of bank loans out of firms' total short-term
finance is an informative index of Federal Reserve policy and loan availability more generally. In
a complementary line of research, Friedman and Kuttner (1992) identify monetary policy and bank
lending as potential sources offluctuationsin the spread between yields on commercial paper and
Treasury bills. While both papers have demonstrated solid empirical links between these financial
indicators and real economic activity, neither hasrigorouslyassessed the extent to which fluctuations in these indicators actually represent exogenous changes in credit conditions, rather than
endogenous responses to changing economic conditions. This paper's goal is to provide such an
assessment.
The paper begins with a sketch of the mechanism through which credit conditions affect firms'
short-termfinancing,drawing a distinction between the effects of the Federal Reserve's open market
operations and other factors influencing banks' willingness to lend. The second section summarizes
the reduced-form relationships between real output, the interest rate, and three alternative indices
1. Senior Economist, Federal Reserve Bank of Chicago. I am grateful to Benjamin Friedman and
David Wilcox for their comments and suggestions.




-1-

Kuttner

of credit conditions: the composition of external finance, the spread between the loan rate and the
commercial paper rate, and the analogous spread between commercial paper and Treasury bills.
The third section turns to a closer examination of the impact of monetary policy and loan
availability on bank and non-bank finance using structural VAR techniques. Identifying monetary
policy with innovations to non-borrowed reserves and controlling for firms'financingrequirements,
the first of the three models estimates the dynamic effects of monetary and lending shocks on the
composition of external finance, the interest rate, and real output. The second structural VAR system assesses the effects of reserves and lending shocks on the paper-bill spread. The third model
identifies lending shocks with innovations in the loan-paper spread. Estimates of these models confirm that all three variables respond appropriately to reserves shocks. In addition, lending shocks,
whether identified through financial flows or via fluctuations in the loan spread, induce a substitution
between bank and non-bank finance.
Less clear is the extent to which any of these measures exclusively reflects the effects of changing loan availability. The fact that positive lending shocks are associated with increases in the interest rate and the paper-bill spread suggests that changes in the composition of external finance have
more to do with firms' financing requirements than with exogenous changes in banks' willingness to
lend. Another slightly puzzling observation is that the largest source of changes to the composition
of external finance seems to be wholly unrelated to both reserves and bank lending. Together, these
two results suggest that while credit conditions are one important determinant of firms' choice of
financing, short-term debt flows may be informative for reasons other than those involving the substitution between bank/non-bank substitution. Although its implications for real activity are rather
weak, the loan spread appears to be a plausible alternative measure of credit conditions.

A model offinancialflowsand interest rate spreads
How do the markets for short-term bank and non-bank finance respond to monetary impulses? And
how do non-monetary shocks affect these markets? And how might one construct an index of the
availability of intermediated funds?




-2-

Kuttner

As a first step towards answering these questions, this section analyzes a simple model of the
markets for commercial paper, bank loans, and Treasury bills in the style of Brainard (1964) or
Bosworth and Duesenberry (1973). While not as detailed as either of those models, it is adapted to
highlight firms' tradeoff between bank and non-bank finance. It also draws an important distinction
between purely monetary influences acting through open market operations, and credit conditions
defined more broadly, which may include other factors affecting banks' willingness to lend.
One of the model's more obvious properties is that an injection of reserves causes the interest
rate to fall — the familiar "liquidity effect." Reserves injections also cause the spread between the
interest rates on bank lending and commercial paper to fall, and leads to increased reliance on bank
finance. Lending shocks, which are assumed to affect only banks' preferences over alternative assets, turn out to have similar effects on the loan-paper spread and the composition of firms' finance.
Lending shocks, by contrast, have no effect on the level of interest rates — only the spreads.
The model also identifies two other factors with implications for the money market. First,
firms' demand for external finance may induce changes in the relevant interest rate spreads and
consequently the composition of finance; controlling for this demand-side influence turns out to be
a major challenge to the construction of an empirical measure of credit availability. Similarly, the
stock of outstanding Treasury bills may have tangible effects on the spreads and the composition of
finance.
The three players in the money market are households, banks, and firms, who participate in
the markets for reserves, commercial paper, Treasury bills, and loans. Specifically, households'
portfolios include demand deposits (DD), commercial paper (P), and Treasury bills (B) according
to




DD* = <Krp) W, 4>' < 0
df
df
P* s flrp, rB)W, — > 0 and — < 0
brp
drg
B^ s (1 - <J) - / f o rB)) W,

-3-

Deposit demand
Paper demand
Bill demand

Kuttner

where W is the sum of deposits, paper, and bills held by households. Households' demand for
non-interest-bearing bank deposits is a decreasing function of the prevailing paper rate, rP. A key
assumption is that households view commercial paper and Treasury bills as imperfect substitutes,
so that changes in their relative supplies affect their respective yields.2 Households require a higher
paper rate (or a lower bill rate) to hold a larger share of their portfolio as commercial paper.
Demand deposits are banks' sole liability. Their assets are divided among Treasury bills, loans
(L), and deposits at the Federal Reserve (R) according to:
K* s p(rp)DDt p ' < 0
Ld = Sin, rP, \)DDf — > 0 and — < 0
drL
drP
B*b s (1 - p(rP) - g(rLf rP, K))DD.

Reserve demand
Loan demand
Bill demand

Banks' demand for non-interest-bearing reserves falls with the prevailing paper rate, while loan
demand is increasing in the loan rate and decreasing in the paper rate.3 The stock of reserves is set
at R' by the Federal Reserve; discount window borrowing is ignored.
Banks' demand for loans is also allowed to depend on the variable X, representing any other
factors affecting banks' willingness to lend. These "lending" shocks lead banks to shift the composition of their portfolios between bills and loans; negative shifts in X may be interpreted as "credit
crunch" episodes. These may occur in reaction to a perceived deterioration in borrowers' creditworthiness, or to more stringent capital requirements as suggested by Bernanke and Lown (1991). They
may also be the result of the "moral suasion" instrument of monetary policy; Owens and Schreft
(1992) identify a number of episodes in which banks contracted their lending in response to Federal
Reserve pressure. Whatever the source, the key feature of these "lending" shocks is that they need
not be accompanied by overt monetary policy in the form of open market operations.4
2. Friedman and Kuttner (1992) discuss some possible reasons for this imperfect substitutability.
Lawler (1978) also finds evidence for imperfect substitutability at seasonal frequencies.
3. Note that throughout the paper, assets are "demanded" while liabilities are "supplied." Hence,
banks "demand" loans and bills, while firms "supply" loans and paper.
4. This point is stressed by Friedman (1991).




-4-

Kuttner

Finally,firmschoose between bank lending and paper issuance as sources of short-term finance
according to
fth

hh

dri

orp

V a h(rL, rP)F, —- < 0 and —- > 0

Loan supply

P* » (1 - h(rL, rP))E

Paper supply

For simplicity, the amount to befinanced,F, is assumed to be exogenous with respect to the various
interest rates. Becausefirmsview loans and paper as imperfect substitutes, they willfinancesome
portion of F through bank lending even though rL generally exceeds />; as discussed by Kashyap,
Stein and Wilcox (hereafter KSW), this presumably reflects some intangible benefit accruing to the
firm from maintaining a relationship with a bank. Firms * share of bankfinance(the KSW "mix")
responds predictably to the loan and paper rates: an increase in the loan rate (or a decrease in the
paper rate), leadsfirmsto substitute away from bankfinancetowards non-bank external finance.5
In equilibrium, the demand for the four assets equals their supply,
p(rpMrP)W = l?
frurpiKftW-hirurpyF^O
Kr»rg)W-{l-h(n,r,)yFmO
(1 - g(rL, rPt \))$W+ (1 - / f a rB) - +)W = B*9
determining yields and quantities as functions of the exogenous /?', X, F, and B*. Walras' law
allows the bill market equation to be dropped. Further simplification is possible by assuming the
asset demand and supply functions to be homogeneous of degree zero with respect to the assets'
5. This model embodies the assumption that bank and commercial paperfinanceare viable alternatives for an economically relevant group offirms.However, there is increasing evidence that this
set offirmsis rather small, and that much of the observed variation in the aggregate composition of
finance is due to the relative availability offinanceto small and largefirms;see Gertler and Gilchrist
(1992) and Oliner and Rudebusch (1992).




-5-

Kuttner

yields, so that (for example) g(n +c,rp + c, X) = gin, />, X) for any constant c. In this case, the/, g
and h functions can be specified in terms of interest rate spreads, and the system reduces to:

gizLP,mrpW-h(zu>)F = 0

(I)

KzpBW-(l-h(zu>))F = 0
where zLP and zPB denote the loan-paper and paper-bill spreads.
Analyzing*the comparative statics of (1) is simplified by its (somewhat artificial) recursive
structure. The interest rate level is entirely determined by supply and demand in the market for
reserves; the fall in reserves resulting from a contractionary open market operation requires a higher
rate to equilibrate the reserves market, as illustrated in Figure l. 6 This higher interest rate leads in
turn to a shrinkage of demand deposits and the banking system as a whole. Banks respond by raising
the loan-paper spread, prompting some of its borrowers to switch to alternative forms of finance—
short-term paper in this model. The increased supply of paper (relative to bills) leads to a widening
spread between the paper and bill rates.
The effects of an adverse lending shock resemble those of a reserves contraction in that both
produce a rising loan spread and a substitution towards non-bank finance. Although both shocks
produce similar effects on banks' portfolios, they differ in one important respect: reserves shocks
affect the level of the short-term interest rate, while lending shocks leave the paper rate unchanged.
A fall in X leads banks to shift the composition of their portfolios away from loans and into Treasury
bills, leaving their reserve demand and the paper rate (and consequently deposits and the banking
system's size) unchanged. Banks increase their spreads relative to the paper rate in order to reduce
their stock of loans. As before,firms'increased reliance on commercial paper drives up the paperbill spread.
6. Total wealth is held constant in an open market operation, as the withdrawal of reserves is
offset by a sale of Treasury securities.




-6-

Kuttner

The observation that both reserves and lending shocks may contribute to real economic fluctuations is one explanation of the widespread interest in constructing a broader measure of credit
conditions than reserves or the interest rate in isolation, which reflect largely those shocks originating from the reserves market The attractive feature of the credit conditions indicators discussed
here is their ability to detect the effects of changes in loan availability and reservesfluctuations:in
this model, the "mix," the loan-paper spread, and the paper-bill all reflect the impact of both types
of shocks. In fact, in the absence of any other shocks, all three of these measures should respond to
monetary and credit factors in qualitatively similar ways.
One problem common to all three of these measures (and the interest rate itself) is their susceptibility to contamination from changes infirms'overall demand forfinancing,which may alter yield
spreads and the composition of externalfinancefor reasons having nothing to do with to exogenous
changes in credit conditions.7 This can be illustrated by examining the comparative statics of (1)
in response to an increase in F, the dollar amount of fundsfirmswish to raise from the short-term
credit markets. A greater demand for loanable funds unambiguously increases the prevailing interest rate, />. Its effects on the loan-paper spread (and therefore the composition of external finance)
is ambiguous, as it depends onfirms9share of bankfinance(/t) relative to households* wealth fraction in bank deposits (<|>), and the share of banks' portfolios held as loans (g). When h(zLP) > tyrp)g
(as is presumably the case), increases in F cause loan demand growth in excess of deposit growth,
driving up the relative cost of bankfinanceand the share of paper infirms'external finance.8 The
same inequality is also relevant for the paper-bill spread; a second sufficient condition for a rising
spread is that (1 - h(zLP)) > J{zpB\ so that the increasing paper demand would require households to
hold a larger share of paper in their portfolios.
7. Under most of the Federal Reserves' post-Accord operating procedures, non-borrowed reserves may also be contaminated in this way; see Strongin (1991).
8. A special feature of the KSW model is that changingfinancingrequirements affect loans and
paper proportionally, leaving the "mix" unchanged.




-7-

Kuttner

One additional complication for interpreting the paper-bill spread as a measure of credit conditions is that it may be affected by changes in the outstanding stock of Treasury bills. In addition,
the wealth effects associated with changes in the volume of Treasuryfinancemay alter the level of
interest rates and loan spread, and consequently the composition of external finance.9 In this model,
an increase in the supply of bills reduces the paper-bill spread, as investors require higher returns
to entice them to hold the additional stock of bills. This increase in banks' demand for loans leads
to a fall in the loan rate relative to the paper rate, and increased reliance on bank finance.
To summarize, the model's main implications are:
• Both reserves and lending shocks alter the relative price of bank and non-bank finance, inducing a substitution between alternative forms of external finance.
• By affecting the supply of commercial paper, this substitution also affects the relative
yields on Treasury bills and commercial paper.
• Changes in reserves affect the level of interest rates, while lending shocks leave the
level unchanged.
• Firms' overallfinancingrequirements may affect interest rate spreads and their composition of short-term finance.
The goal of the paper's subsequent empirical work is to explore these implications. Specifically, it
attempts to identify lending shocks through their impact on the composition of externalfinanceand
interest rate spreads, while controlling for reserves and the overall demand for loanable funds.

Short-term credit markets and real economic activity
One desirable feature of any index of credit conditions is a systematic link between it and subsequent
fluctuations in real economic activity.10 The results below summarize the predictive properties of
the KSW "mix," the prime-paper spread, and the paper-bill spread. The results show that the "mix"
9. Of course, this assumes that households view government bonds as net wealth; see Barro
(1974).
10. Economists and market observers have long recognized the cyclical properties of commercial
paper, bank lending, and their relative yields; see, for example, Foulke (1931), Selden (1963), and
Stigum (1990).




-8-

Kuttner

and the paper-bill spread are good predictors of future changes in real GDP (although this alone does
not justify their interpretation as measures of credit availability).
"Causality" tests
Table 1 examines the incremental information content of the three measures for future changes in
real GDP in the presence of traditional measures of monetary policy: non-borrowed reserves and
the commercial paper rate. Regressions 1-3 are four-variate reduced-form equations of the form
4

4

4

4

Ax, a Ho + Hi* + ] T OjAx^ + ^T pi[A ln(J?),w + ] T Y,Arj>^ + ] T 6 , A ^ + e,
where x is the logarithm of real GDP, R is non-borrowed reserves adjusted for extended credit and
deflated by the GDP deflator, i> is the commercial paper rate, and q denotes, in turn, the "mix", the
loan-paper spread, and the paper-bill spread. As in KSW, the "mix" is computed as the observed
ratio of bank lending to the sum of lending to commercial paper, or L/(L + P).n The results use the
six-month commercial paper and Treasury bill yields, and the prime rate (from the Federal Reserve
H.1S release) is used as the lending rate.
The table reports F-tests for the exclusion of the four 6, terms for the entire 1960:2-1991:4
sample, as well as two shorter samples. One truncated sample begins in 703, when Regulation Q
was eliminated forroostlarge CDs.12 Another begins in 1975:1. Although this date is somewhat
arbitrary, it corresponds roughly to the beginning of a rapid expansion of the commercial paper
market, during which it became a more popular vehicle for non-financialfirms'short-term finance.13
11. The augmented Dickey-Fuller u statistic (computed with eight lags) for the stationarity of the
"mix" is -4.10, rejecting the null hypothesis of nonstationarity at the 1% level. Consequently, it is
included here in levels along with a linear trend term.
12. Regulation Q interest rate ceilings on 30-89 day CDs in denominations of $100,000 were
eliminated on June 24, 1970. Ceilings on CDs with maturities in excess of 90 days remained in
place until March 16,1973.




-9-

Kuttner

The 1975-91 sample also excludes the Penn Central and Franklin National disruptions of 1970 and
1974, and covers the period in which ratings were assigned to commercial paper issues.14
The results of thefirstregression corroborate the strong link between the "mix" and real output
found by KSW, supporting theirfindingthat the composition of finance has significant predictive
power for future real economic activity, even in the presence of reserves and interest rates. The
poor performance of the loan-paper spread in the second regression (again in the presence of reserves and the commercial paper rate) is consistent with the notion that banks' lending rates are
relatively uninformative.15 The third regression demonstrates the incremental information content
of the paper-bill spread — at least in the earlier samples.
Impulse responses
While the F-statistics for "causality" give some indication of the strength of the predictive power of
thesefinancialindicators, they give no indication of the size or direction of their impact. The impulse
response functions plotted in Figure 2 provide a richer description of the effects of innovations
to the financial indicators. Each of the three rows of graphs is from the VAR corresponding to
regressions 1-3 in Table 1. In each case, the system has been orthogonalized (according to the
triangular Cholesky decomposition) with the credit conditions index in last place. Three responses
are plotted for each regression: thefinancialindicator's effects on output and the interest rate, and
the effect of reserves innovations on thefinancialindicator. The dotted lines depict the approximate
95% confidence bounds.
Panels (a) and (b) from the first specification show that "mix" innovations indeed act like
reasonable measures of credit conditions; reserves injections increase the share of bank loans, and
13. At the end of 1974, non-financial commercial paper accounted for only 13.5 billion dollars.
By 1982, thisfigurehad grown 325.2 percent to 57.4 billion. See Hurley (1977,1982), and Stigum
(1990).
14. Moody's and Standard and Poor's began rating commercial paper in 1974.
15. Similar results are obtained with the average of large banks' lending rates obtained from the
Federal Reserve Survey of Terms of Bank Lending reported in release E.2.




-10-

Kuttner

output rises in response to positive "mix" shocks, which might be interpreted as the pure lending
component of credit conditions. The panel (c) plot, however, is something of a puzzle. It shows that
"mix" innovations are associated with a rising commercial paper rate—not what one would expect
from an increased willingness to lend on the part of banks, and inconsistent with the implications
of the model presented earlier.16 However, this pattern is consistent with banks passively supplying
more loans in response to rising demand for credit.
The second row of plots confirm the generally weak relationship between the prime-paper
spread and real output. One interesting feature of the loan spread is that it initially rises in response
to a reserves innovation — clearly inconsistent with the loosening of credit conditions implied by
the reserves injection. The loan spread ultimately falls, however, suggesting that this response is
due to a certain sluggishness in the way banks adjust their lending rates.
The impulse response functions from the paper-bill spread regression are all consistent with
what one would expect from an indicator of credit conditions: positive shocks to the spread generate
declining real output, while reserves injections reduce the spread. Furthermore, unlike the "mix",
innovations in the spread itself have essentially no impact on the level of interest rates.
Comparing the "mix" and the paper-bill spread
Because regressions 1-3 included each of the credit conditions measures in isolation, the results raise
an important question: to what extent are the three indicators measuring the same phenomenon? An
obvious way to address this question is to include more than one indicator in the same regression to
see if the presence of one vitiates the predictive power of the other.
The results from two additional regressions (numbered 4 and 5) are reported in Table 1. The
results from specification 4, which includes both the "mix" and the loan spread, are not surprising
given the weak performance of the loan spread in isolation — the F-statistics for the "mix" remain
virtually unchanged. Somewhat more surprising are the results from specification 5, in which both
the "mix" and the paper-bill spread appear. Here, the relationship between the two variables and real
16. The "mix" terms are significant in the interest rate equation at the 10% level.




-11-

Kuttner

output is uniformly stronger (judged by the F-statistics) than when they are included individually.
Qearly, one (or both) of the indicators is doing something other than simply summarizing the state
of credit market conditions.
The roles of commercial paper and bank loans
The model sketched earlier suggests that flows of commercial paper and bank lending are informative to the extent that they reflect the substitution between the two forms of finance in response to a
monetary or a lending shock. KSW exploit this insight by looking at the ratio of bank loans to the
sum of loans and paper, shocks that affect both forms of debt proportionally are presumed to stem
from sources other than loan availability. A useful check on this specification is to verify that paper
and lending flows enter an unrestricted regression in such a way that the "mix" is the variable that
matters.
This is easily accomplished by differentiating the "mix" (designated h) with respect to time,
P
—
L
dt = (I+P)*

L

P

(I+P) 2

= h{\ - h%/L - /i(l - h)P/P9
decomposing its movements into distinct lending and paper contributions. In discrete time, the
analogous decomposition,
AA, - AM(1 -

AKI)AL/1M

- A M (1 - A,-i)AP/P M -

<&L -

Afip

expresses A/i as a weighted sum of commercial paper and bank loan growth rates, denoted tJxL
and tJip. If AA were in fact the appropriate measure of the impact of credit conditions on the real
economy, the two components would enter real output regressions with equal and opposite signs;
the regression itself would "choose" the KSW specification.
Table 2 displays the results of this experiment. Panel (a) reports the outcome of a regression
of first-differenced log real GDP on four lags of output, tJxL and tJiP over the 1960:2-91:4 sample.
Judged by the F-statistics, the commercial paper terms are much more informative than the lending




-12-

Kuttner

terms; tJip is significant at the 0.01 level, while the tJiL terms are not significant at even the 0.10
level.17 The sum of the estimated coefficients on lending is negative, but statistically insignificant
The regression in panel (b) refines the test by specifying the regression in terms of tJi and
tJip — simply a transformation of the regression in panel (a). Excluding the four lags of &hP is
equivalent to restricting the coefficients on iskL and tJip to have equal and opposite signs. Here,
the tJi terms are statistically insignificant, while the AhP terms are significant at the 0.05 level.
Moreover; the negative estimated sum of the "mix" coefficients is inconsistent with the substitution
hypothesis, although this sum is again statistically insignificant.
To guard against the possibility that the results in the first two panels are an artifact of the
differenced specification, panel (c) reports the results of a regression that includes a linear trend and
h in levels. While not tJ\P terms are not as strong in the levels specification, the coefficients on the
h terms remain statistically insignificant.
These experiments show that the "mix" owes its predictive power in large part to something
other than the substitution between bank and paper finance. In unrestricted equations, h terms are
generally insignificant, while the hypothesis that commercial paper in isolation does not matter for
predicting real output can be rejected. This observation suggests a closer examination of lending and
commercial paper flows individually, and their relation to monetary policy and credit conditions.

A structural approach to identifying lending shocks
The atheoretical results in the preceding section provided some evidence in favor of interpreting the
financing "mix" and the paper-bill spread as measures of credit conditions, although innovations in
the composition offinancewere, contrary to the simple model, are associated with a rising interest
rate. One reason for this pattern may be the result of inadequately controlling for the overall demand
for short-term finance. As demonstrated earlier, an increase in the amount to befinancedneed not
raise bank and non-bankfinanceproportionally. In this case, if increases infirms'demand for funds
17. This is consistent with the results of King (1986).




-13-

Kuttner

are accommodated primarily through bank lending, the "mix" may rise for reasons unrelated to
credit conditions.
Figure 3 plots thefinancinggap (defined as the difference betweenfirms'capital expenditures
less inventory IVA and after-tax internal funds) along with commercial paper and bank loan flows,
demonstrating the close relationship between thefinancinggap and the volume of bank lending
(although commercial paper appears to have become more sensitive to thefinancinggap in the later
part of the sample). To control for credit demand, the results in this section include the financing
gap as an additional determinant offirms'debt issuance.
A more interesting alternative hypothesis is that is that the substitution mechanism inadequately explains the joint behavior of commercial paper and bank lending, and that factors other
than monetary policy are what drive the observedfluctuationsin the composition of short-term external finance. The apparent asymmetry between the effects of loan and paper flows uncovered in
Table 2 provides some circumstantial evidence for this view.
The results presented in this section attempt to address these issues by separately analyzing
flows of lending and commercial paper in a structural VAR setting that controls for the overall
demand for loanable funds. Moving to a more structural approach also addresses the possibility that
the interest rate's odd response to "mix" shocks is as an artifact of the artificial triangular structure
of the Cholesky decomposition employed earlier. Thefirstmodel focuses on the response of lending
and paper flows to reservesfluctuations,and examines the properties of the innovations identified
as lending shocks. The second describes the response of the paper-bill spread to thefinancialflows
generated by reserves and lending shocks. The third usesfluctuationsin the loan-paper spread as
an alternative means of identifying lending shocks.
A review of structural VARs
Beginning with an unrestricted i-variate dynamic simultaneous equation system,




T

-14-

Kuttner

the standard VAR achieves identification by restricting the contemporaneous relationships between
the elements of y, i.e., by setting BQ = 0 and A = /, while placing no restrictions on the covariance
matrix of v, ie., £(w') = Q. The structural VAR introduced by Blanchard and Watson (1986) and
Bernanke (1986) achieves identification by allowing some nonzero elements in thcB0 matrix, while
restricting the covariance matrix of v, the structural disturbances, to be diagonal. Off-diagonal
elements in A can be introduced to allow distinct elements of y to depend on common structural
shocks. Thus, structural VARs differ from traditional structural models by replacing the assumption
of an exogenous instrument set with the assumption of orthogonal structural shocks. At the same
time, the dynamics of the system are left unrestricted, as in the conventional VAR.
Another interpretation of the structural VAR is as a decomposition of the covariance matrix of
VAR residuals. If the structural disturbances are uncorrected with one another, Lc, £(w') = D, Q,
the covariance matrix of the VAR errors becomes a nonlinear function of the structural parameters:
Q=£(J#Av,v,'A'J#)
mBfADA'Bf.
If the system is just-identified, the above equality is exact; B^AD™ is a matrix square root of Q,
and A'1 B0 diagonalizes Q.18
Reserves, lending, and short-term debt flows
Thefirstmodel is a just-identified six-variable system involvingfinancinggap (F)> bank lending,
non-financial commercial paper (P) the commercial paperrate(r>), real GDP (x), and non-borrowed
reserves adjusted for extended credit (R). The interest rate is differenced, while reserves and GDP
enter as log differences. The lending and paper data are again taken from the Flow of Funds accounts
for the non-farm, non-financial corporate and noncorporate sectors. With F, P and L expressed
18. With a total of 2A2 elements in A and B0 and only &(&+1)/2 unique elements in Q, it is clear that
the stnicmral parameters are not identified without additional restrictions on A and£0. The Cholesky
decomposition, which is equivalent to setting B0 = / and making A lower triangular, is but one possibility. In overidentified systems, the problem becomes one of choosing the stnicmral parameters
in Bo and A to generate the bestfitbetween thefittedand the observed covariance matrices.




-15-

Kuttner

as shares of the total dollar volume of outstanding paper and loans, changes in the "mix" can be
constructed as the weighted average of the two flows:
v

'L+P

L+P

The substance of the model is contained in the six equations describing the contemporaneous
relationships between the variables,
R as bU6x + vx

Reserves (2a)

F as bxxR + bz& + V2

Financing gap (2b)

rP ss bx\R + bxiF + V3

Interest rate (2c)

L as b<xR + b^iF + d<30» + v4
P ss b^R + fci2P+*s^> + *MV4 + v5
x as d 0 r P + 644I + fc^P + v6

Lending (2J)
Paper (2e)
Output (20.

No restrictions are placed on the dynamics of the system; consequently, terms dated t-\ and before
are omitted, but implicit.
Equation 2a allows the Federal Reserve to vary reserves contemporaneously with real GDP in
a primitive feedback relationship. Thefinancinggap (equation 2b) also depends on the level of real
economic activity. Consistent with the model presented earlier, the commercial paper rate in 2c is
a function of reserves and thefinancinggap.
The model's key equations are 2d and 2e, describing the behavior of bank lending and commercial paper flows as a function of thefinancinggap, reserves, and the interest rate. The coefficients
on F measure the proportion of the currentfinancinggap satisfiedfinancedthrough loans and paper. The two equations' coefificients on R determine the immediate response, ceteris paribus, of
the two forms of short-termfinanceto changes the banking system's reserve position. The v4 term
in the lending equation represents lending shocks that are orthogonal to reserve andfinancinggap
innovations, which would include factors such as credit crunches. For this interpretation of v4 to be




-16-

Kuttner

legitimate, one of two conditions has to hold: either the observed financing gap must appropriately
control for firms' demand for funds, or the amount of funds banks have available is fixed in the
current quarter.
The v4 innovation also appears in the commercial paper equation with the coefficient a^, allowing commercial paper to respond directly to lending shocks. This parameter determines the
extent to which lending shocks are "recycled" into the commercial paper market within the current quarter. The v5 term in the commercial paper equation accounts for shocks to paper issuance
uncorrected with the other structural disturbances. The final equation for real GDP is a reducedform equation describing the economy's response to the reserves and credit shocks in the preceding
equations.
The parameter estimates in Table 3 summarize the model's contemporaneous behavior, while
the impulse responses functions plotted in Figure 4 describe its dynamics of the system whose orthogonalization is implicit in equations 2a-2f. Like the earlier reduced-form regressions, these
results provide some evidence to support the use of lending flows as an indicator of credit conditions, while confirming the doubts raised in the atheoretical VARs. First, The negative estimate
of the coefficient on R in the lending equation (2d) contradicts the hypothesis that the primary effect of monetary policy is a substitution between bank and non-bank finance; the contemporaneous
response of an injection of non-borrowed reserves, ceteris paribus, is a fall in bank lending.
However, because of the contemporaneous relationship from reserves to the financing gap and
short-term finance via the interest rate and output, the coefficients on R in equations 2d and 2e do not
by themselves determine the overall response of the "mix" to a reserves shock. The actual responses
can be read from the impulse response function, plotted in the top panel of Figure 4.19 This shows
that the net effect of a reserves injection is initially rather small, with the loan share gradually rising
after two to three quarters.
19. The sample average values of h are used to compute the approximate response of the "mix"
from the impulse response functions of the underlying variables.




-17-

Kuttner

Figure 4 also shows that lending shocks seem to have a considerably larger impact on the
composition of externalfinancethan reserves for thefirstfour quarters. Lending shocks9 effect is
strengthened somewhat by the statistically significant negative estimate offl$4>which is consistent
with roughly 10% of the lending shock being "recycled" into the paper market in the current quarter.
The coefficients on F in the paper and lending equations show that neither responds immediately to
fluctuations in thefinancinggap.
A strong liquidity effect is associated with injections of non-borrowed reserves; the paper rate
falls contemporaneously (the negative coefficient on R in equation 2c) and over a longer horizon (the
center panel of Figure 4). These results also confirm the curious positive relation between the "mix"
and the level of interest rates highlighted earlier in the paper. The center panel shows that positive
lending innovations imply a rising interest rate, contradicting the theoretical model's implications
for the effects of lending shocks.
Both monetary and lending shocks are important sources of output fluctuations. Increased
bank lending is contemporaneously associated with more rapid real GDP growth in the short run,
as shown by both the positive (but not quite significant) coefficient on L and the impulse response
function.
What about shocks to commercial paper, v5? The top panel of Figure 4 shows that these shocks
— which are, by construction, orthogonal to the system's other structural disturbances — have the
largest and most persistent impact on the composition of external finance. Interestingly, the center
panel shows that these innovations have essentially no implications for the interest rate, although
they do seem to have a small, negative impact on real output.
Financial flows and interest-rate spreads
Recent papers by Bernanke (1990) and Friedman and Kuttner (1992) suggest that the substitution
between bank and non-bank debt is an important source of fluctuations in the paper-bill spread. As
discussed earlier, monetary contractions reduce lending by shrinking the stock of deposits, leading




-18-

Kuttner

firms to raise the loan rate relative to the paper rate, discouraging intermediated borrowing. Similarly, adverse lending shocks cause banks to shift from loans to Treasury bills. Asfirmsturn to the
paper market to satisfy theirfinancingneeds, the paper supply rises and bill supply to households
falls, raising the paper-bill spread. If this is the way in which credit conditions affect the spread,
one would expect tofindthe mechanism operating through the volume of outstanding non-financial
commercial paper.
The second structural VAR is designed to detect the operation of this mechanism. It augments
thefirstmodel (equations 2a-2f) with the addition of a seventh equation for the paper-bill spread,
x a b&rP + b^JL + b^sP + b^(rP - rB) + v6

Output (3/)

rP-rB = bitiR + 6 7f2 F+b^rp + 67,4^ + bn,$P + v?, Paper-Bill spread (3g)
and also includes the spread in the output equation. The remaining five equations are identical to
those in the earlier model (2a-e).
The parameter estimates reported in Table 4 provide weak evidence for bank/non-bank substitution as a source of paper-bill spread. The positive and marginally significant on the paper term
shows that flows of non-financial paper do exert an influence on the spread.20 However, the very
large, significant coefficient on reserves shows indicates that a great deal of the impact of monetary
policy is transmitted to the spread via other routes.
The impulse responses in the top panel of Figure 5 confirm the spread's strong reaction to
non-borrowed reserves innovations. Positive shocks to thefinancinggap also drive up the spread,
as predicted, while paper shocks have little or no impact. Lending shocks again pose a problem,
however. If the lending innovations identified by the VAR correspond to changes in the availability
of loans, the model suggests that positive shocks should be associated with a falling paper-bill
spread. The opposite is true: lending shocks imply a rising spread. Again, this pattern is consistent
20. By contrast, the results in Table 10 of Friedman and Kuttner (1992) using the total volume of
commercial paper outstanding are consistent with a stronger link between paper issuance and the
spread.




-19-

Kuttner

with bank lending responding passively to changes in the demand for funds inadequately captured
by the financing gap.
The bottom panel of Figure 5 suggests something other than bank/non-bank substitution is
driving the paper-bill spread. Despite the inclusion of a variety of financial variables purporting to
capture the impact of monetary policy on credit markets, the graph shows that the spread continues
have strong implications for future output — comparable in magnitude to those of non-borrowed
reserves. Even accounting for reserves, lending, and paper shocks, orthogonal spread innovations
still result in falling real economic activity.
Identifying lending shocks with loan spread innovations
In light of the conclusion that lending flows (and the "mix") may in part represent endogenous
response to firms' financing demands, the third structural VAR uses an alternative assumption to
identify lending shocks, attributing (orthogonalized) innovations in the loan-paperspread to changes
in banks9 willingness to lend. In the context of the simple model presented earlier, the loan spread
should embody exactly the same information as the "mix." In practice, as KSW note, the loan rate
is likely to be a poor measure of the true cost of bank finance, an observation that motivates their use
of the quantity variables. Indeed, the sluggish response of the loan rate to changes in the paper rate
corroborates this view. The weak response of output to the loan-paper spread makes this approach
seem even less promising.
With these reservations in mind, thefirststructural VAR can be adapted to incorporate the loanpaper spread. An equation for the loan spread is added to the system, and lending and paper flows
are allowed to depend on this spread, as well as on reserves and the financing gap. The covariation
between the flows that is a function of credit conditions is a result of their common dependence on
the loan spread. This identification scheme will work if the financing gap is an imperfect proxy for
the overall demand for funds so long as banks passively accommodate firms' funding requirements
within the quarter at the going spread (that is, if their demand for loans is elastic).




-20-

Kuttner

The modified system is:
R s bi& + Vi

Reserves'(4fl)

F m bZ\R + bZ6x + V2

Financing gap (4b)

r> « buR + b^JF + v3

Interest rate (4c)

rL-rP = b<tR + b^F+b^rP

Loan spread (4rf)

I =fcMrt+ 6 ^ + b & r p + 65,4(^1 - rP) • v5
P « 6Mtf + *42F+fc^r/. -•- A^fo. - TP) + **sV5 + v6
* s ^73r/» + ^ f a , - rP) + 67,5^ • th,*? + ^7

Lending (4e)
Paper (4/)
Output (4g).

Under the assumptions outlined above, the innovations to the loan spread equation are now associated with changes in credit conditions, while the v5 lending innovations represent shocks to firms'
loan supply (that is, their demand for funds).
The parameter estimates in Table 5 accord surprisingly well with the implications of the model.
Although its sluggish response makes the loan spread is subject to large, transitory effects from
the paper rate and reserves, the negative estimated 65,4 and the positive b$4 show that loan and
paper volume respond as they should to the spread. Furthermore, reserves have no discernible
independent impact on financial flows. A rising loan spread is contractionary, although again, the
effect is statistically weak.
The corresponding impulse response functions appear in Figure 6. The top panel again illustrates the consequences of sluggish loan rate adjustment, with reserves injections causing the
loan spread to rise sharply in the current quarter. Over time, reserves innovations produce a falling
spread. The center panel shows the familiar liquidity effect, and the positive impact of lending innovations on the commercial paper rate. In this model, however, with innovations to loan volume
interpreted as shocks to firms' funding requirements, the result is perfectly natural. By contrast,
innovations in the loan spread have quite mild effects on the paper rate.




-21-

Kuttner

Conclusions
This paper has examined the relationship between monetary policy, loan availability, and alternative
indicators of credit market activity. One of is mainfindingsis that the substitution between bank and
non-bank finance is indeed an identifiable effect of monetary policy as measured by innovations to
non-borrowed reserves. This substitution is, however, not the only factor affecting financial flows.
One of the major contributors to the aggregate composition offirms'short-term obligations is flows
of commercial paper unrelated to lending shocks.
Furthermore, the portion of bank lending not attributable to monetary policy is associated with
increases in the commercial paper rate and the paper-bill spread, suggesting that the behavior of the
KSW "mix" is in part due to changes in firms' demand for loanable funds. Despite its apparent
slow adjustment to changes in market interest rates, the loan-paper spread is a plausible alternative
indicator of credit conditions.
The paper-bill spread responds appropriately to monetary shocks, rising in response to a reserves contraction. However, the strength of its response cannot entirely be accounted for by flows
of non-financial paper, suggesting that its informativeness as a predictor of real economic activity
may be due to other sources, such as changes in banks' issuance of negotiable CDs. This is consistent with the observation that non-financial commercial paper comprises a tiny share of the relevant
market—only 25% of total commercial paper, and less than 9% of the sum of paper, CDs and Treasury bills.21 Understanding how Federal Reserve policy and credit conditions affect the paper-bill
spread will require expanding the model to take into account the behavior of other relevant assets,
such as CDs andfinancialpaper.

21. These figures are for 1991:4. The share of non-financial commercial paper is even smaller
earlier in the sample.




-22-

Kuttner

References
Barro, Robert (1974), "Are Government Bonds Net Wealth?** Journal of Political Economy .82,
pp. 1095-1118.
Bernanke, Ben S. (1986), "Alternative Explanations of the Money-Income Correlation,**
Carnegie Rochester Conference Series on Public Policy 25, pp. 49-100.
Bernanke, Ben S. (1990), "On the Predictive Power of Interest Rates and Interest Rate
Spreads,** New England Economic Review November-December, pp. 51-68.
Bernanke, Ben S. and Cara S. Lown (1991), "The Credit Crunch,** Brookings Papers on Economic Activity 2, pp. 205-39.
Blanchard, Olivier, and Mark Watson (1986), "Are Business Cycles All Alike?** in Robert A.
Gordon, ed., The American Business Cycle: Continuity and Change. Chicago: The University of Chicago Press and the NBER.
Bosworth, Barry and James S. Duesenberry (1973), "A Row of Funds Model and its Implications*' in Issues in Federal Debt Management, Federal Reserve Bank of Boston Conference Series 10, pp. 39-149.
Brainard, William C. (1964), "Financial Intermediaries and a Theory of Monetary Control,**
Yale Economic Essays 4, pp. 431-82.
Foulkc, Roy A. (1931), The Commercial Paper Market New York The Bankers Publishing
Company.
Friedman, Benjamin M. (1991), "Comments on Bernanke and Lown,** Brookings Papers on
Economic Activity 2, pp. 240-44.
Friedman, Benjamin M. and Kenneth N. Kuttner (1992), "Why Does the Paper-Bill Spread
Predict Real Economic Activity?** forthcoming in James H. Stock and Mark W. Watson
eds., New Research in Business Cycles, Indicators and Forecasting, Chicago: University
of Chicago Press and the NBER.
Hurley, Evelyn (1977), "The Commercial Paper Market,** Federal Reserve Bulletin 63, June,
pp. 525-536.
Hurley, Evelyn (1982), "The Commercial Paper Market since the Mid-Seventies** Federal Reserve Bulletin 68, June, pp. 327-333.
Gertler, Mark and Simon Gilchrist (1992), "The Role of Credit Market Imperfections in the
Monetary Transmission Mechanism: Arguments and Evidence,*' Manuscript.
Kashyap, Anil, Jeremy C. Stein, and David Wilcox (1992), "Monetary Policy and Credit Conditions: Evidence from the Composition of External Finance,*' NBER Working Paper
#4015, Cambridge: National Bureau of Economic Research.
King, Stephen R. (1986), "Monetary Transmission: Through Bank Loans or Bank Liabilities?"
Journal of Money, Credit and Banking 18, August, pp. 290-303.
Lawler, Thomas A. (1978), "Seasonal Movements in Short-term Yield Spreads,*' Federal Reserve Bank of Richmond Economic Review, July/August,.




-23-

Kuttner

Oliner, Stephen D. and Glenn D. Rudebusch (1992), "The Transmission of Monetary Policy to
Small and Large Firms," Manuscript.
Owens, Raymond E. and Stacey L. Schreft (1992), "Identifying Credit Crunches," Federal Reserve Bank of Richmond Working Paper #92-1.
Selden, Richard T. (1963), Trends and Cycles in the commercial Paper Market," National Bureau of Economic Research Occasional Paper #85.
Stigum, Marcia (1990), The Money Market Homewood: Dow Jones Irwin.
Strongin, Steven H. (1991), "The Identification of Monetary Policy Disturbances: Explaining
the Liquidity Puzzle," Federal Reserve Bank of Chicago Working Paper #91-24.




-24-

Kuttner

1.

F-Statistics for Alternative Measures of Credit Conditions in
Quarterly Real Output Equations

60:2-91:4

70:3-91:4

75:1-91:4

(1) "Mix" alone

3.36"

2.09*

286"

(2) Loan spread alone

an

0.30

0.46

(3) Paper-bill spread alone

3.81"*

2.71"

1.81

(4) "Mix" + loan spread
"mix" terms
loan spread terms

3.37"
0.23

1.81
0.14

3.07"
0.79

(5) "Mix" + paper-ttll spread
"mix" terms
paper-bill spread terms

4.17"*
4.62'"

2.46
3.07"

4.39"*
3.30"

Specification

*
**
***
Notes:




Significant at the 10% level
Significant at the 5% level
Significant at the 1% level
The regressions are based on qtiarterly data for the sample indicated.
In addition to the variables indicated, each regression includes four lags of real GDP
growth, real non-borrowed reserves growth, the differenced commercial paper rate, plus
constant and trend terms.

-25-

Kuttner

2.

Decomposing Changes in the Composition of External Finance

(a) Regression with separate commercial paper and bank lending terms
Exclusion F-stat
(p-value)
Commercial paper (AA/>)
Bank lending (AAL)

Sum of coefficients
(p-value)

4.00
(0.005)

-0.51
(004)

1.39
(Q24)

-0.90
(0.22)

(b) Regression with the differenced "mix" and commercial paper
Exclusion F-stat
(p-value)
"Mix" (AA)
Commercial paper (iJip)

1.45
(0.22)
264
(0.04)

Sum of coefficients
(p-value)
-0.91
(0.23)
-1.38
(0.04)

(c) Regression with the "mix" in levels, commercial paper, and linear trend
Exclusion F-stat
(p-value)
"Mix" (h)
Commercial paper (A/i/>)

Notes:




1.48
(0-21)
227
(0.07)

Sum of coefficients
(p-value)
0.11
(0.16)
-1.77
(0.01)

The regressions are based on quarterly data for 1960:2 through 1991:4.
The specifications include four lags of each included variable and a constant term.

-26-

Kuttner

3.




Structural VAR Estimates, Credit Conditions Identified via Lending Flows
(equations 2a-2f)

2a.
2b.
2c.
2d.
2e.
2f.

Notes:

R= -0.625 x+ v,
(1.94)
F= 0.159 r+ 1.974
(1.05)
(4.11)
/>=-0.208 R+ 0.037
(6.96)
(210)
L = -0.396 * - 0.022
(1.51)
(ttl6)
P = ttl35 rt + 0.027
(1.29)
(0.51)
x = -0.023 r+ 0.019
(0.23)
(1.50)

x+v*
F+vj
F+ 2125 r, + v4
(3.23)
F + 0.444 r>- 0.094 v4 + v5
(1.68)
(271)
1 + 0.004 P + v6
(ttl4)

Estimates are based on quarterly data for 1960:2 through 1991:4.
Regressions include three lags of each variable, constant and trend terms.
Numbers in parentheses are /-statistics.

-27-

Kuttner
4.

Structural VAR Estimates of the Effects of Lending Shocks on the Paper-Bill Spread
(equations 3a-3g)

3a.
3b.
3c.
3d.
3e.
3f.
3g.

Notes:




/?=-0.558 x+V!
(1.51)
F = 0.048 r+ 1.882
(033)
(3.70)
r, * -0.216 R + 0.027
(7.63)
(1.57)
L = -0.306 R- 0.022
(1.18)
(017)
P = 0.110/?+ 0.038
(1.05)
(071)
x= 0.123 r+ 0.016
(1.20)
(1.35)
r/,-r,=

x+vj
F+ v,
F+ 2051 r, + v4
(3.05)
F+ 0.470 r , - 0.091 v4+
v5
(1.73)
(261)
1+ 0.023 P - 0.897 (rP-rB) + v6
(076)
(3.20)

0.016/?+ 0.181 r+ 0.000 F+ 0.002 1 +
(1.40)
(6.20)
(007)
(046)

0.016 P + v,
(1.71)

Estimates are based on quarteriy data for 1960:2 through 1991:4.
Regressions include three lags of each variable, constant and trend terms.
Numbers in parentheses are /-statistics.

-28-

Kuttner
5.

Structural VAR Estimates, Credit Conditions Identified via the Loan Spread
(equations 4a-2g)

4a.
4b.
4c.
4d.
4e.
4f.
4gNotes:




R= -0.347 x+ v,
(0.85)
F= 0.122 r+ 2132 x +
(a89)
(4.46)
i> = -0.202 rt + 0.016 F+
(aiO)
(0.96)
r t - r P = 0.070/?+ 0.014 F (4.99)
(1.77)
1 = -0.039 rt + a021 F +
(0.14)
(ai5)
P= 0.030 J? + a003 F +

V2

»*

a261
(6.56)
a783
(0.96)
0.722
(218)
(a27)
(ao6)
x= -0.047 r - 0.362 ( n - rP)-f 0.016
(0.39)
(1.54)
(1.28)

/> + v4
r, - 4.727 (rL - »>) + vs
(299)
/> + 1.176 (rL-rP)- 0.085 v5 + v6
(1.83)
(242)
1+ 0.015 P+
v7
(0.46)

Estimates are based on quarterly data for 1960:2 through 1991:4.
Regressions include three lags of each variable, constant and trend terms.
Numbers in parentheses are /-statistics.

-29-

Kuttner

Figure 1

reserves market

Rs Cf«0

l-fp




*"f "*"*

loan market

paper market

\yS

-30-

(I

w

(HHi)

Figure 2
Impulse Response Functions of Credit Conditions Indicators
(a) mix -> output

(b) reserves -> mix
0.0048

(c) mix -> interest rate
0.0032

0.00361
0.00161

0.0024
0.00121

0.0000
T^-r

0.0000
-.0012

3

6

1

9

1—r-

0

(d) loan spread -> output

I

I

3

I

I

I

(e) reserves -> loan spread

0.0020

-.0016

6

0.0036

(0 loan spread -> interest rate

i
CO




-.00201

i—i • i

i

(g) paper-bill spread -> output
0.0016

(h) reserves -> paper-bill spread
0.0007 n

0.0000
-.0016
-.0032
-.0046

.0040 • L -i—i—i—i—i—i—r-

r-

«

i—i

3

i

r

- t •

9

r •

0) paper-bill spread -> interest rate
0.00501

Figure 3: Financing Gap and Financial Flows
bank lending and paper issuance, four-quarter moving average

1

v>

• 2
w

c

•o2 18




-36

.• I ' I ' I ' I ' I ' I ' I ' I • I ' I ' I • I • I • I • I • I • I • I ' I ' I ' I ' I ' I ' I ' I ' I ' I ' I

60

63

66

69

72

75

78

81

84

87

90

Kuttner

Figure 4: credit conditions = lending shocks
response of the Mix

0.16
0.08 H
o.oo
-.08 -j

reserves
lending

•.16 -|

paper
-fin gap—

.24

T

1

1

P

I

I

1

8

9

1

10

11

response of the interest rate
0.035

0.000

.035 H

.070

i

1

7

8

11

reserves
Ien3ing
paper

—
0.050 -

^^"^

— —
.-.-—..

.-v^—-•

0.025 -

0.000 —'

1

i

10

response of real output

n f\7K —i
Kj.UfD

-.025 -

1

9

******

.




,

,

**"*

,

2

.

3

,

4

,

5

6

- 33 -

,

7

,

8

,

9

,

1

10

11

Kixttner

Figure 5: credit conditions = lending shocks
response of the paper-bill spread

0.008

0.000

-.008 H

-.016

l
9

r
10

11

response of real output

0.10

0.05 H

0.00

J~^

\
•.05

1




0

1
1

1

1
2

n
3

4

1
5

1
6
- 3H -

1
7

1
8

1
9

1

1
10

11




Kuttner

Figure 6: credit conditions = loan spread shocks
response of the loan spread
reserves
lending
paper

response of the interest rate

s

s

/
" •••••i^>%

response of real output

- 35 -

COMMENTS ON
CREDIT CONDITIONS AND EXTERNAL FINANCE:
INTERPRETING THE BEHAVIOR OF FINANCIAL FLOWS AND INTEREST RATE SPREADS
David Wilcox

Two opposing views have animated much recent research on the
transmission channels of monetary policy. One view (stated in its
extreme form) is that the impulses of monetary policy are transmitted
to the real economy exclusively via the market for reserves. By
manipulating the quantity of available reserves, the Federal Reserve
is able to change the relative supply of money and bonds. Given this
change in relative supply, the interest rate must change in order to
clear the markets for money and bonds. In turn, the change in the
interest rate alters the user cost of capital, and so influences the
investment decisions of businesses and the spending decisions of
households.
An essential assumption implicit in this so-called "money" view
of the transmission mechanism is that bank loans, market-intermediated
privately-issued debt such as commercial paper and corporate bonds,
and privately-held government debt can be treated as perfect
substitutes. Indeed, this assumption is embedded in the conventional
IS-LM model, where the aggregate non-money financial asset is simply
labelled "bonds" for convenience. According to the money view, the
reduction in bank loans that accompanies a reduction in reserves is of
no particular significance in itself because firms can satisfy any
unmet demand for external finance by issuing market-intermediated debt
which is indistinguishable from bank debt. For this reason, the money
view often is summarized by the proposition that bank loans are not
"special."
The opposing view of the transmission mechanism assigns a
central role to bank loans. According to this view, bank loans,
market-intermediated privately-issued debt, and government debt are
not perfect substitutes. The reduction in the volume of bank loans
that accompanies a move toward a more restrictive monetary policy is

1. David Wilcox is on the staff of the Board of Governors of the
Federal Reserve System.




Wilcox

contractionary in itself, even controlling for any associated change
in interest rates. In effect, bank loans behave as if they were a
factor of production. A reduction in their availability increases
their relative price (the spread between the loan rate and the openmarket rate increases). In response, firms seek cheaper alternatives
for their external finance. However, given the imperfect
substitutability of other forms of debt for bank loans, the reduction
in loan availability implies a contraction in real activity.
The important distinction between the money view and the loans
view is that the latter implies that the impulses of monetary policy
are transmitted not only through the overall level of interest rates,
but also through the relative prices and relative quantities of bank
loans and other forms of external finance. If the loans view is
right, fluctuations in the quantities and prices of bank loans,
commercial paper, other private debt, and government debt will be
worth keeping track of separately because they will be informative for
either the current or future state of the economy, or both. Moreover,
the loans view suggests, as Kuttner (this volume) and Friedman (1991)
emphasize, that there is no reason for being uniquely interested in
changes in the stance of monetary policy; other factors (including but
not restricted to the stringency of regulatory oversight) will also be
worthy of study to the extent that they bear on loan availability.
THE IDENTIFICATION PROBLEM
One approach to investigating the empirical significance of the loans
channel has been to regress some measure of real activity (such as
industrial production or GNP) on current and lagged measures of bank
loans. A positive correlation between bank loans and real activity
has sometimes been interpreted as contradicting the money view and
supporting the existence of a separate loans channel. The flaw in
this argument is not hard to spot: A positive correlation between
bank loans and real activity could simply reflect an endogenous
response
of the demand for bank loans to changes in real activity
rather than an exogenous cause of changes in real activity. Even a
finding of a positive correlation between bank loans and
subsequent
changes in activity (as opposed to contemporaneous ones) would not be
convincing evidence of a separate loans channel; such a phenomenon
could reflect, for example, a need to secure financing some months or




-2-

Wilcox

even quarters before the bulk of the associated activity is to take
place.
An important challange taken up in the more recent literature
has been to solve this identification problem in a convincing
manner.2
SUMMARY OF KUTTNER'S PAPER
Ken Kuttner's paper makes two important contributions to the
literature on the monetary policy transmission mechanism: one
theoretical, the other empirical.
On the theoretical front, he presents a very nice compact model
of the flow of funds in a simple economy. He distinguishes five
financial instruments in his model (in contrast to the usual two):
deposits ("money"), bank loans, commercial paper, reserves, and
government debt.
He posits the existence of a representive firm,
a representative bank, and a representative household, and endows each
of them with standard portfolio behavior (households* demand for money
is declining in the opportunity cost of holding money, and so forth).
Then he derives the implications of changes in the stance of monetary
policy, changes in banks' willingness to lend, and changes in firms'
demand for external finance for three quantities: the mix of external
finance, the spread between the loan rate and the commercial paper
rate, and the spread between the paper' rate and the Treasury bill
rate.
The beauty of Kuttner's model is that it delivers sensible
results very directly. For example, a reduction in banks' willingness
4

to lend causes the loan-paper spread to rise.

In response, firms

2. The approach proposed in Kashyap, Stein, and Wilcox (1992) is to
focus on changes in the composition
of external finance rather than
fluctuations in any one component alone. Intuitively, one would not
expect changes in the volume of bank loans relative to the volume of
other debt to be informative for current or future changes in real
activity if bank debt is a perfect substitute for non-bank debt.
3. Implicitly, other corporate liabilities such as medium- and
long-term bonds are treated as perfect substitutes for commercial
paper.
4. Kuttner interprets "negative shifts in X as 'credit crunch'
episodes." He notes, however, that a negative shift in X could
reflect a "perceived deterioration in borrowers' creditworthiness."
In my opinion, it would be more useful to reserve the term "credit
(Footnote continues on next page)




-3-

Wilcox

shift the mix of external finance away from bank loans and toward
market-mediated debt. The increased issuance of commercial paper
drives up the spread between commercial paper rates and bill
rates.
With respect to these three key variables, the effects of
a reduction in banks' willingness to lend are identical to the effects
of a move by the Federal Reserve toward a more restictive monetary
policy, suggesting that any one of the three might be useful as an
index of loan availability.
In fact, it turns out that these three variables also respond
in qualitatively the same manner to the other two exogenous factors in
Kuttner's model (monetary policy and the demand for external finance).
That is, no matter what the conceptual experiment being run in
Kuttner's model, the loan-paper spread will always move in the same
direction as the paper-bill spread, and the two spreads will always
move in the opposite direction of the mix.
In light of these predictions from his theoretical model,
Kuttner's finding that the loan-paper spread significantly
underperforms the mix and the paper-bill spread as indicators for
future real GNP is interesting and a bit puzzling. Kashyap, Stein,

(Footnote continued from previous page)
crunch" for periods in which some potential borrowers are turned away

even though, with Identical

characteristics

in every

respect

(including
"credit
worthiness"),
they would.have been granted credit
in "normal" times.
5. In Kuttner's model, the commercial paper rate is taken as the
benchmark rate over which the Federal Reserve has direct control in
the reserves market. As a result, a reduction in banks' willingness
to lend has no effect on the JeveJ of the commercial paper rate.
As was noted in the text, however, it does increase the loans-paper
spread. As a result, the volume of commercial paper outstanding
rises and the paper-bill spread increases. Given the fixity of the
paper rate in the face of this experiment, it must be that the bill
rate has declined. If the bill rate (rather than the paper rate) were
assumed to clear the market for reserves, all the essential results
still would hold (the mix would shift away from loans, the loans-paper
spread and the paper-bills spread both would rise), but the bill rate
would be fixed and the paper rate would rise.
6. Kuttner notes that the effects of a shift in monetary policy are
not identical in every respect to the effects of a shift in banks'
willingness to lend: The former affects the level of the interest
rate in the market for reserves, whereas the latter does not.




-4-

Wilcox

and Wilcox (1992) argued that the mix might be preferable to the loanpaper spread as an indicator of loan availability (because the stated
loan rate would not adequately reflect changes in non-price terms of
loan contracts such as collateral requirements), but then proceeded to
find in their sample that the predictive power of the two variables
was roughly comparable. It would be worth attempting to reconcile
Kuttner's results with those of KSW, and (assuming Kuttner's results
hold up) attempting to verify the KSW hypothesis about why the loanpaper spread might be an inferior performer.
On the empirical side, Kuttner's paper introduces a new
approach to solving the identification problem. He posits several
simple "structual vector autoregression" models of the markets for
reserves, bank loans, and commercial paper. Kuttner is bold enough to
supply sufficient prior restrictions on the specification of the
various equations, and finds that, for the most part the estimates
that follow are well in line with the predictions that were outlined
in his theoretical section. The major exception--and one that
deserves further investigation--is that increases in banks'
willingness to lend (counterintuitively) appear to cause Increases
in
interest rates.
AN ASYMMETRIC-INFORMATION-BASED ACCOUNT OF SUBSTITUTION BETWEEN LOANS
AND PAPER
In line with most of its recent predecessors, Kuttner's paper adopts
an aggregate perspective: The model is inhabited by representative
banks, households, and non-bank firms, and the empirical work is
conducted using aggregate data. As in the earlier papers, this
perspective--through no fault of the author--sets up certain tensions
of both an expositional sort and a substantive sort. On the
expositional side, the most natural way to tell the story of the loans
channel involves an appeal to heterogeneity among firms: Some are
capable of issuing commercial paper while others are not. Obviously,
a story such as this is difficult to link up directly to a model with
a single representative non-bank firm. On the substantive side, the
representative-agent approach to modelling the problem fuels the
intuition that some firms should be observed to be on the margin
between bank loans and commercial paper. The purpose of the rest of
these comments is to sketch verbally a model that allows for




-5-

Wilcox

heterogeneity among firms, and then to point out two important
implications of such an approach.
The loans view is predicated on the assertion that non-bank
debt is not perfectly substitutable for bank debt. That imperfect
substitutability can be motivated as reflecting market imperfections
that arise when borrowers have more information about their economic
prospects than do prospective lenders. Banks specialize in
"information-intensive" lending--that is, in lending to customers
(such as small businesses) for whom the asymmetric-information problem
is more acute, and hence more difficult for arms-length capital
markets to solve.
A contractionary shift in the stance of monetary policy will
cause banks to reduce the size of their loan portfolios.
Banks
will tend to cut off their most risky customers and continue to
service their most creditworthy ones. Firms that are denied credit by
banks may be unable to borrow from any other lender. Certainly, they
will not be able to issue debt in arms-length capital markets: nor
will they be able to attract financing from other non-bank sources
simply by announcing their willingness to pay a higher rate of
interest on the debt, because potential lenders will recognize that
only the riskiest firms would be willing to offer a higher rate of
return. In the end, these firms are likely to be particularly
vulnerable to the monetary contraction.
After a monetary contraction, a larger fraction of total
external finance will be provided via arms-length capital markets and
a smaller fraction through bank loans. This change in composition may
reflect either (or both) of two factors: First, it may reflect
increased issuance of trade credit by large, financially secure firms
to their smaller, less creditworthy suppliers. An increase in
commercial paper borrowing would be used, in effect, to finance the
rise in trade credit. Large firms may be willing to act, in effect.
as financial intermediaries because they will have accumulated
substantial inside information about the financial stability of their
suppliers in the course of having interacted with them before the

7. A lower level of reserves will only support a lower level of
deposits. The lower level of deposits (which comprise banks*
liabilities) implies that assets will have to decline as well. Given
that banks view loans and securities as imperfect substitutes, some of
that decline in assets will be absorbed in loans.




-6-

Wilcox

credit crunch. Alternatively, the increase in the share of commercial
paper in total external finance may reflect that large firms tend to
expand when their smaller rivals are weakened by financial stringency;
the large firms take the opportunity to seize some portion of the
product market, financing the larger scale of their operations with
the increase in commercial paper issuance.
These two mechanisms show that bank loans and commercial paper
can be substitutes at the aggregate level even though not so for any
individual firm. Failure to observe firms operating on the margin
between bank loans and other market-mediated debt does not constitute
evidence against the heterogeneous-firms version of the loans channel.
IMPLICATIONS OF THE ASYMMETRIC-INFORMATION-BASED APPROACH
The informal discussion in the previous section points to two
important implications for future research. First, the very
motivation of banks specializing in information-intensive lending
suggests that further progress probably would flow from the analysis
of models that allow for heterogeneous non-bank firms. In particular,
it seems likely that most such models will imply that, when the
Federal Reserve adopts a more restrictive monetary policy, banks will
shrink their loan portfolios by refusing credit to their riskiest
(least financially stable) customers. Commercial paper issuance will
rise because firms already issuing paper will issue more--either to
finance their own expanded operations, or to finance the passthrough
of trade credit to their suppliers.
By contrast, a
representative-firm model suggests that all firms should be on the
margin between bank debt and commercial paper, and that when the
Federal Reserve tightens we should observe a rebalancing of
liabilities taking place at the individual firm level. The
implausibility of this account is obvious, given that fewer than 1300
firms in the United States have commercial paper programs rated by
Moody's.

8. Firms that are growing in size will, at some point, find it
possible to issue commercial paper for the first time. If the
profitability of commercial paper issuance is an inverse function of
bank-loan availability, establishment of commercial paper programs
will tend to be bunched into periods immediately following tightenings
of monetary policy. Historically, of course, the commercial paper
market was not always as well-developed as it is now; as the market
deepened and became more efficient, even firms that had been large and
creditworthy for a long time established new programs.




-7-

Wilcox

The second implication of the disaggregated approach is that
future empirical work should focus on micro-level datasets. Such
investigations will be essential for: (1) establishing the identity of
bank customers who are denied credit in the wake of a tightening by
the Federal Reserve; and (2) establishing the source of the
accompanying increase in commercial paper issuance.







DISCOUNT WINDOW BORROWING AND LIQUIDITY

W. J. Coleman. C. Gilles, and P. Labadie1

Three features seem centra] to understanding the relationship between U.S.
monetary policy and the comovements of open market operations, monetary
aggregates, and interest rates. First, shocks to bank reserves affect interest
rates in ways that axe not tightly linked to the Fisherian fundamentals (expected inflation, marginal rate of substitution, and marginal productivity of
capital). Second, banks often respond to reserve shocks by adjusting their
borrowing at the Federal Reserve's discount window. Third, the Federal Reserve often conducts open market operations to smooth interest rates that
would otherwise react to private-sector demand shocks. In this paper, we
study a stochastic general equilibrium model that incorporates these features
in an effort to understand important empirical regularities involving monetary
aggregates and interest rates.
The empirical regularities we have in mind are those documented in the
vast literature aimed at uncovering a negative correlation between short-term
interest rates and exogenous policy shocks to nominal monetary aggregates, a
relationship often referred to as the liquidity effect. Cagan (1972) and Cagan
and Gandolfi (1969), among many others, have reported finding negative correlations between Ml itself and various short-term interest rates. Subsequent
studies have reported similar correlations with innovations in Ml backed out
using a Choleski decomposition of the residuals in a vector autoregression (for
a variety of orderings). More recently, however, Leeper and Gordon (forthcoming) have made a strong case that these innovations probably do not represent
exogenous monetary policy shocks, as the money supply may be endogenously
1

Board of Governors, Federal Reserve System. We gratefully acknowledge helpful discussions with Jim Clouse and Josh Feinman.




Coleman, Gilles, and Labadie
determined in ways that are not captured by any Choleski decomposition. To
support their claim, they noted that the statistical properties of these innovations are sensitive to the other endogenous variables included in the VAR,
the sample period, and the measure of money selected for analysis. Some researchers, for example Bernanke and Blinder (1990) and Sims (forthcoming),
have responded to such criticism by assuming that innovations to interest rates
reflect policy shocks, to which the supply of money responds endogenously. For
our purpose, however, this strategy does not resolve the central question: if
there exists a liquidity effect, then why are these interest rate innovations not
robustly negatively correlated with monetary aggregates (an observation also
made by Leeper and Gordon)?
Christiano and Eichenbaum (1991) and Strongin (1991) have tried to obtain robust negative correlations by using nonborrowed reserves as the measure
of money. This approach contrasts with that of Leeper and Gordon, who experimented with monetary aggregates that are at least as broad as the monetary
base. Christiano and Eichenbaum's rationale for using nonborrowed reserves
is based on the widely held perception that the Fed controls this aggregate.
For this reason they associated policy shocks with innovations to nonborrowed
reserves, which they then showed to be negatively correlated with the federal
funds rate. In fact, using nonborrowed reserves as the measure of money, they
found evidence of a negative correlation regardless of whether money innovations or interest rates innovations were identified as the policy shocks, and
they showed that these correlations are remarkably robust to the sample time
period. To explain why the innovations to broader monetary aggregates do not
exhibit a similar correlation, they noted that these aggregates are largely endogenously determined by the banking system. For example, they argued that
total reserves may be inelastic in the short run, and therefore not correlated
with interest rates at all. In this example, policy shocks to nonborrowed reserves do not affect total reserves immediately. Strongin refined this argument;
he argued that innovations to nonborrowed reserves that are not reflected in
shocks to total reserves should be identified as the policy shocks. He asserted,
in essence, that shocks to required reserves lead to an adjustment in both
- 2 -




Coleman, Gilles. and Labadie
nonborrowed and total reserves, whereas open market operations lead to an
adjustment in only nonborrowed reserves.
We develop a model that is rich enough to address the empirical issues
presented above. To do this, we introduce a banking system, reserve requirements, and a discount window into a model of liquidity based on the works
of Grossman and Weiss (1983), Rotemberg (1984), Lucas (1990) and Fuerst
(1992). In these models, and here, the term liquidity effect refers not merely
to a negative correlation between monetary policy shocks and interest rates
but more generally to any non-Fisherian effect on interest rates. Interest rates
deviate from their Fisherian fundamentals because of shocks to the demand for
bank deposits from businesses to finance new investment projects and perhaps
also because of monetary policy shocks. In our model, the interest rate is also
the cost (both pecuniary and nonpecuniary) of borrowing reserves from the
discount window, so that over time there is a well defined relationship between
borrowed reserves and the interest rate. Monetary policy designed to smooth
interest rates then leads to rather complicated mutual dependencies among
open market operations, both broad and narrow monetary aggregates, and
interest rates; in particular, monetary policy can lead to positive correlations
between broad monetary aggregates and interest rates in spite of the liquidity
effect. When policy shocks are correctly identified, however, the model suggests that broad monetary aggregates are negatively correlated with interest
rates, showing evidence of the liquidity effect. Furthermore, the model always
generates a negative correlation between nonborrowed reserves and short-term
interest rates, regardless of what the policy shocks are and how they are identified. Such a result is due to the way the discount window is operated. In
light of this model, one interpretation of Christiano-Eichenbaum and Strongin's results is that they identified the discount window policy. Since this
policy implies a negative correlation between nonborrowed reserves and interest rates whether or not the model incorporates a liquidity effect, their results
shed little light on the presence of such an effect.

- 3 -




Coleman. Gilles, and Labadie
THE MODEL
DESCRIPTION.

To get an overview of the model, consider the following accounting of the
assets and liabilities of banks. Their liabilities comprise demand deposits of
firms and households as well as savings deposits of households. Their assets
are made up of reserves and a portfolio of government securities and loans
to firms. Banks are required to hold as reserves a fraction of their demand
deposits;'to avoid a deficiency, they can borrow reserves at the discount window. Borrowed reserves incur pecuniary and nonpecuniary costs. To start
building a model around this balance sheet, think of households as dividing
their deposits between demand deposits, which can be used to buy goods, and
savings deposits, which cannot. Assume that this division is made before the
value of the open maxket operation is known, resulting in a liquidity effect as
described by Lucas (1990) and Fuerst (1992). Also assume, as Fuerst (1992)
did, that firms must finance their purchases of investment goods with demand
deposits, so that these deposits represent intermediated capital, as in Freeman
and Huffman (1991).
To view the model in more detail, consider a representative household
that ranks stochastic consumption and leisure streams {ct,lt}

according to

the utility function

Lt=0 \t=0

/

where /3{ is the date-i realization of the random discount factor; /3*+i is unknown at the beginning of period t but is revealed later during that period.
The household begins period t with money balances Mt in an interest-bearing
savings account. It immediately transfers amount Zt to a checking account
which bears no interest but can be used during the period to finance consumption ct; only one transfer during the period is allowed. The household must
choose Zt before it knows the realization of any of the current shocks, or prices
for that matter. Its purchases of goods are subject to the finance constraint
Ptct < Zt.
-4-




Coleman, Gilles. and Labadie
At the end of the period, Mt — Zt remains in the household's savings account
and Zt — PfCt in its checking account.
The household derives income from several sources. It provides labor to
the firm, working a fraction of time equal to 1 — it at wage rate Wt] it earns
interest at rate r\ on the amount Mt — Zt in its savings account; it collects a
transfer Xt from the government; finally, as owner of both the firm and the
bank, it collects Il( and II*, the period's proceeds from the sale of output net
of all costs and bank profit respectively. The household receives its income,
including income from labor performed during the period, at the beginning of
the next period, when it is directly deposited into the savings account. With
unspent checking account balances being transferred back into the savings
account, the law of motion for Mt is
Mt+i = Zt - Ptct + (Mt - Zt)(l + r{) + Wi(l - It) + Xt+

Ii{ + II*.

The firm, the second agent in the economy, combines, capital and labor
inputs to produce a homogeneous product sold to buyers of consumption and
capital goods. The production function is
Vt =

F(kt,nt,0t),

where yt is the output, kt and nt are the inputs of capital and labor, and 9%
is a technological shock. The firm owns the capital stock kt and hires labor at
rate Wt] it makes wage payments at the beginning of the next period using the
receipts from the sale of output. The firm must also acquire investment goods
it; it purchases these goods from other firms in the goods market but cannot use
its sales receipts for this purpose. Instead, it finances investment by borrowing
Bt from a bank, which charges interest at rate r*. The bank provides this
financing by crediting the amount to the firm's checking account, increasing
the balance from its starting level of zero. The firm's finance constraint is
Bt > Ptit.
At the end of the period, the firm has spent Ptit on investment goods and
deposits its current sales receipts, PtVt, leaving Bt -f Pt(yt — U) in its checking
-5-




Coleman, Gilles, and Labadie
account. At the beginning of the next period, the firm repays its bank loan
and transfers wages into the worker's savings account. The amount left in the
firm's account, Il£, is paid to the firm's owner as dividend:
n / = Ptyt - Wtm - Ptit -

rtBt.

The stock of capital depreciates at the constant rate 6. so that its law of motion
obeys
fct+i = (1 -6)kt

+ it.

The firm makes all its decisions (namely, J3t, it, and rtt) with full knowledge
of the current shocks and prices.
The bank, the third agent in the economy, starts period t with liabilities
equal to Mt (the household's savings account) and holds an equal amount of
vault cash as an offsetting asset (we write "vault cash" for definiteness; Mt
could also be thought of as an account at the central bank). The household
immediately transfers Zt from its savings to its checking account, without
affecting the bank's total liabilities or assets. The bank pays interest r\ on
Mt — Zt, the amount left in the savings account, but pays no interest on
checking deposits.

By lending Bt to the firm, an amount that is credited

to the firm's checking account, the bank increases both its liabilities and its
assets from Mt to Mt -r Bt. To buy government bonds and to honor checks
written to finance purchases of consumption and investment goods, the bank
depletes its holding of vault cash, Mt] but it replenishes this cash position by
the amount of the checks that firms receive for selling their output, checks
that they deposit in their account. The amount of vault cash that the bank
holds at the end of the period counts as reserves. Note that for an individual
competitive bank, the loan of Bt to a firm drains reserves (when the firm
spends the proceeds) just as much as if the bank had spent an equal amount
to purchase government securities; therefore, at the same rate of interest, the
bank is indifferent between the two types of lending. For the banking system
as a whole, however, loans to firms involve no net loss of reserves, but merely
a transfer from the borrower's bank to the bank of the producer of investment
goods.
-6-




Coleman. Gilles, and Labadie
Reserves. VJ, pay no interest and are subject to a reserve requirement, a
fixed fraction p of the amount of checking deposits on the books of the bank
at the end of the period:
(1)

Vt > p x [(Zt - Pta) + (Bt - Ptit -t-

Pm)].

If the bank cannot satisfy the reserve requirement with the amount of vault
cash it has at the end of the period (after checks have cleared), it can borrow
the shortfall from the government at the discount window. Therefore, the
following accounting identity must hold
(2)

Mt r D t = qtGt + Pt(it T*-

yt) -f Vu

where G% is the number of one-period pure discount government bonds the
bank acquires, at a unit cost of qt = 1/(1 + rt), and D% is the amount it borrows at the discount window. Government bonds, private loans, and discount
window borrowing carry the same rate of interest rt. The bank's objective is
to maximize its period profit, which is given by
(3)

n j = Tt(Bt + qtGt - Dt) - r\(Mt -

Zt\

The government, the fourth agent in the economy, sells one-period bonds
in the securities market and redeems them at the beginning of the following
period, operates the discount window, and makes transfers to the household's
bank account. During period i, the government announces the open market
operation Gt and the amount of transfers Xt after the household chooses Zt
but before any other decision by any agent has to be made. All money flowing
between the government and the private sector, as well as within the banking industry, takes the form of fiat money. The bank starts period t with an
amount of fiat money (which it calls vault cash) equal to Mt. Nonborrowed
reserves Vt — Dt is the amount left in vault cash after the purchase of government bonds and check clearing but before borrowing at the discount window;
in equilibrium, Vt — Dt = Mt — qtGt as can be seen from eq. (2).
Let Ht denote the outstanding supply of fiat money at the beginning of
period t (Mt is best thought of as the demand for fiat money, so that in
- 7 -




Coleman. Gilles, and Labadie
equilibrium Ht = Mt). The law of motion for Ht, which can also be thought
of as the government budget constraint, is as follows:
i?t+1 = Ht T Tt{qtGt — Dt) — Xt.
Think of government policy as a rule that generates the values of Gt and
Xt and that also sets the rate of interest at the discount window.

Assume

that the government lends reserves at the discount window according to an
upward-sloping function if> : [0, oo) —• [0, oo) that relates the rate of interest
it charges to the fraction of total reserves that it lends. Banks cannot lend
at the discount window, so that when the equilibrium rate of interest is lower
than the minimum rate at which the government is willing to lend, V>(0), there
is no discount window activity:
rt = i)(Dt/Vt)
r

t < V^O)

whenever Dt > 0;
whenever Dt = 0.

The argument of if) ought to be the amount supplied at the window, which in
equilibrium turns out to be equal to Dt, the amount demanded.

Incorporating

this equilibrium relationship directly simplifies the notation, but keep in mind
that banks take as given all interest rates, including the rate they face at the
discount window (which is equal to the rate on government securities).
When the Federal Reserve lends at the discount window, the borrowing
bank pays the discount rate plus a nonpecuniary cost; at the margin, this
sum must equal the cost of borrowing from other banks, which is the federal
funds rate. The marginal nonpecuniary cost is thus captured by the difference
between the federal funds rate and the discount rate, called the spread. Historically, the policy of the Federal Reserve seems to have been to supply funds
at the discount window at an increasing nonpecuniary cost (spread), which is
precisely what the function tp assumes. This type of discount-window policy
has been documented in the empirical literature, and is commonly modeled in
the theoretical literature. 2 Chart 1, which graphs the monthly time series for
2

See for example Polakoff(1960), Goldfeld and Kane (1966), and more recently Goodfriend (1983), Dutkowsky (1984), and Waller (1990). In particular, Fig. 1, p. 346 in Goodfriend depicts an assumed ip function that is strikingly similar to the function that would
best fit the scatter plot of our Chart 2.
-8-




Coleman, Gilles. and Labadie
the federal funds rate and the nonborrowed reserve ratio (the mirror image of
the borrowed reserve ratio), reveals the basis for the findings of the empirical
studies. On closer inspection, a picture of the function ib emerges in a scatter
plot of the borrowed reserve ratio against the spread, shown in Chart 2. Since
this picture suggests that the Federal Reserve is ready to lend its first dollar
at a zero spread, the value of t^(0) corresponds to the discount rate. With this
interpretation of ^(0), the model simply assumes a constant discount rate.
A word about terminology is in order. Vt is total reserves in the banking
system; Dt is borrowed reserves; the difference Vt — Dt is nonborrowed reserves;
and required reserves is p x [Zt + Bt + Pt(yt — it — ct)]. Besides total reserves,
it is possible to identify the analogues of several monetary aggregates. M% (or
Ht) corresponds to the monetary base, MO; the analogue of Ml is the sum of
all reservable accounts, Zt + B%\ the total libilities of the banking sector at
the end of the period, Mt + B^ correspond to M2 (strictly speaking, Ml and
M2 both should include Pt{yt — ct — U) as well, but this is equal to zero in
equilibrium); finally, the difference between M2 and MO, which is Bt, is inside
money.
It is now useful to summarize the timing of information and decisions. During period i, the realizations of four random variables shock the economy—the
technological shock 0t> the preference shock /3t+i, the open market operation
Gt, and the government transfer Xt.

At the beginning of the period, the

household must decide how much to put into its checking account, not knowing the current realization of 0t, /3t+i, Gt, or Xt, and therefore not knowing
what interest rates, prices, output, or consumption will be. After it makes
this decision, all four shocks are revealed and prices are set. On the basis of
these shocks and these prices, the household decides how much to consume
and how much to work; the firm decides how much to borrow, how much to
invest, and how much labor to hire; and the bank decides how much to lend
to the firm and to the government. Then trading takes place and checks clear.
The bank monitors its reserve position and borrows at the discount window
to cover any reserve deficiency (the bank can be thought of as borrowing at
the same time it invests in government bonds or lends to firms, because it
- 9 -




Coleman, Gilles. and Labadie
has the same information when it engages in any of these activities). At the
start of next period, the firm pays its wage bill, repays its bank loan, and
pays out its earnings to its shareholder; the government makes transfers to the
household's savings account and redeems the bonds that the bank holds; the
bank pays interest on its savings account, settles its discount window debt,
and pays out its earnings. These activities determine the new initial balance
in the household's savings account. Then a new cycle starts.
The activities of the four agents that have been described above must, of
course, satisfy the following standard market-clearing conditions.
yt

=

a + it

goods market;

nt

=

1 — it

labor market;

Ht

=

Mt

money market.

The economy is competitive, and agents have rational expectations. An
equilibrium is a set of state-contingent prices and interest rates such that
markets clear when all agents solve their optimization problems, treating prices
as given. In the next subsection, we are more explicit about what this means.
THE MODEL AS A RECURSIVE SYSTEM

The household solves a dynamic program, which is recursive under standard
assumptions about preferences, technology, and the stochastic environment.
ASSUMPTION 1.
tiate,

The period utility function U is twice continuously

strictly increasing in both arguments, and strictly

ASSUMPTION 2.

concave.

The production function F has the form F(k, n, 0) = 9f{k, n),

where f is twice continuously
guments,

differen-

differentiate,

concave, and homogeneous

strictly increasing in both ar-

of degree one.

(Stochastic

constant

returns to scale.)
ASSUMPTION

3.

The preference shocks {/3f} and the technological shocks {6t}

are generated by independent

first-order Markov processes.

The support of

&t is contained in (0,1) and that of &t is contained in (0, oo).
Monetary policy consists of a rule that dictates the value of open market
operations, the size of government transfers, and the level of the discount rate;
-10-




Coleman. Gilles. and Labadie
these instruments are not completely independent of each other. The operation
of the discount window is modeled through a fixed function w that relates the
discount rate to borrowed reserves. Think of the government as announcing
this function and keeping it fixed in all periods, leaving the discount rate itself
endogenousiy determined by the demand for borrowed reserves. Given the
function V>, the values of Gt and Xt in period t are implied by the choices of the
ratios gt = Gt/Ht and 7* = iift+i/^t- To induce stationarity and recursivity,
choose (gujt)
ASSUMPTION

4.

order Markov

as the policy variables and make the following assumption.
The monetary policy shocks { # , 7*} are generated by a firstprocess.

Starting with the optimization problem faced by the bank simplifies both
the notation and the analysis. The bank maximizes its period profit, given
in (3), by choosing an optimal portfolio (Sf,Gt,i?t, Vi), subject to the legal
reserve constraint (1), and the accounting identity (2). Clearly, optimization
requires that V% = p[Zt + Bt + Pt(yt — it — ct)] (no excess reserves) if r* > 0. A
zero-profit condition, the result of perfect competition and constant returns to
scale in the banking industry, implies that r\ = [{Mt + Bt — Vt)/(Mt — Zt)] x rt;
this condition in turn yields r\ = r t [l + (l — p)(Zt + Bt)/(Mt — Zt% which holds
whether or not r* > 0. To obtain the last expression, recall the market-clearing
condition yt = ct-r itSince the firm and the bank belong to the household, it is possible to
integrate the problems faced by the firm, the bank, and the household. Because money growth induces a trend in nominal variables, stationarity of the
equilibrium requires that nominal variables—denoted by uppercase letters—
be divided by the supply of fiat money. The new variables are denoted by
the corresponding lowercase letters; thus, nit = Mt/Ht,

zt = Zt/Ht,

and so

forth. Under assumptions 3 and 4, the evolution of the shocks is determined at
the beginning of period t by the vector (/3t,0t-i> 5t-ij7t-i)> which consists of
the latest known realizations of the shocks. The state of the economy at that
time can then be expressed as st = («t?/3t, 0t-i?5t-i)7i-i)> where Kt is the
aggregate per capita stock of capital (as opposed to fct, which is the individual firm's holding). In equilibrium, of course, individual decisions determine
- 11-




Coleman. Gilles, and Labadie
aggregate outcomes, so that K% = kf. A solution is a set of functions p, w.
and r such that pt = p(st,st+i), wt = tu(st,.st+i), and rt = r(s t ,«st+i) yield
the equilibrium values of the normalized price level, the normalized wage rate,
and the rate of interest on date t (again, pt = Pt/Ht and wt = Wt/Ht).

Since

qt = 1/(1 +Tt)} the equilibrium function r determines a function q satisfying
9t = g ( j t i * t + i ) .

Given such pricing functions, let J(m, &, s) denote the value of the optimal
discounted stream of utility for a household starting a given period with money
balances m, while the firm owns capital stock k and the economy is in state
s = («,/?, 0,(/, 7). The household first chooses z, which is the transfer from
its savings to its checking account, expressed as a fraction of the outstanding
supply of fiat money. Then (/9^0^if^7 , ) are revealed (a prime denotes the
realization of a variable that was unknown at the beginning of the period),
and these shocks determine the current price, wage rate, and rate of interest,
as well as the next-period state s'.

To determine s1, the household must

know how the evolution of the aggregate capital stock depends on the state
of the economy. In equilibrium, of course, this law of motion follows from the
individual optimal decisions. On the basis of an assumed law of motion for
K and of p(s,s ; ), w(s1s')}

and r(s,s'),

the*household makes its consumption

and leisure decisions and the firm makes its labor and investment decisions.
What these optimal decisions are can be studied by considering the Bellman
equation characterizing J, the value function.

J(m,k,s)

= max-Ej

max {C/(c,£)' + / 3 J ( m U V ) }

subject to

(4)

2>P<:;
n* = pO1 f(k1 n) — (1 -J- r)pi — wn;
Jfe' = ( l - * ) J b + i;
w(l - I) + (m - z)(l + rb) -f x1 + 7Tf -»- (z - pc)

,
m

=

-12-




Coleman. Gilles, and Labadie
the last constraint on the problem is the law of motion for K. Here p, w. and r
are short for p($, s'), w(s, s1), and r(s, s'), and E9 is the expectation conditional
on 5. Using the results of the bank's optimization problem, the market-ciearing
condition 6' f{k,n)

= c + i, and the firm's optimization condition b = pi. we

have r > 0, v > p(z + 6), and v = p(z + b) if r > 0.
OPTIMIZATION AND EQUILIBRIUM CONDITIONS.

The Bellman equation for J includes two maximization operators; the first
refers to the choice of z, which is conditional only on s, and the second refers
to the choice of (c, £, n, i) which is conditional on both s and sf. Corresponding
to the latter choice, we have the following four first-order conditions:

(<0
u>(j,5')

u>(a,s') = p ( * , a ' y / 2 ( * , n ) ;

(»)

(0

y

J 2 ( m \ Jb',,') = p(*,,')[! + r(s,

M')]Jl{m''?'J);

where A is the Kuhn-Tucker multiplier associated with the finance constraint
(4), so that A(z — pc) = 0. Indexes to the functions U and J denote partial
derivatives; therefore, U\, for example, is the partial derivative of U with
respect to its first argument, consumption.
The first-order condition associated with the choice of z is
(*)

E.

Ui(c,£)
3

S

IP( > ')\

= E. 0[l + rh(s,*')]

Ji(m',k',s')

r

To solve the dynamic programming problem, we need the following envelope
conditions, which give the marginal values of money and capital:

(m) / l ( m , 4 , ( ) . & [ ^ ] .
(*)

J 8 (m, * , . ) = E. [(U,(c,t)

- pX) ( « 7 i ( * , n ) + (1 + r)(l - * ) ) ] ;

where p is short for p(s, s1), and similarly for w and r.
-13-




Coleman, Gilles, and Labadie
Finally, an equilibrium in this economy is a set of functions

w(s,s!),

p(s,$'), and r(s, s1) [or equivalently 9(3, s1)] and a law of motion for the aggregate capital stock K such that the associated solution of the dynamic programming problem—that is, values for (z, A, c, /, n, 2, i/, d) that solve the first-order
and envelope conditions—satisfies the following equilibrium conditions:

c + i = tC/(i, n);
l - * = n;
qg + v - (f = m;
rn = 1;
fc' = « ' ;
r6 =

m —2

xr;

d x r = (fx ip(d/v).
The last equation states that, when the monetary authorities lend at the discount window (d > 0), they do so in accordance with their supply behavior,
so that r = ^{d/v).

In the third equilibrium condition, qg1 + v — d = m, v is

equal to p(z + pi) unless r = 0, in which case v can exceed required reserves.
SOLVING THE MODEL

Consider initially a slightly simplified version of the model in which labor is
inelastically supplied (I = 0) and money supply'is constant (7 = 1). To solve
this simplified model, first reduce the system of equations that determines the
equilibrium to only three equations in the three unknown functions c, z, and
(a transformation of) J\.
To simplify the notation, define £(/9,.s') = /3Ji(l,/c',y). 3 Then the firstorder condition (c) becomes

Ui(c) = (\ + t)PRecall that K is one of the arguments of 5, so that the function £ is well defined.
-14-




Coleman, Gilles. and Labadie
Here and below £ stands for f(/3,s'); accordingly £' below stands for £(/?', s").
Using this equation and the constraint z > pc, which holds with equality
whenever A > 0, isolate p as
(5)

p = nun

^

{

}

Substitute this equation in £ = (3E5i [U^c^/p1], which follows from the definition of £ and the envelope condition (m), to obtain

( = 0E, max

(6)

{***.<}

this equation is the first of the set of three to be solved (£ now replaces J\).
The second equation follows from substituting the expression (5) for p into the
first-order condition (z), obtaining

(7)

£.|ma*{^,*}]=2<;.l(l + r )t}.
h

The last equation in the system follows from the first-order equation (i) and
the envelope condition (Jk):

mm

{**«}
l£l

(8)

z'e, Ux{c')
\ {9"h{k') + (1 + r')(l - 6))
= 0qEs nun < —
<O}CJ

To write (6) - (8) solely in terms of c, z, and £, express r and r in terms of
these functions as follows:

T = i>(dlv);
and
(9)

r> =

where
-15-




Coleman, Gilles, and Labadie
d = qg ~ v - 1;

v = p(z + 6);
6 =

m

| _ _ _ |

m

;

and finally,

i = 0'/(Jfe)-c.
These equations hold provided d > 0 and r > 0; if d = 0, then r < ^(0), while
if r = 0, then v > p{z + 6). Rather than solving this model explicitly, which
can be done numerically using the methodology presented by Coleman (1992),
we devise an example which admits a closed-form solution.

This example

highlights all the features of the model that are useful in interpreting the
empirical regularities mentioned earlier.
AN EXAMPLE

To develop an intuitive understanding of the model, it is instructive to consider
a parametrization that allows a closed-form solution. Suppose that (a) utility
is logarithmic; (b) production satisfies f(k) = fca, for 0 < a < 1; (c) capital
depreciates completely over each period; and (d) the technological shocks 5,
the policy shocks g, and the preference shocks 0 are all iid (although not
necessarily independent of each other). Now, conjecture that no excess cash
is ever held in the goods market and that z is constant at z.
;

circumstances, 6 = zi/c, i = fc , and equations (6)-(8) simplify to

z
(10)

(11)

1 = E, 'P(l+rb)

" = 0'qEsl

)

'6"a(k')a-1'
J

c
where the interest rate r satisfies

<*>

-*r»''
—1 6 -

Under these




Coleman. Gilles, and Labadie
and rb is given by (9). Further conjecture that the consumption function can
be written as
Q =

TTzrt—r$ k ,

1 + Wq)
for some function h. Note that because k!/c = h(/3'g), the function h can be
thought of as the investment to consumption ratio. Since h depends only on
flq and since q = 1/(1 + r), (12) determines r a s a function of 5, /3', and g1.
Write this function, which implies that r and q are iid and independent of s,
as r = RJ^z^ff^g1) and correspondingly q = Q{z^P\g9)\

now substitute these

equations into (9), and the resulting equation into (10), to obtain

! = £ , tt[i

+

[i^->n*««*<i'-™t)m!,M)

This equation has the important implication that z does not depend on 5,
because s enters only through the conditional expectation, and /?' and g1 are
iid. This observation verifies the conjecture z(a) = z. Tofindfc,substitute the
conjecture about the consumption function into (11) and simplify to obtain
h(l3,q) = a(3'q(l + Esl[h(/3"q')}).
Using the fact that 0' and q are iid (because q — Q(z,0',g'),

and (@,g) is iid),

this equation implies
H{l3q)

-l-Ela0<qy

where E[. ] is the unconditional expectation, taken over the constant distribution of (y9',g). It is then straightforward to verify that the finance constraint
in the goods market is always binding; therefore, all the initial conjectures
were correct.
This example leads to a sharp characterization of the response of monetary
aggregates and the interest rate to supply and demand shocks.
equilibrium value of k'/c =fc,rewrite (12) as

(13)

U^^f^Sl-^d-ElaP'q))
q

^\

pz(l+a0'q-E[a(3'q})
-17-

Using the




Coleman, Gilles, and Labadie
Consider first the effect of technological shocks, &. .Such shocks do not
affect r, as (13) makes clear, and thus they do not affect any of the monetary aggregates. They have real effects, of course, since they affect output,
consumption, and investment. But they fail to move nominal interest rates
(although real rates certainly do) because the demand for consumption and
investment goods shift proportionately. This feature is due to the choice of
utility and production functions, and is not a general feature of the model.
It indicates, however, that in the general case productivity shocks can affect
interest rates and monetary aggregates in either direction. Before turning to
the effect of other shocks, it is helpful to list the relevant equations. The first
is (13), which determines the correlation between each shock and the nominal
rate of interest. The others are:
(14)

total reserves:

v = pz[l + h(/3'q)]]

(15)

nonborrowed reserves:

(16)

borrowed reserves:

(17)

Ml:

z + b=z[l

(18)

M2:

1 + 6 = 1 + 2&(0'g).

t; — d = 1 — qg1]
d = v x ^-1(r);
+ h(0'q)]]

To isolate the effect of policy shocks, assume first that there are no other
shocks (a similar procedure will uncover the effect of preference shocks). Note
that the left side of (13) is decreasing in g, while the right side is increasing both
in q and in g1 (recall that T/J is increasing); therefore g' and q vary inversely. For
the same reason, but considering the right side as a function of q and qg\ q and
qgf vary inversely also. Hence, gf, r, and qg1 all move in the same direction.
In view of (15), then, policy shocks induce a negative correlation between the
nominal rate of interest r and nonborrowed reserves v — d. They also induce
a negative correlation between r and v, total reserves, as (14) reveals since h
increases in q. The correlation between r and v can be entirely attributed to
the variance of inside money, z/i(/3'g); this variance also induces a negative
correlation between r and the broader monetary aggregates Ml and M2, as
shown by (17) and (18). From (16), it is clear that the ratio of borrowed to
total reserves is positively correlated with the interest rate, a relation which
-18-




Coleman, Gilles, and Labadie
has nothing to do with the source of the shock but is due exclusively to the
form of ^, that is, to the operation of the discount window. If total reserves
did not respond to the policy shock (an assumption which is sometimes made
in empirical work), the form of ifr alone would induce a positive correlation
between the interest rate and borrowed reserves.
Suppose now that shocks to /3 are the only shocks in the system. The
left side of (13) is decreasing in g, while the right side is increasing in q and
decreasing in /3'g; therefore, q and /3'q (and therefore q and /3' also) move in
opposite directions, while {31 and /3'q move in the same direction. Equations
(14)—(18) then show that preference shocks induce a positive correlation between the interest rate and any of the reserve or monetary aggregates (total,
nonborrowed, and borrowed reserves; inside money, Ml, and M2).
It is now possible to use the example to study more complicated policies.
Suppose that in response to positive preference shocks that would otherwise
increase interest rates, the government chooses its open market operation to
keep the rate constant, which corresponds to a small realization of g1 (in this
case, /3 and g are still iid, but not independent of each other). With the interest
rate constant, /?' high and gf low, all the reserve and monetary aggregates are
high (but the borrowed reserve ratio is constant). If the policy response only
partially offsets the preference shock, all reserve and monetary aggregates may
still rise, while the rate of interest rises also. In that case, despite the presence
of a liquidity effect in the model, open market operations could be seen as
"inducing" a positive correlation between interest rates and various monetary
aggregates (and nonborrowed reserves as well).
CONCLUSION: INTERPRETING THE EMPIRICAL LITERATURE

As mentioned in the introduction, the empirical literature directed to measuring the effect of monetary policy shocks on interest rates is replete with
seemingly conflicting results. The model provides a framework for thinking
about these results and for interpreting the literature; the example brings out
the important features of the model. First, the model highlights the role of inside money creation as an avenue for total reserves to respond to open market
-19-




Coleman. Gilles, and Labadie
operations. In this sense, the model fails to support Strongin's identifying restrictions that total reserves do not respond to open market operations within
a month or a quarter. Second, the model suggests that the operation of the
discount window, summarized by a fixed and positively sloped supply function,
can alone generate a negative correlation between nonborrowed reserves and
the federal funds rate. Such a correlation has been documented by Christiano
and Eichenbaum (1991). While they identified policy shocks as innovations
to nonborrowed reserves, the model suggests an alternative explanation that
has nothing to do with policy shocks. Third, although the model is designed
to have a liquidity effect, a policy of interest-rate smoothing hinders efforts to
detect its presence. This could explain the difficulties econometricians have
had in measuring this effect. To identify policy shocks, it is not sufficient to
identify a variable (such as nonborrowed reserves) that is under the control of
the Fed, since the Fed may use its instrument to achieve particular objectives.
In this sense, the model points to the familiar need, and provides a framework
for, identifying demand and supply shocks to estimate a liquidity effect.

-20-




Coleman, Gilles, and Labadie
REFERENCES

Bernanke. B., and A. Blinder. "The Federal Funds Rate and the Channels of
Monetary Transmission," Working Paper No. 3487. New York: National
Bureau of Economic Research, October 1990.
Cagan, P. The Channels of Monetary Effects on Interest Rates. New York:
National Bureau of Economic Research, 1972.
Cagan, P., and A. Gandolfi. "The Lag in Monetary Policy as Implied by the
Time Pattern of Monetary effects on Interest Rates," American Economic Review, vol. 59 (Papers and Proceedings, 1969), 277-84.
Christiano, L. J., and M. Eichenbaum. "Identification and the Liquidity effect
of a Monetary Policy Shock." Unpublished manuscript, Federal Reserve
Bank of Minneapolis, November 1991.
Coleman, W. J. "Solving Nonlinear Dynamic Models on Parallel Computers,"
Institute for Empirical Economics working paper. Federal Reserve Bank
of Minneapolis, 1992.
Dutkowsky, D. "The Demand for Borrowed Reserves: A Switching Regression
Model," Journal of Finance, vol. 39 (1984), 407-24.
Freeman, S., and G. W. Huffman. "Inside Money, Output, and Causality,"
International Economic Review, vol. 32 (1991), 645-67.
Fuerst, T. S. "Liquidity, Loanable Funds, and Real Activity," Journal of Monetary Economics, vol. 29 (1992), 3-24.
Goldfeld, S. M., and E. J. Kane. "The Determinants of Member-Bank Borrowing: An Econometric Study," vol. 21 (1966), 499-514.
Goodfriend, M. "Discount Window Borrowing, Monetary Policy, and the PostOctober 6, 1979 Federal Reserve Operating Procedure," Journal of
Monetary Economics, vol. 12 (1983), 343-56.
Grossman, S. J., and L. Weiss. "A Transaction-based Model of the Monetary Transmission Mechanism, " American Economic Review, vol. 73
(1983), 871-80.
King, R., and C. Plosser. "Money, Credit, and Prices in a Real Business Cycle,"
American Economic Review, vol. 74 (1984), 363-80.
-21-




Coleman. Gilles, and Labadie
Leeper, E. M., and D. B. Gordon. "In Search of the Liquidity Effect," Journal
of Monetary Economics (forthcoming).
Lucas, R. E., Jr. "Liquidity and Interest Rates," Journal of Economic Theory,
vol. 50 (1990), 237-64.
Polakoff, M. E. "Reluctance Elasticity, Least Cost, and Member-Bank Borrowing: A Suggested Integration," Journal of Finance, vol. 15 (1960),
1-18.
Rotemberg, J. J. "A Monetary Equilibrium Model with Transaction Costs,"
Journal of Political Economy, vol. 92 (1984), 40-58.
Sims, C. A. "Interpreting the Macroeconomic Time Series Facts: The Effects
of Monetary Policy," European Economic Review (forthcoming).
Strongin, S. "The Identification of Monetary Disturbances: Explaining the
Liquidity Puzzle." Unpublished manuscript, Federal Reserve Bank of
Chicago, December 1991.
Waller, C. J. "Administering the Window: A Game-Theoretic Model of DiscountWindow Borrowing," Journal of Monetary Economics, vol. 25 (1990),
273-87.

-22-

Chart 1. Federal Funds Rate and Nonborrowed Reserves Ratio
Monthly, January 1961 - July 1992

percent

1960




1965

1970

1975

1980

1985

1990

i

Chart 2. The Psi Function; 1961 (1)-1992(7).
Ratio of Borrowed To Total Reserves
U.1U

•

0.08

—

•

0.06

•

•

•

•

•

•

•

•

0.04

*•

• • • :
• • •
• •
•

• § • •

•

•

••

•

•

•

•

• • •••

•

••

•
•

•

•
/

••

•

• •
•• •
•
•-

• •
•

•

1

•
•

•

1

•
•

•

•

0.02

•
%

0.00

hi

•

.

:

*• .wA * * i * '„•.••••
1

•• ••
••

• .
l_

-2




finrftarl fFed Funds - Discount Ratol

L_

l_J




Comments on "Discount Window Borrowing and Liquidity"
by Coleman, Gilles, and Labadie
Michael Dotsey
I have been asked to discuss "Discount Window Borrowing and
Liquidity" which I view as very interesting but preliminary work
toward examining "liquidity effects" in a framework that
incorporates a fairly (primitive) reserves market. I use the term
primitive with regard to the reserves market since no interesting
dynamic behavior is present in this market. Viewing work on BRd,
especially that of Goodfriend (1983) this is a shortcoming that I
hope will be addressed by later generations of the model. The
paper, however, is very rigorous and state of the art on other
dimensions and the authors deserve a lot of credit for moving the
liquidity effects literature in this direction.
The empirical motivation for the paper can be traced to
work by Christiano and Eichenbaum (1992) and especially to that of
Strongin (1991). Strongin's work is fairly persuasive and
indicates that in order for any model to replicate data on
liquidity type effects reserve market behavior is likely to be a
crucial ingredient. This is because the liquidity effect only
shows up in NBR's or to be more accurate, in the part of NBR that
represents independent monetary policy. This paper's novel
inclusion of reserve market behavior represents a commendable
extension of this basic line of research.1
In reading this paper, I found that it raised at least as
many questions as it answered. Much of my confusion is not the

1. One thing I would like to see done in these estimations is
removing settlement day data. This data could potentially contaminate
the results. Suppose for instance the Fed misforecasts float or
treasury balances believing there will be more of these funds available
than are actually there. NBR will be low on the settlement day and the
funds rate will be high, perhaps by a substantial amount. Two such
occurrences in a month (at least 25% probability) could make monthly
average NBR a little low and monthly average rF a little high. While I
doubt this is the reason for Strongin's results it would be nice to
purge the data of what is merely an interbank friction.




Michael Dotsey
result nor the fault of this paper in particular, but rather comes
from a lack of understanding and perhaps misgivings of this
literature in general.

In my comments I will discuss some of

these misgivings and, hopefully, my comments will lead to some
discussion from the rest of the audience.
The paper extends a branch of research that is attempting
to understand the effect of monetary policy on interest rates and
real activity.

In particular these papers7 search for a mechanism

that will explain (1) how contractionary monetary policy raises
short-term interest rates and (2) how it causes declines in
economic activity.

This literature received its impetus from

Lucas's (1990) influential paper.

A common feature of most of

this literature involves cash-in-advance constraints that
constrain the amount of money available for use in a loan or
securities market, however, no two papers seem to use the same
exact specification.
Lucas's original setup and CGL (1991) envision bond
traders as only having limited funds and, therefore, open market
operations affect the price of bonds .and thus interest rates.

The

appeal of Lucas's setup is that it eliminates the differential
wealth effect of open market operations that were present in
earlier literature (eg Grossman and Weiss and Rotemberg).

Fuerst

(1991) extends Lucas's setup to a production economy that places a
CIA constraint on both investment and labor expenditures.

Unlike

households' portfolio decisions, production decisions are made
after the stochastic state of the economy is known.

Since

individuals must choose the portion of their portfolio to lend to
firms via intermediaries prior to observing the monetary transfer
or the market clearing interest rate, the monetary transfer can
affect the tightness or looseness of the loan market.

Hence

liquidity effects that have real consequences result from monetary
policy.

Christiano (1991) subjects the Fuerst model and an

alternative version of that model in which investment decisions

- 2 -




Michael Dotsey
are also made prior to the realization of shocks to a statistical
comparison with a RBC model that contains a standard CIA
constraint.

For reasonable parameter specifications the Fuerst •

model can not produce a liquidity effect that dominates
anticipated inflation effects on the nominal interest rate while
the sluggish capital model can produce a dominant liquidity
effect.

Both these models produce too much variability in

consumption and the counterfactual result that consumption and
prices move in opposite directions.

They also produce very low

interest elasticities of money demand and monetary policy has very
little effect on variations in output.

Furthermore, anticipated

inflation has much too large an effect on labor, consumption, and
output.

To remedy this last result, Christiano and Eichenbaun

(1992) relax the CIA constraint on investment.

They also split

the period into two parts allowing firms to adjust their hiring
decision after observing open market operations while initial
hiring and investment decisions are made prior to observing open
market operations.

They do this with the hope of magnifying the

response of employment and output to liquidity effects.

In CGL's

current paper firms face a CIA constraint on investment but can
pay workers out of end of period revenues.

Also, monetary

transfers are made directly to consumers after their portfolio
decision has been made.

Thus these transfers do not affect the

funds available in the credit market and, therefore, do not give
rise to a "liquidity effect."

Because there is a CIA constraint

on capital, monetary policy can have inflation tax effects as
well.

As their work progresses separating liquidity effects from

inflation tax effects will be important.
Not all of these scenarios can be correct.
constraints placed where they are?

These assumptions of infinite

transactions costs are not innocuous.
in these models.

Why are CIA

They are the driving force

It seems that rather than trying to incorporate

a realistic financial structure into a dynamic macro model and

- 3 -




Michael Dotsey
then testing the model, investigators are trying to find a
mathematical structure that produces the correlations they desire.
Apart from Christiano (1991) very little effort is made to see if
these models are an improvement on basic RBC models or even if
they produce counterfactural predictions along other dimensions.
Since other classes of models can produce negative correlations
between NBR and the funds rate, examining how CIA models fit the
data along other dimensions will be important if the CIA approach
is to gain widespread acceptance.
For example a model like that in Goodfriend's (1987) paper
can potentially produce correlations of the type this literature
is seeking. In that model, which has no rigidities, purposeful
behavior by the Fed can set up negative correlations between the
funds rate and NBR. If the Fed wishes to reduce inflation, it can
do so by reducing the future money supply and in particular future
NBR. Due to anticipated inflation effects, the nominal interest
rate would fall increasing the demand for money and total
reserves. If the Fed wishes to reduce price level surprises it
can supply the necessary NBR to prevent price level movements.
Thus this policy sets up the requisite negative correlation. If
that was all that was going on one would expect this negative
correlation to carry over to broader aggregates. However, M2-M1
components of M2 which involve a large savings motive should be
positively correlated with the real rate of interest and movements
in BR, which are highly variable and positively correlated with
the funds rate, could cause TR to be positively correlated on net
as well.
Alternatively say the Fed is following an exogenous upward
movement in the real rate of interest in an attempt to target
inflation. If the own rate on money balances is sticky then money
(Ml) and hence total reserves will decline along their demand
curve. (Also, M2 could be rising with the real rates.) This
would set up a negative relationship between NBR and the funds

- 4 -




Michael Dotsey
rate.

As rm adjusted, total reserve demand would increase as

would NBR as the Fed defended the new higher funds rate.

If the

Fed did not react instantaneously or vigorously enough to the
increased reserve

demand the funds rate could rise further and

then fall as nonborrowed reserves were pumped into the system
reinforcing the initial negative correlation.

Also, sticky price

models may be able to generate some of the correlations displayed
in the data as well.
Also, the question of what constitutes a period is
somewhat fuzzy in this literature.

Is it a day or perhaps a week?

Most people make some form of cash management decision weekly and
I can not think of any time where a shortage of cash has affected
my real consumption for more than a day or two.

Perhaps I'm

taking the CIA constraint too literally, but if the period is
rather short, as I believe it is, then the propagation mechanisms
needed to match the data would seem incredible by RBC model
standards.
I have strayed a little far afield so let me return to
this paper more specifically.

My primary confusion is linking the

author's major contribution which shows how different measures of
money can have different correlations with interest rates with the
motivation for their paper which appears to be the results found
in Strongin.

In this paper money

(1)

M t ^ - Mt + rt(Gt-Dt) + xt.

The xt portion of measured money provides no liquidity effects.
The 6t portion, that is open market operations has the standard
liquidity effects since it influences the portion of firm
borrowing that must be financed by discount window loans.

The

equilibrium condition that is being used is
(2)

NBRt = Vt - Dt =

Mt - Gt

where Vt * 0(M t +B t ).

An increase in 6t (an open market sale)

requires more discount window borrowing and an increase in
interest rates since r =0(D/TR) is increasing.

- 5 -

Using Mt+1 can




Michael Dotsey
contaminate regression results since it rises by r t (G t -D t ), which
will in general be positive in this model and no liquidity effect
will be present.

Furthermore, growth in money via transfers wiVI

further bias econometric results.
For econometric purposes I see no useful way of isolating
any aggregate to uncover liquidity effects-

Xt type disturbances,

in reality, involve transfers from the Treasury.

These involve a

reduction in Treasury accounts at the Fed and an increase in NBR.
What the model here indicates is that one wants to examine only
changes in reserves

that involve changes in the public's asset

positions and that exclude any interest or lump sum payments.
While these decompositional problems are important for
this model and may in fact be important more generally, they seem
to have little to do with Strongin's empirical strategy nor do
they affect interpretations in other models.

Strongin tries to

separate "pure" supply movements in NBR from those engendered by
policy responses to changes in TR. Whether his identification
procedure is a good one or not could be debated, but he is not
concerned with measurement or decompositional problems in various
reserve measures.
The decompositional problem arises in CGL because of their
modeling of xt as having no liquidity effects.

In Fuerst or

Christiano and Eichenbaun, there is only xt and it enters the
model in a way that produces liquidity effects.

That is NBR

supply disturbances that are not responses to TR shocks produce
liquidity effects.

It seems that Strongin's methodology is more

closely aligned with these models.
Whether decompositional problems are important or not, I
don't know.

They arise in this model by a specification that at

this point seems somewhat arbitrary.

It is no more arbitrary than

any other specification in the literature, but that does not make
it convincing.

I believe the author's need to make a convincing

argument as to why some forms of morjey creation are more likely to

- 6-




Michael Dotsey
involve liquidity effects than others if their message is to carry
weight. After all, in this model one could easily reverse the
roles of Xt and Gt or make them complimentary.
The discussion on page 11 regarding the estimation of rp is
also a little confusing. With

(3)

n-1 -± -±£
v

V

they claim that 0 can be estimated no matter what the shock. But
is that relevant? We would like to know how j> is influenced
contingent on different shocks. Here a positive V shock induced
by a shift in the demand for loans causes 0 to rise and n to
fall, while a decline in NBR due to an open market sale (G up)
also causes n to fall and i> to rise. It is only the latter effect
that one has in mind when discussing liquidity effects, so perhaps
the ratio is not the correct variable to focus on. Rather, in
this model it should be the relationship between the level of NBR
and the funds rate. Also in estimating 0, one would expect shifts
in the function over time since administration of the discount
window has changed over time. For example, I believe window
administration was more lax when the Fed faced a membership
problem.
I would also downplay somewhat figure one. The interest
rate of consequence is the spread between the funds rate and the
discount rate. When one looks at this graph the correlations seem
at least as pronounced. But has anything but a borrowed reserve
demand function been uncovered?
Finally, the discussion concerning adjustably pegging the
interest rate based solely on technological disturbances raises
questions concerning the nominal determinacy of the model (see
McCallum (1981, 1986)).
Overall, I thought this paper was interesting and
represents a nice attempt to start thinking about how behavior in

- 7 -




Michael Dotsey
the market for reserves influences the correlations we observe
between various monetary measures and the funds rate. Given my
qualms concerning this methodology's ability to explain anything
at business cycle frequencies, I would suggest directing the model
in an alternative direction. Perhaps this framework could be used
to help explain short-term term structure movements in interest
rates and examine the so-called "ozone hole." This line of
inquiry would be interesting since it could integrate reserve
market behavior and a tight specification of policy in a fully
developed general equilibrium model.

- 8 -

Michael Dotsey
REFERENCES
Christiano, Lawrence J., "Modeling the Liquidity Effect of a Money
Shock," Federal Reserve Bank of Minneapolis Quarterly Review,
Winter 1991. 15(1). 3-34.
Christiano, L.J., and M. Eichenbaum, (1992) "Liquidity Effects,
Monetary Policy and the Business Cycle," unpublished ms..
Federal Reserve Bank of Minneapolis, Northwestern University,
NBER and Federal Reserve Bank of Chicago, July 1992.
Fuerst, T.S. "Liquidity, Loanable Funds, and Real Activity,"
Journal of Monetary Economics, vol. 29 (1992), 3-24.
Goodfriend, Marvin. "Discount Window Borrowing, Monetary Policy,
and the Post-October 6, 1979 Federal Reserve Operating
Procedure," Journal of Monetary Economics, vol. 12 (1983),
343-56.
, "Interest Rate Smoothing and Price Level TrendStationarity, " Journal of Monetary Economics, March 1987,
19, 335-48.
Grossman, S. J., and L. Weiss. "A Transact ion-based Model of the
Monetary Transmission Mechanism," American Economic Review,
vol. 73 (1983), 871-80.
Lucas, R.E., Jr. "Liquidity and Interest Rates," Journal
Economic Theory, vol. 50 (1990), 237-64.

of

McCallum, Bennett, T. (1981) "Price Level Determinancy with an
Interest Rate Policy Rule and Rational Expectations,"
Journal of Monetary Economics, vol. 8 (November).
, (1986) " Some Issues Concerning Interest Rate
Pegging, Price Level Determinacy, and the Real Bills
Doctrine," Journal of Monetary Economics vol. 17 (January).
Rotemberg, J.J. "A Monetary Equilibrium Model with Transaction
Costs," Journal of Political Economy, vol. 92 (1984), 40-58.
Strongin, S. "The Identification of Monetary Disturbances:
Explaining the Liquidity Puzzle." Unpublished manuscript,
Federal Reserve Bank of Chicago, December 1991.




Credit Conditions and External Finance:
Interpreting the Behavior of Financial Flows and Interest Rate Spreads
Kenneth N.Kuttner1

Aflurryof recent macroeconomic research has drawn attention to the relationship between
monetary policy, credit conditions, and the markets for short-term debt Two recent papers have
focused onfirms'substitution between bank and non-bank externalfinancein particular, proposing
macroeconomic indicators based onfinancialmarket activity. Kashyap, Stein, and Wilcox (1992)
employ quantity data directly, arguing that the share of bank loans out of firms' total short-term
finance is an informative index of Federal Reserve policy and loan availability more generally. In
a complementary line of research, Friedman and Kuttner (1992) identify monetary policy and bank
lending as potential sources offluctuationsin the spread between yields on commercial paper and
Treasury bills. While both papers have demonstrated solid empirical links between these financial
indicators and real economic activity, neither hasrigorouslyassessed the extent to which fluctuations in these indicators actually represent exogenous changes in credit conditions, rather than
endogenous responses to changing economic conditions. This paper's goal is to provide such an
assessment.
The paper begins with a sketch of the mechanism through which credit conditions affect firms'
short-termfinancing,drawing a distinction between the effects of the Federal Reserve's open market
operations and other factors influencing banks' willingness to lend. The second section summarizes
the reduced-form relationships between real output, the interest rate, and three alternative indices
1. Senior Economist, Federal Reserve Bank of Chicago. I am grateful to Benjamin Friedman and
David Wilcox for their comments and suggestions.




-1-

Kuttner

of credit conditions: the composition of external finance, the spread between the loan rate and the
commercial paper rate, and the analogous spread between commercial paper and Treasury bills.
The third section turns to a closer examination of the impact of monetary policy and loan
availability on bank and non-bank finance using structural VAR techniques. Identifying monetary
policy with innovations to non-borrowed reserves and controlling for firms'financingrequirements,
the first of the three models estimates the dynamic effects of monetary and lending shocks on the
composition of external finance, the interest rate, and real output. The second structural VAR system assesses the effects of reserves and lending shocks on the paper-bill spread. The third model
identifies lending shocks with innovations in the loan-paper spread. Estimates of these models confirm that all three variables respond appropriately to reserves shocks. In addition, lending shocks,
whether identified through financial flows or via fluctuations in the loan spread, induce a substitution
between bank and non-bank finance.
Less clear is the extent to which any of these measures exclusively reflects the effects of changing loan availability. The fact that positive lending shocks are associated with increases in the interest rate and the paper-bill spread suggests that changes in the composition of external finance have
more to do with firms' financing requirements than with exogenous changes in banks' willingness to
lend. Another slightly puzzling observation is that the largest source of changes to the composition
of external finance seems to be wholly unrelated to both reserves and bank lending. Together, these
two results suggest that while credit conditions are one important determinant of firms' choice of
financing, short-term debt flows may be informative for reasons other than those involving the substitution between bank/non-bank substitution. Although its implications for real activity are rather
weak, the loan spread appears to be a plausible alternative measure of credit conditions.

A model offinancialflowsand interest rate spreads
How do the markets for short-term bank and non-bank finance respond to monetary impulses? And
how do non-monetary shocks affect these markets? And how might one construct an index of the
availability of intermediated funds?




-2-

Kuttner

As a first step towards answering these questions, this section analyzes a simple model of the
markets for commercial paper, bank loans, and Treasury bills in the style of Brainard (1964) or
Bosworth and Duesenberry (1973). While not as detailed as either of those models, it is adapted to
highlight firms' tradeoff between bank and non-bank finance. It also draws an important distinction
between purely monetary influences acting through open market operations, and credit conditions
defined more broadly, which may include other factors affecting banks' willingness to lend.
One of the model's more obvious properties is that an injection of reserves causes the interest
rate to fall — the familiar "liquidity effect." Reserves injections also cause the spread between the
interest rates on bank lending and commercial paper to fall, and leads to increased reliance on bank
finance. Lending shocks, which are assumed to affect only banks' preferences over alternative assets, turn out to have similar effects on the loan-paper spread and the composition of firms' finance.
Lending shocks, by contrast, have no effect on the level of interest rates — only the spreads.
The model also identifies two other factors with implications for the money market. First,
firms' demand for external finance may induce changes in the relevant interest rate spreads and
consequently the composition of finance; controlling for this demand-side influence turns out to be
a major challenge to the construction of an empirical measure of credit availability. Similarly, the
stock of outstanding Treasury bills may have tangible effects on the spreads and the composition of
finance.
The three players in the money market are households, banks, and firms, who participate in
the markets for reserves, commercial paper, Treasury bills, and loans. Specifically, households'
portfolios include demand deposits (DD), commercial paper (P), and Treasury bills (B) according
to




DD* = <Krp) W, 4>' < 0
df
df
P* s flrp, rB)W, — > 0 and — < 0
brp
drg
B^ s (1 - <J) - / f o rB)) W,

-3-

Deposit demand
Paper demand
Bill demand

Kuttner

where W is the sum of deposits, paper, and bills held by households. Households' demand for
non-interest-bearing bank deposits is a decreasing function of the prevailing paper rate, rP. A key
assumption is that households view commercial paper and Treasury bills as imperfect substitutes,
so that changes in their relative supplies affect their respective yields.2 Households require a higher
paper rate (or a lower bill rate) to hold a larger share of their portfolio as commercial paper.
Demand deposits are banks' sole liability. Their assets are divided among Treasury bills, loans
(L), and deposits at the Federal Reserve (R) according to:
K* s p(rp)DDt p ' < 0
Ld = Sin, rP, \)DDf — > 0 and — < 0
drL
drP
B*b s (1 - p(rP) - g(rLf rP, K))DD.

Reserve demand
Loan demand
Bill demand

Banks' demand for non-interest-bearing reserves falls with the prevailing paper rate, while loan
demand is increasing in the loan rate and decreasing in the paper rate.3 The stock of reserves is set
at R' by the Federal Reserve; discount window borrowing is ignored.
Banks' demand for loans is also allowed to depend on the variable X, representing any other
factors affecting banks' willingness to lend. These "lending" shocks lead banks to shift the composition of their portfolios between bills and loans; negative shifts in X may be interpreted as "credit
crunch" episodes. These may occur in reaction to a perceived deterioration in borrowers' creditworthiness, or to more stringent capital requirements as suggested by Bernanke and Lown (1991). They
may also be the result of the "moral suasion" instrument of monetary policy; Owens and Schreft
(1992) identify a number of episodes in which banks contracted their lending in response to Federal
Reserve pressure. Whatever the source, the key feature of these "lending" shocks is that they need
not be accompanied by overt monetary policy in the form of open market operations.4
2. Friedman and Kuttner (1992) discuss some possible reasons for this imperfect substitutability.
Lawler (1978) also finds evidence for imperfect substitutability at seasonal frequencies.
3. Note that throughout the paper, assets are "demanded" while liabilities are "supplied." Hence,
banks "demand" loans and bills, while firms "supply" loans and paper.
4. This point is stressed by Friedman (1991).




-4-

Kuttner

Finally,firmschoose between bank lending and paper issuance as sources of short-term finance
according to
fth

hh

dri

orp

V a h(rL, rP)F, —- < 0 and —- > 0

Loan supply

P* » (1 - h(rL, rP))E

Paper supply

For simplicity, the amount to befinanced,F, is assumed to be exogenous with respect to the various
interest rates. Becausefirmsview loans and paper as imperfect substitutes, they willfinancesome
portion of F through bank lending even though rL generally exceeds />; as discussed by Kashyap,
Stein and Wilcox (hereafter KSW), this presumably reflects some intangible benefit accruing to the
firm from maintaining a relationship with a bank. Firms * share of bankfinance(the KSW "mix")
responds predictably to the loan and paper rates: an increase in the loan rate (or a decrease in the
paper rate), leadsfirmsto substitute away from bankfinancetowards non-bank external finance.5
In equilibrium, the demand for the four assets equals their supply,
p(rpMrP)W = l?
frurpiKftW-hirurpyF^O
Kr»rg)W-{l-h(n,r,)yFmO
(1 - g(rL, rPt \))$W+ (1 - / f a rB) - +)W = B*9
determining yields and quantities as functions of the exogenous /?', X, F, and B*. Walras' law
allows the bill market equation to be dropped. Further simplification is possible by assuming the
asset demand and supply functions to be homogeneous of degree zero with respect to the assets'
5. This model embodies the assumption that bank and commercial paperfinanceare viable alternatives for an economically relevant group offirms.However, there is increasing evidence that this
set offirmsis rather small, and that much of the observed variation in the aggregate composition of
finance is due to the relative availability offinanceto small and largefirms;see Gertler and Gilchrist
(1992) and Oliner and Rudebusch (1992).




-5-

Kuttner

yields, so that (for example) g(n +c,rp + c, X) = gin, />, X) for any constant c. In this case, the/, g
and h functions can be specified in terms of interest rate spreads, and the system reduces to:

gizLP,mrpW-h(zu>)F = 0

(I)

KzpBW-(l-h(zu>))F = 0
where zLP and zPB denote the loan-paper and paper-bill spreads.
Analyzing*the comparative statics of (1) is simplified by its (somewhat artificial) recursive
structure. The interest rate level is entirely determined by supply and demand in the market for
reserves; the fall in reserves resulting from a contractionary open market operation requires a higher
rate to equilibrate the reserves market, as illustrated in Figure l. 6 This higher interest rate leads in
turn to a shrinkage of demand deposits and the banking system as a whole. Banks respond by raising
the loan-paper spread, prompting some of its borrowers to switch to alternative forms of finance—
short-term paper in this model. The increased supply of paper (relative to bills) leads to a widening
spread between the paper and bill rates.
The effects of an adverse lending shock resemble those of a reserves contraction in that both
produce a rising loan spread and a substitution towards non-bank finance. Although both shocks
produce similar effects on banks' portfolios, they differ in one important respect: reserves shocks
affect the level of the short-term interest rate, while lending shocks leave the paper rate unchanged.
A fall in X leads banks to shift the composition of their portfolios away from loans and into Treasury
bills, leaving their reserve demand and the paper rate (and consequently deposits and the banking
system's size) unchanged. Banks increase their spreads relative to the paper rate in order to reduce
their stock of loans. As before,firms'increased reliance on commercial paper drives up the paperbill spread.
6. Total wealth is held constant in an open market operation, as the withdrawal of reserves is
offset by a sale of Treasury securities.




-6-

Kuttner

The observation that both reserves and lending shocks may contribute to real economic fluctuations is one explanation of the widespread interest in constructing a broader measure of credit
conditions than reserves or the interest rate in isolation, which reflect largely those shocks originating from the reserves market The attractive feature of the credit conditions indicators discussed
here is their ability to detect the effects of changes in loan availability and reservesfluctuations:in
this model, the "mix," the loan-paper spread, and the paper-bill all reflect the impact of both types
of shocks. In fact, in the absence of any other shocks, all three of these measures should respond to
monetary and credit factors in qualitatively similar ways.
One problem common to all three of these measures (and the interest rate itself) is their susceptibility to contamination from changes infirms'overall demand forfinancing,which may alter yield
spreads and the composition of externalfinancefor reasons having nothing to do with to exogenous
changes in credit conditions.7 This can be illustrated by examining the comparative statics of (1)
in response to an increase in F, the dollar amount of fundsfirmswish to raise from the short-term
credit markets. A greater demand for loanable funds unambiguously increases the prevailing interest rate, />. Its effects on the loan-paper spread (and therefore the composition of external finance)
is ambiguous, as it depends onfirms9share of bankfinance(/t) relative to households* wealth fraction in bank deposits (<|>), and the share of banks' portfolios held as loans (g). When h(zLP) > tyrp)g
(as is presumably the case), increases in F cause loan demand growth in excess of deposit growth,
driving up the relative cost of bankfinanceand the share of paper infirms'external finance.8 The
same inequality is also relevant for the paper-bill spread; a second sufficient condition for a rising
spread is that (1 - h(zLP)) > J{zpB\ so that the increasing paper demand would require households to
hold a larger share of paper in their portfolios.
7. Under most of the Federal Reserves' post-Accord operating procedures, non-borrowed reserves may also be contaminated in this way; see Strongin (1991).
8. A special feature of the KSW model is that changingfinancingrequirements affect loans and
paper proportionally, leaving the "mix" unchanged.




-7-

Kuttner

One additional complication for interpreting the paper-bill spread as a measure of credit conditions is that it may be affected by changes in the outstanding stock of Treasury bills. In addition,
the wealth effects associated with changes in the volume of Treasuryfinancemay alter the level of
interest rates and loan spread, and consequently the composition of external finance.9 In this model,
an increase in the supply of bills reduces the paper-bill spread, as investors require higher returns
to entice them to hold the additional stock of bills. This increase in banks' demand for loans leads
to a fall in the loan rate relative to the paper rate, and increased reliance on bank finance.
To summarize, the model's main implications are:
• Both reserves and lending shocks alter the relative price of bank and non-bank finance, inducing a substitution between alternative forms of external finance.
• By affecting the supply of commercial paper, this substitution also affects the relative
yields on Treasury bills and commercial paper.
• Changes in reserves affect the level of interest rates, while lending shocks leave the
level unchanged.
• Firms' overallfinancingrequirements may affect interest rate spreads and their composition of short-term finance.
The goal of the paper's subsequent empirical work is to explore these implications. Specifically, it
attempts to identify lending shocks through their impact on the composition of externalfinanceand
interest rate spreads, while controlling for reserves and the overall demand for loanable funds.

Short-term credit markets and real economic activity
One desirable feature of any index of credit conditions is a systematic link between it and subsequent
fluctuations in real economic activity.10 The results below summarize the predictive properties of
the KSW "mix," the prime-paper spread, and the paper-bill spread. The results show that the "mix"
9. Of course, this assumes that households view government bonds as net wealth; see Barro
(1974).
10. Economists and market observers have long recognized the cyclical properties of commercial
paper, bank lending, and their relative yields; see, for example, Foulke (1931), Selden (1963), and
Stigum (1990).




-8-

Kuttner

and the paper-bill spread are good predictors of future changes in real GDP (although this alone does
not justify their interpretation as measures of credit availability).
"Causality" tests
Table 1 examines the incremental information content of the three measures for future changes in
real GDP in the presence of traditional measures of monetary policy: non-borrowed reserves and
the commercial paper rate. Regressions 1-3 are four-variate reduced-form equations of the form
4

4

4

4

Ax, a Ho + Hi* + ] T OjAx^ + ^T pi[A ln(J?),w + ] T Y,Arj>^ + ] T 6 , A ^ + e,
where x is the logarithm of real GDP, R is non-borrowed reserves adjusted for extended credit and
deflated by the GDP deflator, i> is the commercial paper rate, and q denotes, in turn, the "mix", the
loan-paper spread, and the paper-bill spread. As in KSW, the "mix" is computed as the observed
ratio of bank lending to the sum of lending to commercial paper, or L/(L + P).n The results use the
six-month commercial paper and Treasury bill yields, and the prime rate (from the Federal Reserve
H.1S release) is used as the lending rate.
The table reports F-tests for the exclusion of the four 6, terms for the entire 1960:2-1991:4
sample, as well as two shorter samples. One truncated sample begins in 703, when Regulation Q
was eliminated forroostlarge CDs.12 Another begins in 1975:1. Although this date is somewhat
arbitrary, it corresponds roughly to the beginning of a rapid expansion of the commercial paper
market, during which it became a more popular vehicle for non-financialfirms'short-term finance.13
11. The augmented Dickey-Fuller u statistic (computed with eight lags) for the stationarity of the
"mix" is -4.10, rejecting the null hypothesis of nonstationarity at the 1% level. Consequently, it is
included here in levels along with a linear trend term.
12. Regulation Q interest rate ceilings on 30-89 day CDs in denominations of $100,000 were
eliminated on June 24, 1970. Ceilings on CDs with maturities in excess of 90 days remained in
place until March 16,1973.




-9-

Kuttner

The 1975-91 sample also excludes the Penn Central and Franklin National disruptions of 1970 and
1974, and covers the period in which ratings were assigned to commercial paper issues.14
The results of thefirstregression corroborate the strong link between the "mix" and real output
found by KSW, supporting theirfindingthat the composition of finance has significant predictive
power for future real economic activity, even in the presence of reserves and interest rates. The
poor performance of the loan-paper spread in the second regression (again in the presence of reserves and the commercial paper rate) is consistent with the notion that banks' lending rates are
relatively uninformative.15 The third regression demonstrates the incremental information content
of the paper-bill spread — at least in the earlier samples.
Impulse responses
While the F-statistics for "causality" give some indication of the strength of the predictive power of
thesefinancialindicators, they give no indication of the size or direction of their impact. The impulse
response functions plotted in Figure 2 provide a richer description of the effects of innovations
to the financial indicators. Each of the three rows of graphs is from the VAR corresponding to
regressions 1-3 in Table 1. In each case, the system has been orthogonalized (according to the
triangular Cholesky decomposition) with the credit conditions index in last place. Three responses
are plotted for each regression: thefinancialindicator's effects on output and the interest rate, and
the effect of reserves innovations on thefinancialindicator. The dotted lines depict the approximate
95% confidence bounds.
Panels (a) and (b) from the first specification show that "mix" innovations indeed act like
reasonable measures of credit conditions; reserves injections increase the share of bank loans, and
13. At the end of 1974, non-financial commercial paper accounted for only 13.5 billion dollars.
By 1982, thisfigurehad grown 325.2 percent to 57.4 billion. See Hurley (1977,1982), and Stigum
(1990).
14. Moody's and Standard and Poor's began rating commercial paper in 1974.
15. Similar results are obtained with the average of large banks' lending rates obtained from the
Federal Reserve Survey of Terms of Bank Lending reported in release E.2.




-10-

Kuttner

output rises in response to positive "mix" shocks, which might be interpreted as the pure lending
component of credit conditions. The panel (c) plot, however, is something of a puzzle. It shows that
"mix" innovations are associated with a rising commercial paper rate—not what one would expect
from an increased willingness to lend on the part of banks, and inconsistent with the implications
of the model presented earlier.16 However, this pattern is consistent with banks passively supplying
more loans in response to rising demand for credit.
The second row of plots confirm the generally weak relationship between the prime-paper
spread and real output. One interesting feature of the loan spread is that it initially rises in response
to a reserves innovation — clearly inconsistent with the loosening of credit conditions implied by
the reserves injection. The loan spread ultimately falls, however, suggesting that this response is
due to a certain sluggishness in the way banks adjust their lending rates.
The impulse response functions from the paper-bill spread regression are all consistent with
what one would expect from an indicator of credit conditions: positive shocks to the spread generate
declining real output, while reserves injections reduce the spread. Furthermore, unlike the "mix",
innovations in the spread itself have essentially no impact on the level of interest rates.
Comparing the "mix" and the paper-bill spread
Because regressions 1-3 included each of the credit conditions measures in isolation, the results raise
an important question: to what extent are the three indicators measuring the same phenomenon? An
obvious way to address this question is to include more than one indicator in the same regression to
see if the presence of one vitiates the predictive power of the other.
The results from two additional regressions (numbered 4 and 5) are reported in Table 1. The
results from specification 4, which includes both the "mix" and the loan spread, are not surprising
given the weak performance of the loan spread in isolation — the F-statistics for the "mix" remain
virtually unchanged. Somewhat more surprising are the results from specification 5, in which both
the "mix" and the paper-bill spread appear. Here, the relationship between the two variables and real
16. The "mix" terms are significant in the interest rate equation at the 10% level.




-11-

Kuttner

output is uniformly stronger (judged by the F-statistics) than when they are included individually.
Qearly, one (or both) of the indicators is doing something other than simply summarizing the state
of credit market conditions.
The roles of commercial paper and bank loans
The model sketched earlier suggests that flows of commercial paper and bank lending are informative to the extent that they reflect the substitution between the two forms of finance in response to a
monetary or a lending shock. KSW exploit this insight by looking at the ratio of bank loans to the
sum of loans and paper, shocks that affect both forms of debt proportionally are presumed to stem
from sources other than loan availability. A useful check on this specification is to verify that paper
and lending flows enter an unrestricted regression in such a way that the "mix" is the variable that
matters.
This is easily accomplished by differentiating the "mix" (designated h) with respect to time,
P
—
L
dt = (I+P)*

L

P

(I+P) 2

= h{\ - h%/L - /i(l - h)P/P9
decomposing its movements into distinct lending and paper contributions. In discrete time, the
analogous decomposition,
AA, - AM(1 -

AKI)AL/1M

- A M (1 - A,-i)AP/P M -

<&L -

Afip

expresses A/i as a weighted sum of commercial paper and bank loan growth rates, denoted tJxL
and tJip. If AA were in fact the appropriate measure of the impact of credit conditions on the real
economy, the two components would enter real output regressions with equal and opposite signs;
the regression itself would "choose" the KSW specification.
Table 2 displays the results of this experiment. Panel (a) reports the outcome of a regression
of first-differenced log real GDP on four lags of output, tJxL and tJiP over the 1960:2-91:4 sample.
Judged by the F-statistics, the commercial paper terms are much more informative than the lending




-12-

Kuttner

terms; tJip is significant at the 0.01 level, while the tJiL terms are not significant at even the 0.10
level.17 The sum of the estimated coefficients on lending is negative, but statistically insignificant
The regression in panel (b) refines the test by specifying the regression in terms of tJi and
tJip — simply a transformation of the regression in panel (a). Excluding the four lags of &hP is
equivalent to restricting the coefficients on iskL and tJip to have equal and opposite signs. Here,
the tJi terms are statistically insignificant, while the AhP terms are significant at the 0.05 level.
Moreover; the negative estimated sum of the "mix" coefficients is inconsistent with the substitution
hypothesis, although this sum is again statistically insignificant.
To guard against the possibility that the results in the first two panels are an artifact of the
differenced specification, panel (c) reports the results of a regression that includes a linear trend and
h in levels. While not tJ\P terms are not as strong in the levels specification, the coefficients on the
h terms remain statistically insignificant.
These experiments show that the "mix" owes its predictive power in large part to something
other than the substitution between bank and paper finance. In unrestricted equations, h terms are
generally insignificant, while the hypothesis that commercial paper in isolation does not matter for
predicting real output can be rejected. This observation suggests a closer examination of lending and
commercial paper flows individually, and their relation to monetary policy and credit conditions.

A structural approach to identifying lending shocks
The atheoretical results in the preceding section provided some evidence in favor of interpreting the
financing "mix" and the paper-bill spread as measures of credit conditions, although innovations in
the composition offinancewere, contrary to the simple model, are associated with a rising interest
rate. One reason for this pattern may be the result of inadequately controlling for the overall demand
for short-term finance. As demonstrated earlier, an increase in the amount to befinancedneed not
raise bank and non-bankfinanceproportionally. In this case, if increases infirms'demand for funds
17. This is consistent with the results of King (1986).




-13-

Kuttner

are accommodated primarily through bank lending, the "mix" may rise for reasons unrelated to
credit conditions.
Figure 3 plots thefinancinggap (defined as the difference betweenfirms'capital expenditures
less inventory IVA and after-tax internal funds) along with commercial paper and bank loan flows,
demonstrating the close relationship between thefinancinggap and the volume of bank lending
(although commercial paper appears to have become more sensitive to thefinancinggap in the later
part of the sample). To control for credit demand, the results in this section include the financing
gap as an additional determinant offirms'debt issuance.
A more interesting alternative hypothesis is that is that the substitution mechanism inadequately explains the joint behavior of commercial paper and bank lending, and that factors other
than monetary policy are what drive the observedfluctuationsin the composition of short-term external finance. The apparent asymmetry between the effects of loan and paper flows uncovered in
Table 2 provides some circumstantial evidence for this view.
The results presented in this section attempt to address these issues by separately analyzing
flows of lending and commercial paper in a structural VAR setting that controls for the overall
demand for loanable funds. Moving to a more structural approach also addresses the possibility that
the interest rate's odd response to "mix" shocks is as an artifact of the artificial triangular structure
of the Cholesky decomposition employed earlier. Thefirstmodel focuses on the response of lending
and paper flows to reservesfluctuations,and examines the properties of the innovations identified
as lending shocks. The second describes the response of the paper-bill spread to thefinancialflows
generated by reserves and lending shocks. The third usesfluctuationsin the loan-paper spread as
an alternative means of identifying lending shocks.
A review of structural VARs
Beginning with an unrestricted i-variate dynamic simultaneous equation system,




T

-14-

Kuttner

the standard VAR achieves identification by restricting the contemporaneous relationships between
the elements of y, i.e., by setting BQ = 0 and A = /, while placing no restrictions on the covariance
matrix of v, ie., £(w') = Q. The structural VAR introduced by Blanchard and Watson (1986) and
Bernanke (1986) achieves identification by allowing some nonzero elements in thcB0 matrix, while
restricting the covariance matrix of v, the structural disturbances, to be diagonal. Off-diagonal
elements in A can be introduced to allow distinct elements of y to depend on common structural
shocks. Thus, structural VARs differ from traditional structural models by replacing the assumption
of an exogenous instrument set with the assumption of orthogonal structural shocks. At the same
time, the dynamics of the system are left unrestricted, as in the conventional VAR.
Another interpretation of the structural VAR is as a decomposition of the covariance matrix of
VAR residuals. If the structural disturbances are uncorrected with one another, Lc, £(w') = D, Q,
the covariance matrix of the VAR errors becomes a nonlinear function of the structural parameters:
Q=£(J#Av,v,'A'J#)
mBfADA'Bf.
If the system is just-identified, the above equality is exact; B^AD™ is a matrix square root of Q,
and A'1 B0 diagonalizes Q.18
Reserves, lending, and short-term debt flows
Thefirstmodel is a just-identified six-variable system involvingfinancinggap (F)> bank lending,
non-financial commercial paper (P) the commercial paperrate(r>), real GDP (x), and non-borrowed
reserves adjusted for extended credit (R). The interest rate is differenced, while reserves and GDP
enter as log differences. The lending and paper data are again taken from the Flow of Funds accounts
for the non-farm, non-financial corporate and noncorporate sectors. With F, P and L expressed
18. With a total of 2A2 elements in A and B0 and only &(&+1)/2 unique elements in Q, it is clear that
the stnicmral parameters are not identified without additional restrictions on A and£0. The Cholesky
decomposition, which is equivalent to setting B0 = / and making A lower triangular, is but one possibility. In overidentified systems, the problem becomes one of choosing the stnicmral parameters
in Bo and A to generate the bestfitbetween thefittedand the observed covariance matrices.




-15-

Kuttner

as shares of the total dollar volume of outstanding paper and loans, changes in the "mix" can be
constructed as the weighted average of the two flows:
v

'L+P

L+P

The substance of the model is contained in the six equations describing the contemporaneous
relationships between the variables,
R as bU6x + vx

Reserves (2a)

F as bxxR + bz& + V2

Financing gap (2b)

rP ss bx\R + bxiF + V3

Interest rate (2c)

L as b<xR + b^iF + d<30» + v4
P ss b^R + fci2P+*s^> + *MV4 + v5
x as d 0 r P + 644I + fc^P + v6

Lending (2J)
Paper (2e)
Output (20.

No restrictions are placed on the dynamics of the system; consequently, terms dated t-\ and before
are omitted, but implicit.
Equation 2a allows the Federal Reserve to vary reserves contemporaneously with real GDP in
a primitive feedback relationship. Thefinancinggap (equation 2b) also depends on the level of real
economic activity. Consistent with the model presented earlier, the commercial paper rate in 2c is
a function of reserves and thefinancinggap.
The model's key equations are 2d and 2e, describing the behavior of bank lending and commercial paper flows as a function of thefinancinggap, reserves, and the interest rate. The coefficients
on F measure the proportion of the currentfinancinggap satisfiedfinancedthrough loans and paper. The two equations' coefificients on R determine the immediate response, ceteris paribus, of
the two forms of short-termfinanceto changes the banking system's reserve position. The v4 term
in the lending equation represents lending shocks that are orthogonal to reserve andfinancinggap
innovations, which would include factors such as credit crunches. For this interpretation of v4 to be




-16-

Kuttner

legitimate, one of two conditions has to hold: either the observed financing gap must appropriately
control for firms' demand for funds, or the amount of funds banks have available is fixed in the
current quarter.
The v4 innovation also appears in the commercial paper equation with the coefficient a^, allowing commercial paper to respond directly to lending shocks. This parameter determines the
extent to which lending shocks are "recycled" into the commercial paper market within the current quarter. The v5 term in the commercial paper equation accounts for shocks to paper issuance
uncorrected with the other structural disturbances. The final equation for real GDP is a reducedform equation describing the economy's response to the reserves and credit shocks in the preceding
equations.
The parameter estimates in Table 3 summarize the model's contemporaneous behavior, while
the impulse responses functions plotted in Figure 4 describe its dynamics of the system whose orthogonalization is implicit in equations 2a-2f. Like the earlier reduced-form regressions, these
results provide some evidence to support the use of lending flows as an indicator of credit conditions, while confirming the doubts raised in the atheoretical VARs. First, The negative estimate
of the coefficient on R in the lending equation (2d) contradicts the hypothesis that the primary effect of monetary policy is a substitution between bank and non-bank finance; the contemporaneous
response of an injection of non-borrowed reserves, ceteris paribus, is a fall in bank lending.
However, because of the contemporaneous relationship from reserves to the financing gap and
short-term finance via the interest rate and output, the coefficients on R in equations 2d and 2e do not
by themselves determine the overall response of the "mix" to a reserves shock. The actual responses
can be read from the impulse response function, plotted in the top panel of Figure 4.19 This shows
that the net effect of a reserves injection is initially rather small, with the loan share gradually rising
after two to three quarters.
19. The sample average values of h are used to compute the approximate response of the "mix"
from the impulse response functions of the underlying variables.




-17-

Kuttner

Figure 4 also shows that lending shocks seem to have a considerably larger impact on the
composition of externalfinancethan reserves for thefirstfour quarters. Lending shocks9 effect is
strengthened somewhat by the statistically significant negative estimate offl$4>which is consistent
with roughly 10% of the lending shock being "recycled" into the paper market in the current quarter.
The coefficients on F in the paper and lending equations show that neither responds immediately to
fluctuations in thefinancinggap.
A strong liquidity effect is associated with injections of non-borrowed reserves; the paper rate
falls contemporaneously (the negative coefficient on R in equation 2c) and over a longer horizon (the
center panel of Figure 4). These results also confirm the curious positive relation between the "mix"
and the level of interest rates highlighted earlier in the paper. The center panel shows that positive
lending innovations imply a rising interest rate, contradicting the theoretical model's implications
for the effects of lending shocks.
Both monetary and lending shocks are important sources of output fluctuations. Increased
bank lending is contemporaneously associated with more rapid real GDP growth in the short run,
as shown by both the positive (but not quite significant) coefficient on L and the impulse response
function.
What about shocks to commercial paper, v5? The top panel of Figure 4 shows that these shocks
— which are, by construction, orthogonal to the system's other structural disturbances — have the
largest and most persistent impact on the composition of external finance. Interestingly, the center
panel shows that these innovations have essentially no implications for the interest rate, although
they do seem to have a small, negative impact on real output.
Financial flows and interest-rate spreads
Recent papers by Bernanke (1990) and Friedman and Kuttner (1992) suggest that the substitution
between bank and non-bank debt is an important source of fluctuations in the paper-bill spread. As
discussed earlier, monetary contractions reduce lending by shrinking the stock of deposits, leading




-18-

Kuttner

firms to raise the loan rate relative to the paper rate, discouraging intermediated borrowing. Similarly, adverse lending shocks cause banks to shift from loans to Treasury bills. Asfirmsturn to the
paper market to satisfy theirfinancingneeds, the paper supply rises and bill supply to households
falls, raising the paper-bill spread. If this is the way in which credit conditions affect the spread,
one would expect tofindthe mechanism operating through the volume of outstanding non-financial
commercial paper.
The second structural VAR is designed to detect the operation of this mechanism. It augments
thefirstmodel (equations 2a-2f) with the addition of a seventh equation for the paper-bill spread,
x a b&rP + b^JL + b^sP + b^(rP - rB) + v6

Output (3/)

rP-rB = bitiR + 6 7f2 F+b^rp + 67,4^ + bn,$P + v?, Paper-Bill spread (3g)
and also includes the spread in the output equation. The remaining five equations are identical to
those in the earlier model (2a-e).
The parameter estimates reported in Table 4 provide weak evidence for bank/non-bank substitution as a source of paper-bill spread. The positive and marginally significant on the paper term
shows that flows of non-financial paper do exert an influence on the spread.20 However, the very
large, significant coefficient on reserves shows indicates that a great deal of the impact of monetary
policy is transmitted to the spread via other routes.
The impulse responses in the top panel of Figure 5 confirm the spread's strong reaction to
non-borrowed reserves innovations. Positive shocks to thefinancinggap also drive up the spread,
as predicted, while paper shocks have little or no impact. Lending shocks again pose a problem,
however. If the lending innovations identified by the VAR correspond to changes in the availability
of loans, the model suggests that positive shocks should be associated with a falling paper-bill
spread. The opposite is true: lending shocks imply a rising spread. Again, this pattern is consistent
20. By contrast, the results in Table 10 of Friedman and Kuttner (1992) using the total volume of
commercial paper outstanding are consistent with a stronger link between paper issuance and the
spread.




-19-

Kuttner

with bank lending responding passively to changes in the demand for funds inadequately captured
by the financing gap.
The bottom panel of Figure 5 suggests something other than bank/non-bank substitution is
driving the paper-bill spread. Despite the inclusion of a variety of financial variables purporting to
capture the impact of monetary policy on credit markets, the graph shows that the spread continues
have strong implications for future output — comparable in magnitude to those of non-borrowed
reserves. Even accounting for reserves, lending, and paper shocks, orthogonal spread innovations
still result in falling real economic activity.
Identifying lending shocks with loan spread innovations
In light of the conclusion that lending flows (and the "mix") may in part represent endogenous
response to firms' financing demands, the third structural VAR uses an alternative assumption to
identify lending shocks, attributing (orthogonalized) innovations in the loan-paperspread to changes
in banks9 willingness to lend. In the context of the simple model presented earlier, the loan spread
should embody exactly the same information as the "mix." In practice, as KSW note, the loan rate
is likely to be a poor measure of the true cost of bank finance, an observation that motivates their use
of the quantity variables. Indeed, the sluggish response of the loan rate to changes in the paper rate
corroborates this view. The weak response of output to the loan-paper spread makes this approach
seem even less promising.
With these reservations in mind, thefirststructural VAR can be adapted to incorporate the loanpaper spread. An equation for the loan spread is added to the system, and lending and paper flows
are allowed to depend on this spread, as well as on reserves and the financing gap. The covariation
between the flows that is a function of credit conditions is a result of their common dependence on
the loan spread. This identification scheme will work if the financing gap is an imperfect proxy for
the overall demand for funds so long as banks passively accommodate firms' funding requirements
within the quarter at the going spread (that is, if their demand for loans is elastic).




-20-

Kuttner

The modified system is:
R s bi& + Vi

Reserves'(4fl)

F m bZ\R + bZ6x + V2

Financing gap (4b)

r> « buR + b^JF + v3

Interest rate (4c)

rL-rP = b<tR + b^F+b^rP

Loan spread (4rf)

I =fcMrt+ 6 ^ + b & r p + 65,4(^1 - rP) • v5
P « 6Mtf + *42F+fc^r/. -•- A^fo. - TP) + **sV5 + v6
* s ^73r/» + ^ f a , - rP) + 67,5^ • th,*? + ^7

Lending (4e)
Paper (4/)
Output (4g).

Under the assumptions outlined above, the innovations to the loan spread equation are now associated with changes in credit conditions, while the v5 lending innovations represent shocks to firms'
loan supply (that is, their demand for funds).
The parameter estimates in Table 5 accord surprisingly well with the implications of the model.
Although its sluggish response makes the loan spread is subject to large, transitory effects from
the paper rate and reserves, the negative estimated 65,4 and the positive b$4 show that loan and
paper volume respond as they should to the spread. Furthermore, reserves have no discernible
independent impact on financial flows. A rising loan spread is contractionary, although again, the
effect is statistically weak.
The corresponding impulse response functions appear in Figure 6. The top panel again illustrates the consequences of sluggish loan rate adjustment, with reserves injections causing the
loan spread to rise sharply in the current quarter. Over time, reserves innovations produce a falling
spread. The center panel shows the familiar liquidity effect, and the positive impact of lending innovations on the commercial paper rate. In this model, however, with innovations to loan volume
interpreted as shocks to firms' funding requirements, the result is perfectly natural. By contrast,
innovations in the loan spread have quite mild effects on the paper rate.




-21-

Kuttner

Conclusions
This paper has examined the relationship between monetary policy, loan availability, and alternative
indicators of credit market activity. One of is mainfindingsis that the substitution between bank and
non-bank finance is indeed an identifiable effect of monetary policy as measured by innovations to
non-borrowed reserves. This substitution is, however, not the only factor affecting financial flows.
One of the major contributors to the aggregate composition offirms'short-term obligations is flows
of commercial paper unrelated to lending shocks.
Furthermore, the portion of bank lending not attributable to monetary policy is associated with
increases in the commercial paper rate and the paper-bill spread, suggesting that the behavior of the
KSW "mix" is in part due to changes in firms' demand for loanable funds. Despite its apparent
slow adjustment to changes in market interest rates, the loan-paper spread is a plausible alternative
indicator of credit conditions.
The paper-bill spread responds appropriately to monetary shocks, rising in response to a reserves contraction. However, the strength of its response cannot entirely be accounted for by flows
of non-financial paper, suggesting that its informativeness as a predictor of real economic activity
may be due to other sources, such as changes in banks' issuance of negotiable CDs. This is consistent with the observation that non-financial commercial paper comprises a tiny share of the relevant
market—only 25% of total commercial paper, and less than 9% of the sum of paper, CDs and Treasury bills.21 Understanding how Federal Reserve policy and credit conditions affect the paper-bill
spread will require expanding the model to take into account the behavior of other relevant assets,
such as CDs andfinancialpaper.

21. These figures are for 1991:4. The share of non-financial commercial paper is even smaller
earlier in the sample.




-22-

Kuttner

References
Barro, Robert (1974), "Are Government Bonds Net Wealth?** Journal of Political Economy .82,
pp. 1095-1118.
Bernanke, Ben S. (1986), "Alternative Explanations of the Money-Income Correlation,**
Carnegie Rochester Conference Series on Public Policy 25, pp. 49-100.
Bernanke, Ben S. (1990), "On the Predictive Power of Interest Rates and Interest Rate
Spreads,** New England Economic Review November-December, pp. 51-68.
Bernanke, Ben S. and Cara S. Lown (1991), "The Credit Crunch,** Brookings Papers on Economic Activity 2, pp. 205-39.
Blanchard, Olivier, and Mark Watson (1986), "Are Business Cycles All Alike?** in Robert A.
Gordon, ed., The American Business Cycle: Continuity and Change. Chicago: The University of Chicago Press and the NBER.
Bosworth, Barry and James S. Duesenberry (1973), "A Row of Funds Model and its Implications*' in Issues in Federal Debt Management, Federal Reserve Bank of Boston Conference Series 10, pp. 39-149.
Brainard, William C. (1964), "Financial Intermediaries and a Theory of Monetary Control,**
Yale Economic Essays 4, pp. 431-82.
Foulkc, Roy A. (1931), The Commercial Paper Market New York The Bankers Publishing
Company.
Friedman, Benjamin M. (1991), "Comments on Bernanke and Lown,** Brookings Papers on
Economic Activity 2, pp. 240-44.
Friedman, Benjamin M. and Kenneth N. Kuttner (1992), "Why Does the Paper-Bill Spread
Predict Real Economic Activity?** forthcoming in James H. Stock and Mark W. Watson
eds., New Research in Business Cycles, Indicators and Forecasting, Chicago: University
of Chicago Press and the NBER.
Hurley, Evelyn (1977), "The Commercial Paper Market,** Federal Reserve Bulletin 63, June,
pp. 525-536.
Hurley, Evelyn (1982), "The Commercial Paper Market since the Mid-Seventies** Federal Reserve Bulletin 68, June, pp. 327-333.
Gertler, Mark and Simon Gilchrist (1992), "The Role of Credit Market Imperfections in the
Monetary Transmission Mechanism: Arguments and Evidence,*' Manuscript.
Kashyap, Anil, Jeremy C. Stein, and David Wilcox (1992), "Monetary Policy and Credit Conditions: Evidence from the Composition of External Finance,*' NBER Working Paper
#4015, Cambridge: National Bureau of Economic Research.
King, Stephen R. (1986), "Monetary Transmission: Through Bank Loans or Bank Liabilities?"
Journal of Money, Credit and Banking 18, August, pp. 290-303.
Lawler, Thomas A. (1978), "Seasonal Movements in Short-term Yield Spreads,*' Federal Reserve Bank of Richmond Economic Review, July/August,.




-23-

Kuttner

Oliner, Stephen D. and Glenn D. Rudebusch (1992), "The Transmission of Monetary Policy to
Small and Large Firms," Manuscript.
Owens, Raymond E. and Stacey L. Schreft (1992), "Identifying Credit Crunches," Federal Reserve Bank of Richmond Working Paper #92-1.
Selden, Richard T. (1963), Trends and Cycles in the commercial Paper Market," National Bureau of Economic Research Occasional Paper #85.
Stigum, Marcia (1990), The Money Market Homewood: Dow Jones Irwin.
Strongin, Steven H. (1991), "The Identification of Monetary Policy Disturbances: Explaining
the Liquidity Puzzle," Federal Reserve Bank of Chicago Working Paper #91-24.




-24-

Kuttner

1.

F-Statistics for Alternative Measures of Credit Conditions in
Quarterly Real Output Equations

60:2-91:4

70:3-91:4

75:1-91:4

(1) "Mix" alone

3.36"

2.09*

286"

(2) Loan spread alone

an

0.30

0.46

(3) Paper-bill spread alone

3.81"*

2.71"

1.81

(4) "Mix" + loan spread
"mix" terms
loan spread terms

3.37"
0.23

1.81
0.14

3.07"
0.79

(5) "Mix" + paper-ttll spread
"mix" terms
paper-bill spread terms

4.17"*
4.62'"

2.46
3.07"

4.39"*
3.30"

Specification

*
**
***
Notes:




Significant at the 10% level
Significant at the 5% level
Significant at the 1% level
The regressions are based on qtiarterly data for the sample indicated.
In addition to the variables indicated, each regression includes four lags of real GDP
growth, real non-borrowed reserves growth, the differenced commercial paper rate, plus
constant and trend terms.

-25-

Kuttner

2.

Decomposing Changes in the Composition of External Finance

(a) Regression with separate commercial paper and bank lending terms
Exclusion F-stat
(p-value)
Commercial paper (AA/>)
Bank lending (AAL)

Sum of coefficients
(p-value)

4.00
(0.005)

-0.51
(004)

1.39
(Q24)

-0.90
(0.22)

(b) Regression with the differenced "mix" and commercial paper
Exclusion F-stat
(p-value)
"Mix" (AA)
Commercial paper (iJip)

1.45
(0.22)
264
(0.04)

Sum of coefficients
(p-value)
-0.91
(0.23)
-1.38
(0.04)

(c) Regression with the "mix" in levels, commercial paper, and linear trend
Exclusion F-stat
(p-value)
"Mix" (h)
Commercial paper (A/i/>)

Notes:




1.48
(0-21)
227
(0.07)

Sum of coefficients
(p-value)
0.11
(0.16)
-1.77
(0.01)

The regressions are based on quarterly data for 1960:2 through 1991:4.
The specifications include four lags of each included variable and a constant term.

-26-

Kuttner

3.




Structural VAR Estimates, Credit Conditions Identified via Lending Flows
(equations 2a-2f)

2a.
2b.
2c.
2d.
2e.
2f.

Notes:

R= -0.625 x+ v,
(1.94)
F= 0.159 r+ 1.974
(1.05)
(4.11)
/>=-0.208 R+ 0.037
(6.96)
(210)
L = -0.396 * - 0.022
(1.51)
(ttl6)
P = ttl35 rt + 0.027
(1.29)
(0.51)
x = -0.023 r+ 0.019
(0.23)
(1.50)

x+v*
F+vj
F+ 2125 r, + v4
(3.23)
F + 0.444 r>- 0.094 v4 + v5
(1.68)
(271)
1 + 0.004 P + v6
(ttl4)

Estimates are based on quarterly data for 1960:2 through 1991:4.
Regressions include three lags of each variable, constant and trend terms.
Numbers in parentheses are /-statistics.

-27-

Kuttner
4.

Structural VAR Estimates of the Effects of Lending Shocks on the Paper-Bill Spread
(equations 3a-3g)

3a.
3b.
3c.
3d.
3e.
3f.
3g.

Notes:




/?=-0.558 x+V!
(1.51)
F = 0.048 r+ 1.882
(033)
(3.70)
r, * -0.216 R + 0.027
(7.63)
(1.57)
L = -0.306 R- 0.022
(1.18)
(017)
P = 0.110/?+ 0.038
(1.05)
(071)
x= 0.123 r+ 0.016
(1.20)
(1.35)
r/,-r,=

x+vj
F+ v,
F+ 2051 r, + v4
(3.05)
F+ 0.470 r , - 0.091 v4+
v5
(1.73)
(261)
1+ 0.023 P - 0.897 (rP-rB) + v6
(076)
(3.20)

0.016/?+ 0.181 r+ 0.000 F+ 0.002 1 +
(1.40)
(6.20)
(007)
(046)

0.016 P + v,
(1.71)

Estimates are based on quarteriy data for 1960:2 through 1991:4.
Regressions include three lags of each variable, constant and trend terms.
Numbers in parentheses are /-statistics.

-28-

Kuttner
5.

Structural VAR Estimates, Credit Conditions Identified via the Loan Spread
(equations 4a-2g)

4a.
4b.
4c.
4d.
4e.
4f.
4gNotes:




R= -0.347 x+ v,
(0.85)
F= 0.122 r+ 2132 x +
(a89)
(4.46)
i> = -0.202 rt + 0.016 F+
(aiO)
(0.96)
r t - r P = 0.070/?+ 0.014 F (4.99)
(1.77)
1 = -0.039 rt + a021 F +
(0.14)
(ai5)
P= 0.030 J? + a003 F +

V2

»*

a261
(6.56)
a783
(0.96)
0.722
(218)
(a27)
(ao6)
x= -0.047 r - 0.362 ( n - rP)-f 0.016
(0.39)
(1.54)
(1.28)

/> + v4
r, - 4.727 (rL - »>) + vs
(299)
/> + 1.176 (rL-rP)- 0.085 v5 + v6
(1.83)
(242)
1+ 0.015 P+
v7
(0.46)

Estimates are based on quarterly data for 1960:2 through 1991:4.
Regressions include three lags of each variable, constant and trend terms.
Numbers in parentheses are /-statistics.

-29-

Kuttner

Figure 1

reserves market

Rs Cf«0

l-fp




*"f "*"*

loan market

paper market

\yS

-30-

(I

w

(HHi)

Figure 2
Impulse Response Functions of Credit Conditions Indicators
(a) mix -> output

(b) reserves -> mix
0.0048

(c) mix -> interest rate
0.0032

0.00361
0.00161

0.0024
0.00121

0.0000
T^-r

0.0000
-.0012

3

6

1

9

1—r-

0

(d) loan spread -> output

I

I

3

I

I

I

(e) reserves -> loan spread

0.0020

-.0016

6

0.0036

(0 loan spread -> interest rate

i
CO




-.00201

i—i • i

i

(g) paper-bill spread -> output
0.0016

(h) reserves -> paper-bill spread
0.0007 n

0.0000
-.0016
-.0032
-.0046

.0040 • L -i—i—i—i—i—i—r-

r-

«

i—i

3

i

r

- t •

9

r •

0) paper-bill spread -> interest rate
0.00501

Figure 3: Financing Gap and Financial Flows
bank lending and paper issuance, four-quarter moving average

1

v>

• 2
w

c

•o2 18




-36

.• I ' I ' I ' I ' I ' I ' I ' I • I ' I ' I • I • I • I • I • I • I • I ' I ' I ' I ' I ' I ' I ' I ' I ' I ' I

60

63

66

69

72

75

78

81

84

87

90

Kuttner

Figure 4: credit conditions = lending shocks
response of the Mix

0.16
0.08 H
o.oo
-.08 -j

reserves
lending

•.16 -|

paper
-fin gap—

.24

T

1

1

P

I

I

1

8

9

1

10

11

response of the interest rate
0.035

0.000

.035 H

.070

i

1

7

8

11

reserves
Ien3ing
paper

—
0.050 -

^^"^

— —
.-.-—..

.-v^—-•

0.025 -

0.000 —'

1

i

10

response of real output

n f\7K —i
Kj.UfD

-.025 -

1

9

******

.




,

,

**"*

,

2

.

3

,

4

,

5

6

- 33 -

,

7

,

8

,

9

,

1

10

11

Kixttner

Figure 5: credit conditions = lending shocks
response of the paper-bill spread

0.008

0.000

-.008 H

-.016

l
9

r
10

11

response of real output

0.10

0.05 H

0.00

J~^

\
•.05

1




0

1
1

1

1
2

n
3

4

1
5

1
6
- 3H -

1
7

1
8

1
9

1

1
10

11




Kuttner

Figure 6: credit conditions = loan spread shocks
response of the loan spread
reserves
lending
paper

response of the interest rate

s

s

/
" •••••i^>%

response of real output

- 35 -

COMMENTS ON
CREDIT CONDITIONS AND EXTERNAL FINANCE:
INTERPRETING THE BEHAVIOR OF FINANCIAL FLOWS AND INTEREST RATE SPREADS
David Wilcox

Two opposing views have animated much recent research on the
transmission channels of monetary policy. One view (stated in its
extreme form) is that the impulses of monetary policy are transmitted
to the real economy exclusively via the market for reserves. By
manipulating the quantity of available reserves, the Federal Reserve
is able to change the relative supply of money and bonds. Given this
change in relative supply, the interest rate must change in order to
clear the markets for money and bonds. In turn, the change in the
interest rate alters the user cost of capital, and so influences the
investment decisions of businesses and the spending decisions of
households.
An essential assumption implicit in this so-called "money" view
of the transmission mechanism is that bank loans, market-intermediated
privately-issued debt such as commercial paper and corporate bonds,
and privately-held government debt can be treated as perfect
substitutes. Indeed, this assumption is embedded in the conventional
IS-LM model, where the aggregate non-money financial asset is simply
labelled "bonds" for convenience. According to the money view, the
reduction in bank loans that accompanies a reduction in reserves is of
no particular significance in itself because firms can satisfy any
unmet demand for external finance by issuing market-intermediated debt
which is indistinguishable from bank debt. For this reason, the money
view often is summarized by the proposition that bank loans are not
"special."
The opposing view of the transmission mechanism assigns a
central role to bank loans. According to this view, bank loans,
market-intermediated privately-issued debt, and government debt are
not perfect substitutes. The reduction in the volume of bank loans
that accompanies a move toward a more restrictive monetary policy is

1. David Wilcox is on the staff of the Board of Governors of the
Federal Reserve System.




Wilcox

contractionary in itself, even controlling for any associated change
in interest rates. In effect, bank loans behave as if they were a
factor of production. A reduction in their availability increases
their relative price (the spread between the loan rate and the openmarket rate increases). In response, firms seek cheaper alternatives
for their external finance. However, given the imperfect
substitutability of other forms of debt for bank loans, the reduction
in loan availability implies a contraction in real activity.
The important distinction between the money view and the loans
view is that the latter implies that the impulses of monetary policy
are transmitted not only through the overall level of interest rates,
but also through the relative prices and relative quantities of bank
loans and other forms of external finance. If the loans view is
right, fluctuations in the quantities and prices of bank loans,
commercial paper, other private debt, and government debt will be
worth keeping track of separately because they will be informative for
either the current or future state of the economy, or both. Moreover,
the loans view suggests, as Kuttner (this volume) and Friedman (1991)
emphasize, that there is no reason for being uniquely interested in
changes in the stance of monetary policy; other factors (including but
not restricted to the stringency of regulatory oversight) will also be
worthy of study to the extent that they bear on loan availability.
THE IDENTIFICATION PROBLEM
One approach to investigating the empirical significance of the loans
channel has been to regress some measure of real activity (such as
industrial production or GNP) on current and lagged measures of bank
loans. A positive correlation between bank loans and real activity
has sometimes been interpreted as contradicting the money view and
supporting the existence of a separate loans channel. The flaw in
this argument is not hard to spot: A positive correlation between
bank loans and real activity could simply reflect an endogenous
response
of the demand for bank loans to changes in real activity
rather than an exogenous cause of changes in real activity. Even a
finding of a positive correlation between bank loans and
subsequent
changes in activity (as opposed to contemporaneous ones) would not be
convincing evidence of a separate loans channel; such a phenomenon
could reflect, for example, a need to secure financing some months or




-2-

Wilcox

even quarters before the bulk of the associated activity is to take
place.
An important challange taken up in the more recent literature
has been to solve this identification problem in a convincing
manner.2
SUMMARY OF KUTTNER'S PAPER
Ken Kuttner's paper makes two important contributions to the
literature on the monetary policy transmission mechanism: one
theoretical, the other empirical.
On the theoretical front, he presents a very nice compact model
of the flow of funds in a simple economy. He distinguishes five
financial instruments in his model (in contrast to the usual two):
deposits ("money"), bank loans, commercial paper, reserves, and
government debt.
He posits the existence of a representive firm,
a representative bank, and a representative household, and endows each
of them with standard portfolio behavior (households* demand for money
is declining in the opportunity cost of holding money, and so forth).
Then he derives the implications of changes in the stance of monetary
policy, changes in banks' willingness to lend, and changes in firms'
demand for external finance for three quantities: the mix of external
finance, the spread between the loan rate and the commercial paper
rate, and the spread between the paper' rate and the Treasury bill
rate.
The beauty of Kuttner's model is that it delivers sensible
results very directly. For example, a reduction in banks' willingness
4

to lend causes the loan-paper spread to rise.

In response, firms

2. The approach proposed in Kashyap, Stein, and Wilcox (1992) is to
focus on changes in the composition
of external finance rather than
fluctuations in any one component alone. Intuitively, one would not
expect changes in the volume of bank loans relative to the volume of
other debt to be informative for current or future changes in real
activity if bank debt is a perfect substitute for non-bank debt.
3. Implicitly, other corporate liabilities such as medium- and
long-term bonds are treated as perfect substitutes for commercial
paper.
4. Kuttner interprets "negative shifts in X as 'credit crunch'
episodes." He notes, however, that a negative shift in X could
reflect a "perceived deterioration in borrowers' creditworthiness."
In my opinion, it would be more useful to reserve the term "credit
(Footnote continues on next page)




-3-

Wilcox

shift the mix of external finance away from bank loans and toward
market-mediated debt. The increased issuance of commercial paper
drives up the spread between commercial paper rates and bill
rates.
With respect to these three key variables, the effects of
a reduction in banks' willingness to lend are identical to the effects
of a move by the Federal Reserve toward a more restictive monetary
policy, suggesting that any one of the three might be useful as an
index of loan availability.
In fact, it turns out that these three variables also respond
in qualitatively the same manner to the other two exogenous factors in
Kuttner's model (monetary policy and the demand for external finance).
That is, no matter what the conceptual experiment being run in
Kuttner's model, the loan-paper spread will always move in the same
direction as the paper-bill spread, and the two spreads will always
move in the opposite direction of the mix.
In light of these predictions from his theoretical model,
Kuttner's finding that the loan-paper spread significantly
underperforms the mix and the paper-bill spread as indicators for
future real GNP is interesting and a bit puzzling. Kashyap, Stein,

(Footnote continued from previous page)
crunch" for periods in which some potential borrowers are turned away

even though, with Identical

characteristics

in every

respect

(including
"credit
worthiness"),
they would.have been granted credit
in "normal" times.
5. In Kuttner's model, the commercial paper rate is taken as the
benchmark rate over which the Federal Reserve has direct control in
the reserves market. As a result, a reduction in banks' willingness
to lend has no effect on the JeveJ of the commercial paper rate.
As was noted in the text, however, it does increase the loans-paper
spread. As a result, the volume of commercial paper outstanding
rises and the paper-bill spread increases. Given the fixity of the
paper rate in the face of this experiment, it must be that the bill
rate has declined. If the bill rate (rather than the paper rate) were
assumed to clear the market for reserves, all the essential results
still would hold (the mix would shift away from loans, the loans-paper
spread and the paper-bills spread both would rise), but the bill rate
would be fixed and the paper rate would rise.
6. Kuttner notes that the effects of a shift in monetary policy are
not identical in every respect to the effects of a shift in banks'
willingness to lend: The former affects the level of the interest
rate in the market for reserves, whereas the latter does not.




-4-

Wilcox

and Wilcox (1992) argued that the mix might be preferable to the loanpaper spread as an indicator of loan availability (because the stated
loan rate would not adequately reflect changes in non-price terms of
loan contracts such as collateral requirements), but then proceeded to
find in their sample that the predictive power of the two variables
was roughly comparable. It would be worth attempting to reconcile
Kuttner's results with those of KSW, and (assuming Kuttner's results
hold up) attempting to verify the KSW hypothesis about why the loanpaper spread might be an inferior performer.
On the empirical side, Kuttner's paper introduces a new
approach to solving the identification problem. He posits several
simple "structual vector autoregression" models of the markets for
reserves, bank loans, and commercial paper. Kuttner is bold enough to
supply sufficient prior restrictions on the specification of the
various equations, and finds that, for the most part the estimates
that follow are well in line with the predictions that were outlined
in his theoretical section. The major exception--and one that
deserves further investigation--is that increases in banks'
willingness to lend (counterintuitively) appear to cause Increases
in
interest rates.
AN ASYMMETRIC-INFORMATION-BASED ACCOUNT OF SUBSTITUTION BETWEEN LOANS
AND PAPER
In line with most of its recent predecessors, Kuttner's paper adopts
an aggregate perspective: The model is inhabited by representative
banks, households, and non-bank firms, and the empirical work is
conducted using aggregate data. As in the earlier papers, this
perspective--through no fault of the author--sets up certain tensions
of both an expositional sort and a substantive sort. On the
expositional side, the most natural way to tell the story of the loans
channel involves an appeal to heterogeneity among firms: Some are
capable of issuing commercial paper while others are not. Obviously,
a story such as this is difficult to link up directly to a model with
a single representative non-bank firm. On the substantive side, the
representative-agent approach to modelling the problem fuels the
intuition that some firms should be observed to be on the margin
between bank loans and commercial paper. The purpose of the rest of
these comments is to sketch verbally a model that allows for




-5-

Wilcox

heterogeneity among firms, and then to point out two important
implications of such an approach.
The loans view is predicated on the assertion that non-bank
debt is not perfectly substitutable for bank debt. That imperfect
substitutability can be motivated as reflecting market imperfections
that arise when borrowers have more information about their economic
prospects than do prospective lenders. Banks specialize in
"information-intensive" lending--that is, in lending to customers
(such as small businesses) for whom the asymmetric-information problem
is more acute, and hence more difficult for arms-length capital
markets to solve.
A contractionary shift in the stance of monetary policy will
cause banks to reduce the size of their loan portfolios.
Banks
will tend to cut off their most risky customers and continue to
service their most creditworthy ones. Firms that are denied credit by
banks may be unable to borrow from any other lender. Certainly, they
will not be able to issue debt in arms-length capital markets: nor
will they be able to attract financing from other non-bank sources
simply by announcing their willingness to pay a higher rate of
interest on the debt, because potential lenders will recognize that
only the riskiest firms would be willing to offer a higher rate of
return. In the end, these firms are likely to be particularly
vulnerable to the monetary contraction.
After a monetary contraction, a larger fraction of total
external finance will be provided via arms-length capital markets and
a smaller fraction through bank loans. This change in composition may
reflect either (or both) of two factors: First, it may reflect
increased issuance of trade credit by large, financially secure firms
to their smaller, less creditworthy suppliers. An increase in
commercial paper borrowing would be used, in effect, to finance the
rise in trade credit. Large firms may be willing to act, in effect.
as financial intermediaries because they will have accumulated
substantial inside information about the financial stability of their
suppliers in the course of having interacted with them before the

7. A lower level of reserves will only support a lower level of
deposits. The lower level of deposits (which comprise banks*
liabilities) implies that assets will have to decline as well. Given
that banks view loans and securities as imperfect substitutes, some of
that decline in assets will be absorbed in loans.




-6-

Wilcox

credit crunch. Alternatively, the increase in the share of commercial
paper in total external finance may reflect that large firms tend to
expand when their smaller rivals are weakened by financial stringency;
the large firms take the opportunity to seize some portion of the
product market, financing the larger scale of their operations with
the increase in commercial paper issuance.
These two mechanisms show that bank loans and commercial paper
can be substitutes at the aggregate level even though not so for any
individual firm. Failure to observe firms operating on the margin
between bank loans and other market-mediated debt does not constitute
evidence against the heterogeneous-firms version of the loans channel.
IMPLICATIONS OF THE ASYMMETRIC-INFORMATION-BASED APPROACH
The informal discussion in the previous section points to two
important implications for future research. First, the very
motivation of banks specializing in information-intensive lending
suggests that further progress probably would flow from the analysis
of models that allow for heterogeneous non-bank firms. In particular,
it seems likely that most such models will imply that, when the
Federal Reserve adopts a more restrictive monetary policy, banks will
shrink their loan portfolios by refusing credit to their riskiest
(least financially stable) customers. Commercial paper issuance will
rise because firms already issuing paper will issue more--either to
finance their own expanded operations, or to finance the passthrough
of trade credit to their suppliers.
By contrast, a
representative-firm model suggests that all firms should be on the
margin between bank debt and commercial paper, and that when the
Federal Reserve tightens we should observe a rebalancing of
liabilities taking place at the individual firm level. The
implausibility of this account is obvious, given that fewer than 1300
firms in the United States have commercial paper programs rated by
Moody's.

8. Firms that are growing in size will, at some point, find it
possible to issue commercial paper for the first time. If the
profitability of commercial paper issuance is an inverse function of
bank-loan availability, establishment of commercial paper programs
will tend to be bunched into periods immediately following tightenings
of monetary policy. Historically, of course, the commercial paper
market was not always as well-developed as it is now; as the market
deepened and became more efficient, even firms that had been large and
creditworthy for a long time established new programs.




-7-

Wilcox

The second implication of the disaggregated approach is that
future empirical work should focus on micro-level datasets. Such
investigations will be essential for: (1) establishing the identity of
bank customers who are denied credit in the wake of a tightening by
the Federal Reserve; and (2) establishing the source of the
accompanying increase in commercial paper issuance.




TABLE OF CONTENTS
VOLUME 1
Session 1. Historical Overview
Ann-Marie Meulendyke
"Federal Reserve Tools in the Monetary Policy Process in Recent
Decades"
Comments: Robert L. Hetzel
Marvin Goodfriend
"Interest Rate Policy and the Inflation Scare Problem: 1979-1992"
Comments: R. Alton Gilbert

Session 2. International Comparisons
John Morton and Paul Wood
"Interest Rate Operating Procedures of Foreign Central Banks"
Bruce Kasman
"A Comparison of Monetary Policy Operating Procedures in Six
Industrial Countries"
Comments (on both John Morton and Paul Wood and on Bruce Kasman'
Stephen A. Meyer
Robert B. Kahn and Linda S. Kole
"Monetary Transmission Channels in Major Foreign Industrial
Countries"
Comments: Craig S. Hakkio

Session 3: Time Series Econometric Issues
William Roberds, David Runkle, and Charles H. Whiteman
"Another Hole in the Ozone Layer: .Changes in FOMC Operating
Procedure and the Term Structure
Comments: Glenn D. Rudebusch
Charles Evans, Steven Strongin, and Francesca Eugeni
"A Policymaker's Guide to Indicators of Economic Activity"




Comments: Richard W. Kopcke

-2-

Session 4: Operating Issues Related to Banking
Wilbur John Coleman II, Christian Gilles, and Pamela Labadie
"Discount Window Borrowing and Liquidity"
Comments: Michael Dotsey
Kenneth N. Kuttner
"Credit Conditions and External Finance: Interpreting the Behavior
of Financial Flows and Interest Rate Spreads"
Comments: David Wilcox

VOLUME 2
Session 5: Reserve Targeting. Interest Rate Targeting, and Term
Structure Volatility
Joseph E. Gagnon and Ralph W. Tryon
"Price and Output Stability Under Alternative Monetary Policy
Rules"
Comments: Satyajit Chatterjee
Steven Russell
"Monetary Policy Experiments in a Stochastic Overlapping
Generations Model of the Term Structure"
Comments: Eric M. Leeper
Jeffrey Fuhrer and George Moore
"Inflation Persistence"
Comments: John B. Taylor
Session 6: Adapting to Regulatory Change
Allan D. Brunner and Cara S. Lown
"Implementing Short-Run Monetary Policy with Lower Reserve
Requirements"
Comments: Edward J. Stevens
John Wenninger and William Lee
"Federal Reserve Operating Procedures and Institutional Change"
Comments: Daniel L. Thornton




-3-

Session 7; Feedback Rules for Monetary Policy
John P. Judd and Brian Motley
"Controlling Inflation with an Interest Rate Instrument"
Comments: Evan F. Koenig
Gregory D. Hess, David H. Small, and Flint Brayton
"Nominal Income Targeting with the Monetary Base as Instrument: An
Evaluation of McCallum's Rule"
(with an appendix by Richard D. Porter)
Comments: Bennett T. McCallum

Summary and Overview
John B. Taylor
"New Directions in Monetary Policy Research: Comments on the
Federal Reserve System's Special Meeting on Operating Procedures"
Bennett T. McCallum
"Concluding Observations"




PRICE AND OUTPUT STABILITY UNDER ALTERNATIVE MONETARY POLICY RULES
Joseph E. Gagnon and Ralph W. Tryon

This paper is an empirical study of alternative monetary policy regimes in
the United States using stochastic simulation of the MX3 multicountry
rational-expectations macro model developed by the staff of the Board of
Governors.

We focus on the implications of interest rate smoothing and

incomplete information for the stability of prices, output, and long-term
interest rates when the monetary authority targets nominal income. We
also conduct a limited number of simulations with a modified version of
our model that incorporates staggered real price contracts in the manner
of Fuhrer and Moore (1992).

The paper builds on the methods and results

of an earlier paper (Gagnon and Tryon (1992)) that examined monetary
policy rules using stochastic simulations of the MX3 model.
There are several findings.

First, we confirm our earlier result

that the variabilities of prices and output are roughly equal whether the
monetary authority targets the monetary base or nominal income, and we obtain confidence intervals for this result.

Second, we find that interest

rate smoothing provides a significant reduction in interest rate
variability with almost no increase in the variability of price and output.

Third, it appears that random errors in the observation of the

target variable may not significantly increase the variability of price
and output.

Fourth, while staggered real price contracts tend to increase

the size and persistence of price and output deviations, they do not lead
to different conclusions about the relative effects of nominal income targetting with and without interest rate smoothing.

Finally, we find that

the variability of the long-term interest rate is much lower than that of
the short-term interest rate in all the monetary regimes studied.

1. Division of International Finance, Board of Governors of the
Federal Reserve System. We are grateful to Mark Unferth for very capable
assistance in running the simulations and preparing the tables.



Gagnon and Tryon
STOCHASTIC SIMULATION FRAMEWORK
The paper uses stochastic simulations of the MX3 multicountry model to
evaluate different monetary policy rules. MX3 is a medium-sized rationalexpectations model of the United States, Japan, Germany, and the rest of
2
the world.
We analyze the effectiveness of different monetary policy
rules in stabilizing the economy.

The policy rules are simple feedback

relations between the short-term interest rate and deviations of the target variable from its target value.

The target value in each case is the

baseline path for the target variable; the baseline path is the deterministic solution for the model.

The functional form and the parameters

of the policy rules are chosen arbitrarily at plausible values, rather
than as the solution to an optimization problem.
We simulate the MX3 model for multiple replications of each rule,
using random shocks drawn from a joint normal distribution using the estimated covariance matrix of the model residuals for the period 1976-88.
The simulation range for each replication is over 20 quarters, from 1989
through 1993.

3

The baseline path is the simulation over the same period

without any stochastic shocks, converging toward a steady state.

For each

replication, we calculate the deviation of each variable of interest, including the (log) levels and growth rates of income, prices, and interest
rates, from the baseline values.

The root-mean-squared deviation (RMSD)

across replications is calculated for each rule; this measure of
variability is compared across rules.
Comparison of rules
Using the RMSD to compare different rules implies that the monetary
authority's objective is stated in terms of the second moments, rather
than the first moments, of the data.

The choice of this objective

reflects our conviction that the average levels of real economic variables
are invariant to any well-specified monetary policy rule in the long run.

2. For a description of the theory and estimation of the model, see
Gagnon (1991). We do not believe that any of our results are strongly
dependent on the use of a multicountry model rather than a purely domestic
model. Nonetheless, it is a property of the model that foreign responses
to U.S. shocks can have feedback effects on the United States through the
exchange rate and the trade balance.
3. For a description of the historical residuals and their calculation, see Gagnon and Tryon (1992).



- 2-

Gagnon and Tryon
Although nominal variables do depend on monetary policy, this study ignores the factors involved in choosing a long-run inflation rate and
focuses solely on deviations from the long-run rate.
The use of second moments as measures of economic performance may
be rationalized on two grounds.

First, fluctuations of variables around

their expected values give rise to adjustment costs as agents adapt their
behavior to the new conditions.

Second, agents may be risk averse, so

that their utility is increased when monetary policy succeeds in reducing
the variance of an important variable.

Of course it is possible that, by

reducing the variance of one variable, policy may increase the variance of
some other variable.

In conducting the analysis it is necessary to con-

sider all of the most important variables.

Implicitly or explicitly,

policymakers may have to weigh stabilization of one variable against the
destabilization of another.
The transition from one policy regime to another is likely to involve significant costs as agents learn gradually about the new regime.
It would be of interest to consider the problem of making such a regime
shift less costly, but we do not pursue that topic here.

The assumption

behind all the stochastic simulations in this paper is that the regime
shift is understood perfectly by the private sector and is fully credible.
Thus, comparisons of economic performance across policy regimes reflect
differences in the long-run stochastic behavior of the economy and not the
short-run transition costs.
Number of replications
To begin a stochastic simulation, residuals are drawn for one period from
a normal distribution with mean zero and the estimated historical
variance-covariance matrix.

The model is solved in 1989Q1 by using these

residuals and the fixed lags and exogenous variables.
tions are computed by the Fair-Taylor algorithm.
assumed to be z ero.

The future expecta-

Future residuals are

The stochastic solution for 1989Q1 is then used for

the necessary lags in solving 1989Q2.

In solving 1989Q2, a new draw of

residuals is taken from their estimated distribution, but future residuals
are again assumed to be zero.

This process is repeated for twenty

quarters, thus completing one stochastic replication over the baseline




Gagnon and Tryon
period.

Twenty stochastic replications are conducted for each policy
4
rule, for a total of 400 draws of the residuals.
In order to make more accurate comparisons across policy rules, we
repeated the same sequence of stochastic shocks for each rule.

Somewhat

to our surprise we found that differences in the computed RMSDs across
replications for the same rule were two orders of magnitude greater than
differences in the RMSDs across policy rules for the same replication.

To

test whether the RMSD under one rule differed significantly from the RMSD
under another rule, we computed the difference between the two RMSDs for
each replication.

Assuming that these differences are normally dis-

tributed, we were able to test the null hypothesis that their mean value
is zero, i.e. that there is no difference in the stability of the given
variable under either policy rule.
Because of the adjustment lags present in many equations in MX3,
the random shocks tend to have persistent effects.

Within a given

replication, as shocks are drawn in successive periods their effects are
combined with the gradually declining effects of earlier shocks.

Since

each replication begins without the effects of any lagged shocks, the RMSD
of most variables increases over the first few years of a replication.

In

order to focus on the long-run stability of the variables, we computed
RMSDs for only the last two years (eight quarters) of each five-year
replication.
There were four replications for which we were unable to obtain a
solution in at least one period for at least one policy rule.

In order to

obtain 20 complete replications with identical shocks under all rules, we
simulated the model for a total of 24 sets of stochastic shocks.
Difficulty in solving the model for a particular set of shocks is clearly
not independent of the nature of those shocks.
more likely to lead to solution problems.

Unusually large shocks are

Thus, our results may be biased

by the exclusion of those replications that could not be solved under all
policy rules.

4. These replications were conducted with TROLL 13.1 software using
the stochastic simulator package. Each replication requires about 40
minutes of processing (CPU) time on an Amdahl 5850.
5. The RMSDs of a few variables continued to increase mildly over the
fourth year of the replications, but we did not believe that the increase
was strong enough to warrant discarding an additional year of data.



- 4-

Gagnon and Tryon
The number of replications run for each rule was determined
primarily by limits on available computer time.

In order to gauge the

significance of our results, we report confidence intervals and t-tests
calculated on the assumption that the MX3 model is approximately lognormal.

This assumption was tested using 100 replications of a single

period stochastic disturbance; the deviations from baseline for all variables of interest were checked for normality using the Jarque-fiera test
statistic.

We were unable to reject the null hypothesis of normality at

the 952 confidence level for all variables except consumption and the real
exchange rate.
To economize on computation time, the Fair-Taylor algorithm was allowed only one type-III iteration over a forecast horizon of twenty
quarters.

The type-II convergence criterion used was 0.02 percent.

In

most cases type-II convergence was achieved, but sometimes the solution
stops at the iteration limit of 100. A series of test solutions indicated
that these restrictions allow reasonably accurate results.
MONETARY BASE AND NOMINAL INCOME TARGETTING
We begin with two simple alternatives, monetary base and nominal income
targetting:
(1)

RSt

- RS* - 1.5 [log(MBt)

(2)

RSt - RS*t - 1.5 [log(GDPVt)

- log(tfB*)]

- log(GDPV^)]

where RS is the short-term interest rate in decimal form at an annual
rate; MB is the monetary base; and GDPV is nominal GDP. Asterisks denote
target values.

In rule (1) the monetary authorities target the monetary

base; in rule (2) the monetary authorities target nominal GDP. Unlike in
our previous work, in this paper we implement the monetary policy reaction
functions only for the United States:

in the other regions monetary

policy is assumed to hold the monetary base fixed (on its baseline path).
Columns 1 and 2 of Table 1 summarize the results of the stochastic
simulations for these rules. The table shows the root-mean-square deviation (RMSD) of each variable from its baseline path over the period
1992:1-1993:4 averaged over the 20 replications.




- 5-

Below each RMSD is a 90

Gagnon and Tryon
percent confidence interval for the true RMSD based on a sample of 20 independent observations.

PGDP is the GDP deflator; CAB is the current

account balance divided by nominal GDP; REFW is a weighted real exchange
rate; RL is the long-term interest rate; C is real private consumption;
and MBR is the monetary base deflated by the consumption deflator.

The

variables are measured in logarithms, except for the interest rates and
the ratio of the current account balance to nominal GDP, which are decimal
fractions.

We also report statistics for the first differences (quarterly

growth rates) of some of these variables.

Column 1 of Table 2 shows the

mean values of the differences in RMSDs for these two policy rules. An
asterisk denotes that the difference is significantly different from zero
at the 10 percent level; two asterisks denote significance at the 5 percent level.
The most striking aspect of these results is that the monetary base
and nominal income rules are very similar.

As expected, the RMSD of

nominal GDP is lower with nominal income targetting (by 0.9 percentage
points), and the RMSD of the monetary base is lower when it is targetted
(by 0.4 percentage points).

(The monetary base rule does not require that

the money base target be met exactly, so there is partial accomodation of
money demand shocks in this case.)

In each case the RMSD of real GDP from

baseline is about 6 percentage points, while the RMSD in the growth rate
of real GDP is around 2 percentage points. The variability of the growth
rate is significantly lower for the nominal income rule, but the magnitude
of the difference is slight (0.2 percentage points).

The variability of

the price level and inflation is essentially the same under both regimes.
Thus, the reduction in variability of nominal income is not passed through

6. The confidence intervals are computed under the assumption that the
deviations of each variable from baseline are normally distributed. For
normal deviations the sample variance follows a chi-square distribution.
Although the sample contains 160 observations, we allowed for only 20
degrees of freedom in computing our confidence intervals because the
deviations within each of the 20 replications were highly autocorrelated.
Our intervals are therefore larger than true 90 percent confidence
intervals.
7. The discussion of our results focuses on the RMSDs for output,
prices, and interest rates. The real exchange rate and the current
account balance are presented as macro variables of general interest.
Consumption and real money balances are included because they represent an
alternative pair of variables that the monetary authority might wish to
stabilize. Generally speaking, consumption variability is highly
correlated with output variability. The variability of the real monetary
base is less easy to characterize.




- 6-

Gagnon and Tryon
to its components (real output and prices); instead, this regime effectively exploits offsetting variations in the components to meet the target
for nominal income.
There is a significant difference in the variability of the interest rate: under nominal income targetting the RMSD of the short-term
interest rate is higher by 117 basis points and the variability of its
first difference is 175 basis points higher.

This increased variability

is passed through to the long-term interest rate and the real exchange
rate.
o

These results are consistent with our earlier findings.

A

priori, one might expect nominal income targetting to stabilize prices and
output better than monetary base targetting, because (log) nominal income
is simply an equal-weighted average of the price level and output in
logarithms.

The similarity of nominal income and monetary base targetting

implies that money demand shocks in MX3 are not too large relative to
other disturbances.

Since money demand in MX3 is roughly proportional to

nominal income, and the adjustment lag is relatively short, it is perhaps
not surprising that these two rules have similar effects on prices and
output.
INTEREST RATE SMOOTHING
An important feature of feedback rules of the form used in (1) and (2) is
that they can lead to "instrument instability," i.e., substantial variation in short-term interest rates from one period to the next.

This is

not necessarily a theoretical problem, since the interest rate need not
enter directly into private agents' utility or the monetary authority's
objective function.

However, excessive interest rate (or exchange rate)

volatility is sometimes viewed as undesirable in and of itself.

To ad-

dress this question, we consider a variation of the nominal income rule

8. These results are also robust to changes in the feedback coefficient by a factor of two or three. Gagnon and Tryon (1992) show that
increasing the feedback coefficient reduces the RMSD of the target
variable and increases the RMSD of the short-term interest rate. A higher
feedback coefficient on nominal income does not reduce the RMSD of the
price level or output separately.




- 7-

Gagnon and Tryon
that smooths fluctuations in the short-term interest rate.

The rule is

calibrated so that a persistent deviation in nominal GDP will provide the
same interest rate response as rule (2) in the long run, but not in the
short run:
(3)

RSt

- RS*t - 0.8 ks t .2 - ^ . j ]
+ (1-0.8) 1.5 [logiGDPVJ

- log(GDPV*)J

The results are shown in column 3 of Table 1 and columns 2 and 3 of
Table 2.

The addition of interest rate smoothing to the nominal income

rule reduces the RMSD of the change in the short-term interest rate by 200
basis points.

This reduction is not at the expense of any importanc in-

crease in volatility in the targets; the RMSD of nominal income rises by
only 0.2 percentage points; the changes in the variability of real output
and prices are of the same order of magnitude.

Nominal income targetting

combined with interest rate smoothing produces almost exactly the same
results as the monetary base rule, as shown in column 2 of Table 2.
The ability of interest rate smoothing to dampen fluctuations in
interest rates without substantial increases in the RMSDs of other variables is noteworthy.

We believe that this result is due to two basic

properties of the MX3 model.

First, adjustment lags are quite large

throughout the model, implying that the short-run response of the model to
monetary policy is much less than the long-run response.

Second, the be-

havioral equations are forward-looking, so that future monetary policy has
a strong impact on current behavior.

If the interest smoothing rule is

credible, agents believe that a sustained upward shock to nominal income
will cause the monetary authority to initiate a series of increases in the
short-term interest rate.

These expected future increases in the interest

9. Alternatively, this rule could be motivated by a desire to include
lagged information in the target. In a model with adjustment lags it is
optimal in principle to react to both current and lagged shocks. However,
it is infeasible to compute the optimal response pattern to lagged
information in a model of this size.
10. This result appears to be robust with respect to the parameters of
the smoothing rule. We performed a limited number of trials with
different feedback coefficients on the lagged interest rate and nominal
GDP; we found that increasing either coefficient always yielded a smaller
RMSD of the associated variable at the expense of the other variable. The
RMSDs of real output and prices were unaffected in these trials.



- 8-

Gagnon and Tryon
rate will dampen current consumption and investment more than if agents
were not forward-looking.
OBSERVATION ERRORS
Rules (4) and (5) incorporate observation error into rules (2) and (3).
These rules are motivated by the fact that the monetary authority cannot
accurately observe current nominal GDP. Many components of nominal GDP
are observable contemporaneously, but many others are measured only with a
lag.

We postulate that the monetary authority uses the contemporaneously

available indicators of nominal GDP to make an unbiased estimate of current nominal GDP.

The contemporaneous indicators may include--but are

not limited to--asset prices, commodity prices, interest rate spreads, and
in-house surveys of business activity.

The error in estimated nominal GDP

is captured by c, which is assumed to be normally distributed with zero
mean and no autocorrelation.

(4)

RSt - RS*t - 1.5 [log(GDPVc) + et - logiCDPV^)]

(5)

RSt

- RS*t - 0.8 [*Stml

- *S*.2]

+ (1-0.8) 1.5 [log(GDPVt)

+ € t - log(GDPV*)J

To estimate the magnitude of the monetary authority's observation
error, we calculated the difference between the consensus forecast for
current quarter nominal GDP growth published in Blue Chip
Indicators

Economic
12
and BEA's final estimate of nominal GDP growth.
The standard

deviation of the error (in logarithms) is 0.0068, or 2/3 of a percent.
The results are shown in the fourth and fifth columns of Tables 1
and 2.

The addition of observation error degrades the ability of the

authority to control the target variable, and without interest rate
smoothing, the RMSD of nominal income increases by 0.4 percentage points.

11. We believe that the interaction between the private sector and the
monetary authority should be modeled symmetrically: if the private sector
can respond simultaneously to innovations in the monetary instrument, then
the monetary authority should be allowed to react simultaneously to
innovations in private variables. We conjecture that dynamic instability
of macroeconomic models under some policy rules may be due specifications
that do not allow the monetary authority to respond to any contemporaneous
information.
12. The forecasts were published at the beginning of the third month of
the quarter and were based on information collected and analyzed during
the second month. The sample period was 1980:1 through 1988:4.




- 9-

Gagnon and Tryon
The RMSD of the short-term interest rate rises by about 70 basis points,
and the RMSD of the real exchange rate also rises, by about 1.5 percentage
points.

However, none of these differences is statistically significant.

With interest rate smoothing, there is virtually no difference between the nominal income rules with and without observation error.

This

is because with smoothing, the authorities do not respond nearly as much
to temporary shocks, and the impact of observation errors is correspondingly reduced.

The existence of observation error thus strengthens the

case for interest rate smoothing.

As in the standard signal extraction

problem, the monetary authority's optimal response to an innovation in the
target variable is reduced by the presence of noise in the observation.
Because the effect of an observation error is much less persistent than
the effect of a structural disturbance, a partially delayed response of
monetary policy helps to filter out the effect of noise on the policy instrument.
REAL PRICE CONTRACTS
In another paper presented at this conference, Fuhrer and Moore argue that
U.S. macroeconomic time series are better modeled with staggered price
contracts in real, as opposed to nominal,

terms.

In particular, Fuhrer

and Moore provide evidence that the U.S. inflation rate is much more persistent in the face of shocks than can be explained by staggered nominal
price contracts.

We wanted to explore the implications of real price con-

tracts in the MX3 model.

We were especially interested to see whether our

conclusions about nominal income targetting with and without interest rate
smoothing are robust to this alternate specification of the model. The
real contracting model is described in equations (6)-(8).
(6)

log(PCTPt) - ajlogCXp + a2logUrt_2) + c ^ l o g U ^ ) + c ^ l o g U ^ )

(7)

log(Vt) - o1log(Xt/PGDPt) +
+ a3log(Xt_2/PGDPt_2)

(8)

<*2log(Xtml/PGDPtl)
+

ailog{Xt3/PGDPt3)

logU t /PGDP t ) - c^logd^) + « 2 l o g ( V l )

+ a

3log(V2)

+ a 4 log(y t+3 ) + 7(ajlog(CUt) + * 2 log<a/ t+I )
+ a3log<0/t+2) + a4log(07t+3))
The GDP deflator, PGDP, is a geometric average of the contract
prices, X, that are still in effect in the current period.




- 10 -

Equation (6)

Gagnon and Tryon
implies that the longest contract price lasts for four quarters. The
coefficients a,- a, sum to unity and equal the proportion of contracts
outstanding that were negotiated at times t, t-1, t-2, and t-3, respectively.

V is the average real contract price currently in effect.

Equation (8) states that the current real contract price depends on the
expected future real contract price as well as the expected future level
of capacity utilization, CU.

The parameter 7 reflects the sensitivity of

the real contract price to excess demand.
We tried to run stochastic simulations of the model using the coefficient values estimated by Fuhrer and Moore, however, we were unable to
complete a single replication with those coefficient values.

We were able

to complete 20 replications using the coefficients in MX3's nominal price
contract equations.

The problem appears to be associated with the coeffi-

cient -y, which is nearly two orders of magnitude larger in MX3 than in
Fuhrer and Moore.
The results are displayed in columns 6 and 7 of Tables 1 and 2.
The RMSDs of all variables are larger with real price contracts than with
nominal price contracts.

Although the difference is sometimes quite

large, it is never significant.

Because of our uncertainty about the ap-

propriate coefficients to use, we choose not to focus on the effect of
real price contracts per se. We are interested, however, in the comparison of policy rules with and without interest rate smoothing when
contracts are written in real terms.

Because inflation is more persistent

with real price contracts, we were concerned that interest rate smoothing
might prove to be destabilizing under real contracts even though it is not
destabilizing under nominal contracts.

Column 7 of Table 2 demonstrates

that this concern appears unwarranted.

Only the nominal monetary base

shows any large increase in variability, and that increase is not statistically significant.
actually declines.

Moreover, the RMSD of the real monetary base
The short-term interest rate exhibits a large and sig-

nificant decrease in variability under interest rate smoothing, just as it
did with nominal price contracts.
LONG-TERM INTEREST RATES
It is of some interest to understand the implications of various monetary
policy rules for the behavior of long-term interest rates.

If long-term

interest rates are determined simply by expectations of future short-term
interest rates, we should find that long-term rates are less variable than




- 11 -

Gagnon and Tryon
short-term rates.

This conclusion follows from that fact that future

shocks are expected to equal zero and that the model is expected to
gradually return to baseline in the absence of shocks.

In order for long-

term rates to be more variable than short-term rates, the model would have
to allow for permanent shocks to the inflation rate or to the real interest rate.

(Alternatively, we could incorporate an ad hoc risk premium in

the long-term interest rate.)
The basic MX3 model includes only a one-period interest rate.

For

this paper, the model was modified to define a long-term interest rate,
modeled as an exponentially declining weighted sum of expected future
short-term interest rates:
CO

(9)

RL - (1-7) 2 7 1 E [USt+1] - (l-7> RSt + 7 E t [ ^ t + I l
i-0

The weights were chosen to approximate a ten-year bond (7 - 0.975, at a
quarterly rate).

The long-term interest rate does not enter into any

other equation in the model, and there is no stochastic term in this
definition.
The results for the long-term interest rate are shown in the last
two rows of both tables. As expected from the experimental design, the
variability of the long-term rate is in all cases substantially less than
the short-term rate.
CONCLUSIONS
This paper contains three main findings.

First, in the MX3 model the

variabilities of prices and output are roughly equal whether the monetary
authority targets the monetary base or nominal income.

Second, interest

rate smoothing provides a significant reduction in interest rate
variability with almost no increase in the variability of prices and output, and this conclusion is not affected by a modification of the model
that increases the persistence of the inflation rate.

Third, it appears

that random errors in the observation of a nominal income target may not
significantly increase the variability of prices and output.

13. The consumption and investment equations depend on the long-term
interest rate implicitly as a function of expected future short-term
rates.




- 12 -

Gagnon and Tryon
The result that monetary base and nominal income targetting are
bi^adly equivalent is robust within the framework of our model, but may
not be entirely conclusive.

A priori, there are reasons to prefer nominal

income targetting, because nominal income is closer to the ultimate goals
of policy than is the money supply.

This paper does not resolve the issue

of the optimal target variable(s) for monetary policy, nor does it derive
optimal coefficients for the rules considered here.

Calculation of op-

timal rules would be prohibitively expensive with our model, and the
conclusions would be particularly sensitive to specification and estimation errors in the model.
We believe that our other two findings are robust to the specification of the model and policy rules. The ability of the monetary authority
to smooth interest rates to a significant extent without destabilizing
other variables depends mainly on the existence of forward-looking agents
and adjustment costs in economic activity.

The result that observation

error does not significantly affect the outcomes under different rules
depends on adjustment lags and on the relatively small size of the observation error.

Since we were able to measure the observation error

directly, we have a high degree of confidence in our finding that it does
not destabilize real output and prices significantly.




- 13 -

Gagnon and Tryon

1.

Summary of Stochastic Simulations of Policy Rules

Rule 1:
Variables

m

Rule 2:
GDPV

Rule 3:

Rule 4:

RS & GDPV

GDPV & e

GDP

0.0730
(0.058-0.098)

0.074
(0.060-0.100)

0.075
(0.060-0.102)

0.069
(0.055-0.094)

AGDP

0.022
(0.018-0.030)

0.020
(0.016-0.027)

0.022
(0.017-0.029)

0.021
(0.017-0.028)

PGDP

0.087
(0.069-0.117)

0.086
(0.069-0.117)

0.084
(0.067-0.113)

0.087
(0.069-0.117)

APGDP

0.014
(0.011-0.020)

0.014
(0.011-0.019)

0.014
(0.011-0.018)

0.014
(0.011-0.019)

GDPV

0.037
(0.030-0.050)

0.027
(0.021-0.036)

0.029
(0.023-0.039)

0.035
(0.028-0.048)

MB

0.020
(0.016-0.030)

0.023
(0.019-0.032)

0.031
(0.025-0.042)

0.030
(0.024-0.041)

CAB

0.004
(0.003-0.005)

0.003
(0.003-0.005)

0.003
(0.003-0.005)

0.004
(0.003-0.005)

RS

0.030
(0.023-0.040)

0.040
(0.032-0.054)

0.026
(0.020-0.035)

0.053
(0.042-0.071)

ARS

0.010
(0.008-0.013)

0.028
(0.022-0.038)

0.007
(0.006-0.010)

0.034
(0.027-0.046)

RL

0.004
(0.003-0.006)

0.005
(0.004-0.007)

0.004
(0.003-0.006)

0.007
(0.006-0.010)

ARL

0.002
(0.001-0.002)

0.002
(0.002-0.003)

0.001
(0.001-0.002)

0.003
(0.003-0.005)

RERW

0.130
(0.103-0.175)

0.130
(0.104-0.177)

0.130
(0.104-0.176)

0.157
(0.125-0.213)

ARERW

0.058
(0.046-0.079)

0.063
(0.050-0.085)

0.060
(0.048-0.082)

0.072
(0.057-0.097)

C

0.061
(0.049-0.083)

0.062
(0.050-0.084)

0.623
(0.050-0.084)

0.067
(0.054-0.091)

AC

0.016
(0.012-0.021)

0.015
(0.012-0.021)

0.015
(0.012-0.021)

0.016
(0.013-0.022)

MBR

0.077
(0.061-0.104)

0.087
(0.069-0.118)

0.078
(0.062-0.106)

0.090
(0.072-0.122)

AMBR

0.013
(0.010-0.018)

0.015
(0.012-0.020)

0.013
(0.011-0.018')

0.016
(0.013-0.022)




- 14 -

Gagnon and Tryon

Summary of Stochastic Simulations of Policy Rules (cont'd)
Root-mean-squared deviation from baseline
Rule 5:
Rule 6:

Rule 7:

Variables

GDPV & RS & e

GDPV & FM

GDP

0.075
(0.060-0.101)

0.134
(0.107-0.182)

0.139
(0.111-0.188)

AGDP

0.022
(0.017-0.029)

0.025
(0.020-0.034)

0.028
(0.022-0.037)

PGDP

0.840
(0.067-0.113)

0.161
(0.128-0.218)

0.156
(0.124-0.211)

APGDP

0.014
(0.011-0.018)

0.026
(0.021-0.036)

0.027
(0.021-0.036)

GDPV

0.029
(0.023-0.039)

0.047
(0.038-0.064)

0.063
(0.050-0.085)

MB

0.031
(0.024-0.041)

0.035
(0.028-0.048)

0.058
(0.047-0.079)

CAB

0.003
(0.002-0.005)

0.004
(0.003-0.006)

0.004
(0.003-0.005)

RS

0.025
(0.020-0.034)

0.071
(0.056-0.096)

0.056
(0.045-0.076)

ARS

0.008
(0.006-0.010)

0.031
(0.025-0.042)

0.013
(0.011-0.018)

RL

0.004
(0.003-0.006)

0.010
(0.008-0.013)

0.010
(0.008-0.013)

ARL

0.001
(0.001-0.002)

0.003
(0.002-0.004)

0.003
(0.002-0.003)

RERW

0.132
(0.110-0.178)

0.153
(0.122-0.207)

0.164
(0.131-0.222)

ARERW

0.061
(0.048-0.082)

0.064
(0.051-0.087)

0.062
(0.050-0.084)

C

0.062
(0.049-0.083)

0.101
(0.080-0.136)

0.101
(0.081-0.137)

AC

0.016
(0.012-0.021)

0.020
(0.016-0.027)

0.021
(0.017-0.028)

MBR

0.076
(0.061-0.104)

0.157
(0.125-0.212)

0.141
(0.112-0.191)

AMBR

0.013
(0.010-0.018')

0.027
(0.021-0.036')

0.024
(0.019-0 .OSS')




- 15 -

GDPV & RS & FM

Gagnon and Tryon

2.

Differences In RMSDs across Policy Rules

GDP

-0.0011

-0.0022

-0.0011

0.0026

AGDP

0.0018*

0.0004

-0.0014*

-0.0006

PGUP

0.0009

0.0029

0.0020

-0.0001

APGDP

0.0004

0.0006

0.0002

-0.0001

GDPV

0.0092

0.0074

-0.0018

-0.0039

-0.0036

-0.0106

-0.0070

-0 0042

0.0001

0.0001

-0.0001

-0.0002

RS

-0.0117

0.0026

0.0143*

-0.0067

ARS

-0.0175**

0.0025

0.0200**

-0.0059

RL

-0.0008

-0.0002

0.0006

-0.0009

ARL

-0.0007**

-0.0001

0.0006**

-0.0003

RERW

-0.0016

-0.0005

0.0011

-0.0148

ARERW

-0.0049*

-0.0024

0.0026

-0.0057

C

-0.0007

-0.0008

-0.0001

-0.0020

0.0003

0.0001

-0.0003

-0.0008

MBR

-0.0086

-0.0001

0.0085

-0.0010

AMBR

-0.0017

-0.0002

0.0015

-0.0012

MB
CAB

AC

* significant at 10 percent level,
** significant at 5 percent level.




- 16 -

Gagnon and Tryon

2.

Differences In BMSDs across Policy Rules (cont'd)

Variables

Rule 3-Rule 5

Rule 2-Rule 6

Rule 6-Rule 7

GDP

0.0003

-0.0497

-0.0044

AGDP

0.0001

-0.0047

-0.0021

PGDP

0.0001

-0.0626

0.0031

LPGDP

-0.0000

-0.0104

0.0000

GDPV

-0.0000

-0.0172

-0.0089

MB

0.0003

-0.0093

-0.0213

CAB

0.0000

-0.0008

0.0004

RS

0.0002

-0.0259

0.0171*

-0.0004

-0.0028

0.0182**

RL

0.0002

-0.0035

0.0009

LRL

0.0002

-0.0006

0.0005

RERW

-0.0011

-0.0169

-0.0045

LRERU

-0.0003

-0.0008

0.0020

C

0.0006

-0.0320

0.0000

LC

0.0000

-0.0037

-0.0007

HBR

0.0014

-0.0586

0.0148

-0.0099

0.0024

&RS

MBR

.

o.oooi

* significant at 10 percent level.
** significant at 5 percent level.




- 17 -

Gagnon and Tryon
APPENDIX: Some Detailed Results
Tables Al through A5 present more detail on the results of our stochastic
simulations for a small subset of the policy rules.

The top panel of

Table Al shows the sample autocorrelations of the deviations of several
variables from their baseline values under nominal income targetting (Rule
2).

All of the variables have a high degree of autocorrelation in the

levels, but none of them have highly autocorrelated growth rates (first
differences).

This autocorrelation is due to the presence of adjustment

lags in most of the model's equations.

The autocorrelation is par-

ticularly high for the price level, real money balances, and output.
The bottom panel of Table Al presents statistics relating to the
sample distribution of the deviations.

These statistics were computed
14
from 100 replications of a one-quarter solution.
Of the variables
tested, only the weighted real exchange rate and consumption reject our

null hypothesis of normality.
Table A2 displays the RMSDs for each quarter, computed across 20
replications.

For the levels of the variables, the RMSDs tend to increase

over successive quarters.

In most cases the RMSD appears to stabilize

after 12 quarters, but there is still some slight increase in the RMSDs
for real output and the price level after 12 quarters.

For the dif-

ferences of the variables, the RMSDs appear quite stable over time. The
average RMSDs differ from those presented in Table 1 because they were
computed using observations from all 20 quarters and 20 replications,
rather than the last eight quarters of 20 replications.

Table A3 contains

the RMSDs for each replication, computed across 20 quarters.

This table

shows how different the results were for each set of stochastic shocks.
Tables A4 and A5 show that the differences in RMSDs across rules
are quite small, despite the large differences in RMSDs across quarters
and across replications that are documented in Tables A2 and A3.

14. The model was solved for only one quarter because the persistence
of deviations in the model implies that the distribution of a variable in
a given quarter is dependent on the number of quarters that have been
simulated stochastically prior to the given quarter. Hence, if we had
included observations from different quarters we would have been sampling
from populations with different distributions. As the number of
stochastically simulated quarters increases, the distributions of the
observed deviations should approach a stationary distribution.



- 18 -

Gagnon and Tryon
Al.

Properties of Simulated Deviations from Baseline
Rule 2: GDPV
First through Fifth Order Autocorrelation

Variables
GDP
LGDP
PGDP
LPGDP
GDPV

MB
CAB
RS
ARS
RL
ARL
RERW
LRERW

C

AC
MBR
AMBR

1
0.976
-0.161
0.997
0.862
0.618
0.617
0.867
0.618
0.230
0.798
-0.108
0.775
0.269
0.964
0.015
0.997
0.592

2

4

3

0.962
0.201
0.991
0.704
0.265
0.062
0.721
0.265
-0.247
0.708
-0.076
0.506
-0.042
0.925
-0.125
0.992
0.432

5

0.963
0.023
0.979
0.650
0.315
-0.274
0.591
0.315
0.130
0.848
-0.399
0.504
-0.181
0.943
-0.322
0.988
0.478

0.966
0.094
0.983
0.640
0.240
-0.337
0.605
0.240
0.076
0.722
0.215
0.325
-0.059
0.943
-0.050
0.989
0.346

0.950
0.189
0.979
0.723
0.396
-0.409
0.487
0.396
0.077
0.730
0.116
0.595
-0.186
0.911
-0.042
0.986
0.579

Test of Normality'
Variables
GDP
PGDP

MB
RS
RL
CAB
RERW

C
MBR

Statistic

P-value

2.64
2.69
3.23
2.93
1.70
1.51
8.16*
9.39*
0.28

(.733)
(.740)
(.801)
(.769)
(.573)
(.530)
(.983)
(.991)
(.128)

Skewness

Excess Kurtosis

-.396
-.136
.410
-.383
-.027
-.229
.697
-.745
.114

* Reject normality at 5 percent significance level.

Autocorrelations computed from 20 replications of 20 quarters.
Normality tests based on 100 replications of one quarter.




- 19 -

-.075
-.756
-.320
-.342
-.637
-.391
.120
.191
.120

Gagnon and Tryon
A2.

RMSD by Quarter Across 20 Replications
Rule 1: MA

GDP
AGDP
PGDP
APGDP
GDPV
MB
CAB
RS
ARS
RL
ARL
RERW
ARERW
C
AC
MBR
AMBR

GDP
AGDP
PGDP
APGDP
GDPV
MB
CAB
RS
ARS
RL
ARL
RERW
ARERW
C
AC
MBR
AMBR




89 1

89 2

89 3

89 4

90 1

90 2

90 3

90 4

0.0179
0.0185
0.0046
0.0068
0.0186
0.0037
0.0019
0.0055
0.0086
0.0012
0.0014
0.0513
0.0552
0.0134
0.0134
0.0071
0.0055

0.0154
0.0185
0.0105
0.0068
0.0121
0.0058
0.0019
0.0086
0.0086
0.0013
0.0014
0.0552
0.0552
0.0123
0.0134
0.0083
0.0055

0.0234
0.01670.0195
0.0101
0.0181
0.0085
0.0021
0.0127
0.0097
0.0019
0.0015
0.0870
0.0606
0.0233
0.0170
0.0149
0.0094

0.0375
0.0204
0.0291
0.0106
0.0343
0.0125
0.0028
0.0188
0.0107
0.0023
0.0017
0.1141
0.0590
0.0356
0.0179
0.0222
0.0096

0.0280
0.0207
0.0384
0.0104
0.0289
0.0132
0.0020
0.0198
0.0086
0.0019
0.0015
0.1102
0.0642
0.0338
0.0145
0.0278
0.0105

0.0359
0.0192
0.0455
0.0095
0.0265
0.0150
0.0020
0.0225
0.0083
0.0015
0.0014
0.0921
0.0545
0.0341
0.0145
0.0326
0.0095

0.0442
0.0160
0.0495
0.0100
0.0301
0.0137
0.0028
0.0206
0.0085
0.0022
0.0019
0.1075
0.0565
0.0423
0.0150
0.0396
0.0089

0.0487
0.0254
0.0531
0.0119
0.0286
0.0154
0.0032
0.0231
0.0110
0.0029
0.0018
0.1085
0.0698
0.0449
0.0183
0.0435
0.0095

91 1

91 2

91 3

91 4

92 1

92 2

92 3

92 4

0.0443
0.0235
0.0577
0.0125
0.0314
0.0157
0.0035
0.0235
0.0100
0.0025
0.0013
0.1312
0.0722
0.0434
0.0143
0.0479
0.0104

0.0536
0.0236
0.0638
0.0119
0.0336
0.0173
0.0030
0.0260
0.0096
0.0024
0.0017
0.1363
0.0562
0.0491
0.0183
0.0545
0.0115

0.0582
0.0244
0.0690
0.0109
0.0338
0.0162
0.0030
0.0243
p.0075
0.0028
0.0015
0.1342
0.0586
0.0529
0.0190
0.0597
0.0119

0.060C
0.0194
0.0736
0.0104
0.0313
0.0157
0.0024
0.0236
0.0092
0.0026
0.0016
0.1449
0.0448
0.0541
0.0122
0.0643
0.0079

0.0666
0.0197
0.0757
0.0113
0.0319
0.0151
0.0027
0.0226
0.0090
0.0035
0.0017
0.1335
0.0486
0.0596
0.0154
0.0684
0.0098

0.0656
0.0233
0.0773
0.0131
0.0310
0.0147
0.0029
0.0220
0.0103
0.0037
0.0015
0.1438
0.0509
0.0580
0.0126
0.0704
0.0113

0.0687
0.0208
0.0793
0.0144
0.0353
0.0165
0.0030
0.0247
0.0107
0.0043
0.0012
0.1200
0.0627
0.0587
0.0148
0.0731
0.0133

0.0671
0.0282
0.0816
0.0139
0.0316
0.0198
0.0028
0.0297
0.0090
0.0042
0.0014
0.1297
0.0508
0.0589
0.0153
0.0736
0.0103

Gagnon and Tryon
A2.

RMSD by Quarter Across 20 Replications (cont'd)
Rule 1: MA

GDP
AGDP
PGDP
APGDP
GDPV
MB
CAB
RS
ARS
RL
ARL
RERW
ARERW
C
AC
MBR
AMBR




93 1

93 2

93 3

93 4

Avg

0.0688
0.0142
0.0847
0.0154
0.0366
0.0209
0.0034
0.0314
0.0085
0.0045
0.0014
0.1340
0.0542
0.0570
0.0145
0.0728
0.0138

0.0714
0.0259
0.0908
0.0160
0.0391
0.0219
0.0044
0.0328
0.0110
0.0038
0.0020
0.1331
0.0678
0.0563
0.0178
0.0762
0.0172

0.0819
0.0222
0.0963
0.0138
0.0417
0.0222
0.0045
0.0333
0.0085
0.0039
0.0017
0.1280
0.0623
0.0674
0.0175
0.0839
0.0151

0.0875
0.0181
0.1032
0.0154
0.0453
0.0231
0.0039
0.0347
0.0112
0.0043
0.0014
0.1044
0.0637
0.0707
0.0155
0.0909
0.0137

0.0561
0.0212
0.0662
0.0120
0.0319
0.0161
0.0030
0.0242
0.0095
0.0031
0.0016
0.1178
0.0588
0.0490
0.0157
0.0574
0.0111

Gagnon and Tryon
A3.

RMSD by Replication Across 20 Quarters
Rule 1: MA

GDP
AGDP
PGDP
APGDP
GDPV
MB
CAB
RS
ARS
RL
ARL
RERW
ARERW
C
AC
MBR
AMBR

GDP
AGDP
PGDP
APGDP
GDPV
MB
CAB
RS
ARS
RL
ARL
RERW
ARERW
C
AC
MBR
AMBR

TR1

TR2

TR3

TR4

TR5

TR6

TR7

TR8

0.0436
0.0173
0.0694
0.0081
0.0347
0.0205
0.0034
0.0307
0.0072
0.0028
0.0017
0.1824
0.0518
0.0285
0.0118
0.0465
0.0087

0.0302
0.0195
0.0355
0.0094
0.0176
0.0115
0.0018
0.0172
0.0084
0.0028
0.0016
0.0729
0.0432
0.0271
0.0139
0.0338
0.0095

0.0573
0.0221
0.0545
0.0098
0.0235
0.0126
0.0024
0.0189
0.0096
0.0033
0.0015
0.0575
0.0471
0.0526
0.0149
0.0547
0.0112

0.0291
0.0163
0.0470
0.0120
0.0316
0.0136
0.0018
0.0204
0.0100
0.0017
0.0014
0.0835
0.0512
0.0198
0.0155
0.0298
0.0097

0.0490
0.0233
0.0618
0.0107
0.0295
0.0106
0.0026
0.0159
0.0079
0.0023
0.0015
0.0911
0.0571
0.0569
0.0178
0.0641
0.0101

0.0285
0.0167
0.0381
0.0121
0.0212
0.0143
0.0032
0.0214
0.0081
0.0030
0.0016
0.0814
0.0552
0.0326
0.0136
0.0402
0.0110

0.1074
0.0194
0.1342
0.0117
0.0358
0.0188
0.0019
0.0282
0.0071
0.0026
0.0013
0.1150
0.0594
0.0912
0.0143
0.1144
0.0114

0.0314
0.0217
0.0107
0.0060
0.0259
0.0077
0.0021
0.0115
0.0062
0.0030
0.0019
0.1379
0.0485
0.0304
0.0118
0.0168
0.0063

TR 9

TR 10

TR 11

TR 12

TR 13

TR 14

TR 15

TR 16

0.0239
0.0244
0.0375
0.0112
0.0401
0.0135
0.0050
0.0202
0.0102
0.0022
0.0014
0.1354
0.0655
0.0408
0.0133
0.0373
0.0110

0.0538
0.0210
0.0727
0.0089
0.0333
0.0093
0.0043
0.0140
0.0107
0.0025
0.0018
0.1181
0.0732
0.0713
0.0186
0.0755
0.0087

0.0809
0.0183
0.0780
0.0110
0.0254
0.0145
0.0020
0.0218
0.0083
0.0036
0.0016
0.1447
0.0633
0.0631
0.0131
0.0697
0.0099

0.0447
0.0192
0.0404
0.0129
0.0310
0.0134
0.0033
0.0201
0.0114
0.0027
0.0015
0.0761
0.0559
0.0262
0.0112
0.0331
0.0080

0.0614
0.0144
0.0714
0.0179
0.0311
0.0173
0.0016
0.0260
0.0117
0.0048
0.0016
0.0689
0.0473
0.0415
0.0161
0.0591
0.0118

0.0346
0.0332
0.0309
0.0064
0.0374
0.0087
0.0032
0.0131
0.0091
0.0022
0.0015
0.2531
0.0794
0.0311
0.0184
0.0238
0.0097

0.0195
0.0224
0.0275
0.0108
0.0248
0.0127
0.0020
0.0191
0.0101
0.0022
0.0015
0.1003
0.0657
0.0244
0.0182
0.0230
0.0106

0.0606
0.0197
0.0754
0.0123
0.0257
0.0179
0.0016
0.0268
0.0090
0.0020
0.0013
0.0724
0.0571
0.0523
0.0145
0.0634
0.0127




Gagnon and Tryon
A3.

RMSD by Replication Across 20 Quarters (cont'd)
Rule 1: MA

GDP
AGDP
PGDP
^PGDP
CDPV
MB
CAB
RS
ARS
RL
ARL
RERW
ARERW
C
AC
MBR
AMBR




TR 17

TR 18

TR 19

TR 20

0.0258
0.0191
0.0290
0.0108
0.0319
0.0137
0.0026
0.0206
0.0114
0.0022
0.0016
0.0965
0.0585
0.0303
0.0157
0.0190
0.0086

0.0590
0.0236
0.0833
0.0201
0.0537
0.0298
0.0031
0.0447
0.0127
0.0045
0.0013
0.0832
0.0741
0.0479
0.0211
0.0579
0.0182

0.0665
0.0277
0.0701
0.0164
0.0243
0.0227
0.0052
0.0341
0.0082
0.0029
0.0016
0.0910
0.0621
0.0561
0.0182
0.0576
0.0164

0.1033
0.0167
0.1162
0.0124
0.0398
0.0217
0.0028
0.0326
0.0093
0.0053
0.0017
0.1186
0.0433
0.0774
0.0174
0.1046
0.0121

Avg
0.0561
0.0212
0.0662
0.0120
0.0319
0.0161
0.0030
0.0242
0.0095
0.0031
0.0016
0.1178
0.0588
0.0490
0.0157
0.0574
0.0111

Gagnon and Tryon
A4.

Difference in RMSD Between Rules 1 and 2 Across Replications

GDP
LGDP
PGUP
APGDP
GDPV
MB
CAB
RS
ARS
RL
ARL
RERW
ARERW
C
AC
MBR
AMBR

GDP
LGDP
PGDP
APGDP
GDPV
MB
CAB
RS
ARS
RL
ARL
RERW
ARERW
C
AC
MBR
AMBR

89 1

89 2

89 3

89 4

90 1

90 2

0.0024
0.0018
0.0001
0.0001
0.0027
-0.0079
0.0002
-0.0184
-0.0132
-0.0007
-0.0008
-0.0065
-0.0067
0.0006
0.0003
-0.0022
-0.0006

0.0012
0.0018
0.0003
0.0001
0.0023
-0.0088
0.0002
-0.0060
-0.0132
-0.0004
-0.0008
-0.0030
-0.0067
0.0004
0.0003
-0.0037
-0.0006

0.0020
0.0025
0.00030.0002
0.0037
-0.0114
0.0003
-0.0089
-0.0121
-0.0007
-0.0008
-0.0051
-0.0062
0.0009
0.0007
0.0013
0.0003

0.0044
0.0031
0.0004
0.0001
0.0063
-0.0170
0.0006
-0.0232
-0.0179
-0.0013
-0.0007
-0.0098
-0.0066
0.0020
0.0011
0.0020
-0.0009

0.0010
0.0019
0.0004
0.0001
0.0052
-0.0095
0.0002
-0.0158
-0.0207
-0.0008
-0.0010
-0.0045
-0.0066
0.0019
0.0004
0.0009
-0.0024

0.0004
0.0018
0.0003
0.0001
0.0049
-0.0074
-0.0000
-0.0099
-0.0156
-0.0005
-0.0007
-0.0018
-0.0056
0.0014
0.0004
-0.0010
-0.0006

91 1

91 2

91 3

91 4

92 1

92 2

92 3

92 4

-0.0002
0.0021
0.0003
0.0004
0.0058
-0.0059
0.0004
-0.0148
-0.0221
-0.0008
-0.0008
-0.0056
-0.0044
0.0005
0.0005
-0.0029
-0.0019

-0.0001
0.0026
0.0004
0.0004
0.0071
-0.0049
0.0003
-0.0138
-0.0208
-0.0010
-0.0009
-0.0038
-0.0085
-0.0001
0.0008
-0.0036
-0.0029

0.0002
0.0024
0.0005
0.0003
0.0085
-0.0098
0.0004
-0.0136
-0.0245
-0.0012
-0.0008
-0.0050
-0.0078
0.0000
0.0006
-0.0046
-0.0015

-0.0004
0.0014
0.0004
0.0001
0.0086
-0.0068
0.0000
-0.0104
-0.0153
-0.0007
-0.0006
-0.0001
-0.0052
-0.0002
0.0004
-0.0054
-0.0025

0.0002
0.0020
0.0001
0.0003
0.0084
-0.0052
0.0002
-0.0127
-0.0157
-0.0008
-0.0006
0.0005
-0.0063
-0.0002
0.0005
-0.0052
-0.0023

-0.0007
0.0019
-0.0002
0.0003
0.0084
-0.0040
0.0001
-0.0119
-0.0198
-0.0006
-0.0005
-0.0004
-0.0050
-0.0004
0.0004
-0.0072
-0.0019

-0.0008
0.0018
-0.0002
0.0005
0.0082
0.0003
-0.0002
-0.0160
-0.0206
-0.0009
-0.0005
-0.0007
-0.0043
-0.0008
0.0001
-0.0093
-0.0031

-0.0017
0.0026
-0.0000
0.0005
0.0077
-0.0010
-0.0000
-0.0062
-0.0204
-0.0005
-0.0010
0.0001
-0.0054
-0.0011
0.0005
-0.0111
0.0012




90 "*
0.0014
0.0019
0.0002
0.0002
0.0053
-0.0105
0.0002
-0.0165
-0.0152
-0.0008
-0.0007
-O.OO'-l
-0.0048
0.0012
0.0006
-0.0011
-0.0023

90 4
0.0016
0.0032
0.0002
0.0004
0.0058
-0.0083
0.0003
-0.0111
-0.0235
-0.0011
-0.0011
-0.0053
-0.0099
0.0010
0.0012
0.0003
-0.0007

Gagnon and Tryon
A4.

Difference in RKSD Betveen Rules 1 and 2 Across Replications (cont'd)

GDP
LGDP
PGDP
LPGDP
GDPV

MB
CAB
RS
&RS
RL

ARL
RERW
ARERW

C
AC
MBR
AMBR




93 1

93 2

93 3

93 4

-0.0017
0.0011
0.0002
0.0005
0.0094
-0.0014
0.0002
-0.0094
-0.0133
-0.0009
-0.0006
-0.0006
-0.0029
-0.0011
0.0003
-0.0120
-0.0019

-0.0024
0.0022
0.0006
0.0004
0.0115
-0.0046
0.0006
-0.0086
-0.0203
-0.0009
-0.0008
-0.0025
-0.0078
-0.0016
0.0005
-0.0140
-0.0007

-0.0015
0.0020
0.0007
0.0003
0.0132
-0.0062
0.0003
-0.0094
-0.0210
-0.0012
-0.0008
-0.0049
-0.0046
-0.0015
0.0002
-0.0118
-0.0018

-0.0018
0.0009
0.0008
0.0003
0.0137
-0.0070
0.0003
-0.0127
-0.0124
-0.0011
-0.0003
-0.0036
-0.0036
-0.0015
-0.0001
-0.0111
-0.0033

- 25 -

Avg
-0.0003
0.0021
0.0003
0.0003
0.0077
-0.0066
0.0002
-0.0121
-0.0181
-0.0008
-0.0008
-0.0031
-0.0059
-0.0002
0.0005
-0.0064
-0.0016

Gagnon and Tryon
A5.

Difference in RMSD Between Rules 1 and 2 Across Quarters

TR1

GDP
LGDP
PGDP
APGDP
GDPV
MB
CAB
RS
ARS
RL
ARL
RERW
ARERW
C
AC
MBR
AMBR

TR3

TR4

TR5

TR6

-0.0034 -0.0008
0.0020 -0.0007 -0.0011 -0.0008
0.0017
0.0021
0.0025
0.0015
0.0027
0.0020
0.0036
0.0001 -0.0012 -0.0012
0.0032
0.0018
0.0004
0.0004
0.0002
0.0002
0.0006
0.0005
0.0086
0.0029
0.0050
0.0053
0.0078
0.0044
0.0065 -0.0101 -0.0045 -0.0079 -0.0208 -0.0084
-0.0003
0.0005
0.0001 -0.0001 0.0008
0.0002
-0.0085 -0.0049 -0.0087 -0.0191 -0.0166 -0.0038
-0.0160 -0.0185 -0.0186 -0.0118 -0.0220 -0.0137
-0.0008 -0.0004 -0.0008 -0.0010 -0.0005 -0.0005
-0.0005 -0.0008 -0.0007 -0.0007 -0.0008 -0.0009
-0.0044 -0.0026 -0.0045 -0.0080 0.0002 -0.0033
-0.0044 -0.0053 -0.0053 -0.0047 -0.0079 -0.0055
-0.0018 -0.0006
0.0009
0.0004 -0.0012
0.0000
0.0005
0.0002
0.0006
0.0004
0.0008
0.0003
-0.0128 -0.0077
0.0021 -0.0169
0.0019
0.0007
-0.0006 -0.0006 -0.0026 -0.0059 -0.0039 -0.0015
TR 9

GDP
AGDP
PGDP
APGDP
GDPV
MB
CAB
RS
ARS
RL
ARL
RERW
ARERW
C
AC
MBR
AMBR

TR2

0.0014
0.0025
0.0054
0.0007
0.0142
-0.0047
0.0005
-0.0186
-0.0221
-0.0014
-0.0008
0.0045
-0.0062
-0.0029
0.0006
-0.0081
-0.0035




TR 10

TR 11

-0.0055
0.0027
0.0022
0.0010
0.0052 -0.0006
0.0005
0.0002
0.0115
0.0053
-0.0199
0.0038
0.0006 -0.0001
-0.0187 -0.0085
-0.0176 -0.0146
-0.0002 -0.0007
-0.0007 -0.0004
-0.0007
0.0034
-0.0052 -0.0040
-0.0030 0.0014
0.0005
0.0003
-0.0088
0.0031
-0.0019
0.0002

TR7

TR8

-0.0047
0.0017
0.0001
0.0001
0.0072
-0.0087
0.0004
-0.0147
-0.0168
-0.0004
-0.0006
-0.0031
-0.0038
-0.0015
0.0006
-0.0117
-0.0025

0.0062
0.0020
-0.0007
0.0003
0.0091
-0.0140
0.0000
-0.0138
-0.0195
-0.0009
-0.0007
-0.0038
-0.0034
0.0034
0.0005
0.0044
-0.0005

TR 12

TR 13

TR 14

TR 15

TR 16

0.0035
0.0018
0.0001
0.0006
0.0074
-0.0114
0.0005
-0.0152
-0.0184
-0.0010
-0.0009
-0.0051
-0.0067
0.0015
0.0001
0.0047
-0.0048

0.0000
0.0007
-0.0052
0.0005
0.0053
-0.0046
-0.0004
-0.0128
-0.0115
-0.0010
-0.0004
-0.0075
-0.0052
-0.0001
0.0003
-0.0232
-0.0027

0.0033
0.0040
0.0044
0.0003
0.0085
-0.0118
0.0005
-0.0303
-0.0360
-0.0014
-0.0014
-0.0076
-0.0116
0.0022
0.0014
-0.0035
-0.0008

0.0001
0.0028
0.0006
0.0003
0.0048
-0.0011
0.0002
-0.0109
-0.0215
-0.0009
-0.0010
-0.0046
-0.0087
-0.0001
0.0006
-0.0052
-0.0007

-0.0026
0.0014
0.0002
0.0002
0.0044
-0.0020
0.0002
-0.0052
-0.0152
-0.0007
-0.0006
-0.0014
-0.0019
-0.0012
0.0002
-0.0011
-0.0022

Gagnon and Tryon
A5.

Difference in RHSD Between Rules 1 and 2 Across Quarters (cont'd)

TR 17
GDP
LGDP
PGDP
LPGDP
GDPV

MB
CAB
RS
LRS
RL
ARL
RERW
ARERW

C
AC
MBR
AMBR




0.0025
0.0019
0.0014
0.0001
0.0083
0.0008
0.0003
-0.0148
-0.0144
-0.0010
-0.0008
-0.0051
-0.0061
0.0028
0.0007
-0.0117
0.0003

TR 18

TR 19

TR 20

Avg

0.0001
0.0027 -0.0014 -0.0003
0.0014
0.0011
0.0027
0.0021
0.0034 -0.0021 -0.0037
0.0003
0.0003
0.0005 -0.0004
0.0002
0.0176
0.0077
0.0006
0.0080
-0.0033 -0.0133
0.0080 -0.0066
0.0002
0.0004 -0.0002
0.0002
-0.0095 -0.0016 -0.0151 -0.0121
-0.0131 -0.0242 -0.0089 -0.0181
-0.0014 -0.0003 -0.0012 -0.0008
-0.0005 -0.0011 -0.0006 -0.0008
-0.0075 -0.0025 -0.0029 -0.0031
-0.0071 -0.0083 -0.0048 -0.0059
0.0001
0.0012 -0.0002 -0.0002
0.0008
0.0002
0.0006
0.0005
-0.0084
0.0143 -0.0207 -0.0064
-0.0003
0.0032 -0.0013 -0.0016

- 27 -

Gagnon and Tryon
REFERENCES

Blue Chip Economic Indicators

(Alexandria, VA: Capitol Publications)

various issues, 1980-1989.
Fuhrer, Jeff, and George Moore "Inflation Persistence," in this conference
series, 1992.
Gagnon, Joseph E. "A Forward-Looking Multicountry Model for Policy
Analysis: MX3," Economic and Financial

Computing

1, 1991, pp. 311-61.

Gagnon, Joseph E., and Ralph W. Tryon "Stochastic Behavior of the World
Economy under Alternative Policy Regimes," International Finance
Discussion Papers No. 428, March 1992.




- 28 -




COMMENTS ON PRICE AND OUTPUT STABILITY UNDER ALTERNATIVE MONETARY
POLICY RULES
Satyajit Chatterjee1

The goal of this paper is to use a medium sized econometric model
of the U.S., West Germany, and Japan, to evaluate the performance
of the U.S. economy under alternative monetary policy rules. This
model has been developed by the first author and discussed in
detail in Gagnon (1991).

Two key aspects of the model are that

expectation formation is forward looking and approximately
rational (in the sense that it approximately corresponds to the
predictions of the model) and that the long run properties of the
model match those of a standard real neoclassical growth model.
These features make the model attractive for monetary
policy analysis. Adherence to rational expectations means that
the policy analysis exercise is immune from the Lucas critique.
The fact that the long run properties of the model are those of a
real neoclassical growth model means that the short run dynamics
(for which monetary factors matter) do not extrapolate into
bizarre long run behavior.

In addition, the behavioral equations

of the model, while not grounded in explicit optimization, are
carefully motivated.

The reader gets a sense of the kind of

structure (in terms of preferences, technology and market
opportunities) that would generate these decision rules.

In what

follows I will not comment any further on the structure of the
model.
What I will focus on instead are the different monetary
policy rules considered in the paper and the manner in which they

Economist, Federal Reserve Bank of Philadelphia.

Chatterjee

are evaluated.

Let me begin with the latter point.

The authors

"evaluate" policy ruies in terms of their effect on output, price
and interest rate variability.

The presumption is that a rule

that generates more variability is inferior to one that generates
less.

However, a more appealing way to evaluate policy rules is

to determine how they affect the variability of utility.

From the

discussion in Gagnon (1991), it would appear that the authors have
in mind a situation where consumption goods and real money
balances are the only arguments in people's utility function.
Therefore, it is the variability in these quantities that ought to
matter.

Indeed, since the authors actually estimate the

parameters of the utility function they can evaluate the different
policy rules directly in terms of expected utility.

My sense is

that the ranking of rules would be significantly affected by this
choice of metric.

In particular, rules that smooth interest rates

might generate greater variability in real money balances and
hence be less desirable.
I turn now to the different policy rules evaluated in the
paper.

Presumably, the ultimate object of interest here is the

character of the optimal monetary policy rule given the structure
of the model.

However, due to its complexity it is

computationally infeasible (but not impossible) to calculate the
optimal policy rule for a given criterion function.

Instead, the

authors provide us with the operating characteristics for a
collection of reasonable looking rules.

While this is

understandable, we are left nevertheless with an uncomfortable
imbalance in the paper: while a lot of care has gone into
modelling individual decision rules as resembling the result of
some sort of optimization exercise, no attempt is made to model
the monetary policy rule as resembling the result of some sort of
an optimization exercise on part of the monetary authority.




2

Chatterjee

Consequently, 1 find it difficult to get interested in the
operating characteristics of the economy under any of these rules.
What could be done to alleviate this problem?

One

possibility, which I find personally attractive is to pose the
optimal monetary policy question in a model for which it is
computationally feasible to obtain an answer.

I am thinking about

fairly abstract general equilibrium monetary models like the
representative agent cash-in-advance models of the type
popularized by Lucas (Lucas (1984), Lucas and Stokey (1987)).
Recently, researchers (Cooley and Hansen (1989)) have used
calibrated versions of this model to obtain answers to question
like: what happens to the operating characteristics of the economy
if the monetary growth rate is raised from 3% to 6%? It is not too
difficult to extend this kind of analysis to compute optimal
feedback rules. We would then have a numerical candidate rule
which we know to be optimal for a simpler economy.

It would then

be of interest to see how this rule performs when it is used in an
econometric model of the kind that the authors have estimated and
which incorporates real world frictions absent from the simpler
model.
To summarize, I find the model estimated by the authors to
be reasonable and certainly worth taking seriously.

My sole

concern has to do with the manner in which the model is used.

I

would liked to have seen different policies ranked according to
expected utility (or failing that, at least in terms of
variability of consumption and real money balances).

I would also

liked to have seen some attempt at studying the operating
characteristics of an approximately optimal monetary policy rule.




3

Chatterjee

REFERENCES
Cooley Thomas F., Hansen Gary D., "The Inflation Tax in a Real
Business Cycle Model." American

Economic Review,

vol. 79,

(September 1989), pp. 733-48.
Gagnon, Joseph E., "A Forward Looking Multicountry Model for
Policy Analysis: MX3." Economic

and Financial

Computing,

vol

1, (1991), pp. 311-61.
Lucas, Robert E., Jr., "Money in a Theory of Finance."

Rochester

Conference Series

on Public Policy,

Carnegie

vol. 21,

(1984), pp. 9-46.
..-«.-». f Stokey Nancy L. , "Money and Interest in a Cash-in-Advance
Economy." Econometrica,




vol. 55, (May 1987), pp. 491-514.

4

MONETARY POLICY EXPERIMENTS
IN A STOCHASTIC OVERLAPPING GENERATIONS MODEL
OF THE TERM STRUCTURE
Steven Russell1

In recent years economists have begun to experiment with
the construction of dynamic, stochastic general equilibrium
models designed to confront the data provided by
macroeconomic time series — models that can explain, or
help explain, the relationships between and within various
series that constitute the "stylized facts" of the business
cycle. The exercise of data confrontation seems to take
place in two steps. First, an investigator specifies a
model, and chooses its parameters in a way that seems
empirically plausible. This step is sometimes called
"calibration." Next, the model is simulated, and the
properties of the artificial time series it generates are
compared to those of actual time series data, paying special
attention to the particular stylized facts emphasized by the
investigator. In practice there is usually some (and often
a great deal of) interaction between first and second steps:
parameter values are very often chosen with an eye towards
producing artificial data with the desired properties.
The model of choice for these sorts of exercises has
been the representative agent, infinite-horizon capital
accumulation model, augmented by positing stochastic
variation in technological productivity. The augmented model
has become known as the "real business cycle" (RBC) model.
RBC modeling has provided valuable insights into the nature

1

Federal Reserve Bank of St. Louis.
research assistance.




Lynn Dietrich provided

Russell

and sources of the business cycle, and is clearly a growth
industry among macroeconomists.
One common criticism of RBC models is that they
emphasize real sources of cyclical variation at the expense
of sources that are monetary in nature. Some RBC modelers
have attempted to respond to this criticism by introducing
money into the model and examining the effects of various
assumptions about monetary policy. Unfortunately, the model
does not seem well suited to this purpose. Since it is
devoid of the sorts of exchange frictions that are necessary
to provide a "natural" role for money, money demand must be
induced via ad hoc devices such as placing real balances in
the utility function or imposing cash-in-advance
constraints. One characteristic finding is that monetary
policy is entirely neutral in the long run, and relatively
ineffectual even in the short run. Results of this sort
have led critics to allege that RBC practitioners introduce
money into their models only in order to demonstrate its
unimportance.
One problem with "monetary" RBC models that has
attracted a good deal of attention has been their inability
to produce liquidity effects — their inability, that is,
to generate nominal interest rates that decline in response
to monetary injections. Recently, investigators such as
Fuerst (1992) and Christiano and Eichenbaum (1992) have
succeeded in constructing RBC models that produce liquidity
effects. While this is certainly a very interesting
development, a skeptic might view the scope and intricacy
of the assumptions these investigators must make in order to
achieve such effects as a testament to the limitations of
the RBC model as a framework for monetary analysis.
The most popular dynamic general equilibrium
alternative to the representative-agent, infinite horizon




-2-

Russell

model is the overlapping generations (OLG) model. The OLG
model has been, and for the most part remains, the model of
choice for theorists interested in monetary issues.
However, OLG modelers have rarely attempted to confront
business cycle data in anything like the sense that RBC
modelers attempt to do so. The principal reason for this is
probably the "time horizon question." RBC models can be
calibrated, by appropriate choice of the representative
agent's rate of time preference, so that a "period" seems to
represent an interval of a quarter or a year, and in
particular so that business-cycle-like variation occurs over
intervals that are short relative to the decision horizon of
the agent. This cannot be done with conventional OLG
models, since the agents that populate them live for only
two or three periods.
The relative length of agents' decision horizons may
not be the only (or best) criterion according to which the
empirical plausibility of a model can be evaluated, however.
Another criterion is whether the model is flexible enough to
capture the features of the economy that most economists
consider critical to understanding the phenomena under
study. When this criterion is applied, overlapping
generations models compare quite favorably to RBC models.
The generational heterogeneity of the population of agents
creates the potential for an active money market, and active
primary and secondary markets for government securities —
active in the sense that the agents in the model actually
trade these objects with other agents. In addition, it is
relatively easy to introduce the sorts of intragenerational
heterogeneity necessary to produce active private credit
markets. In OLG models, monetary injections can easily take
the form of open market purchases, rather than the
"helicopter drops" favored by RBC modelers. Partly as a
result, it is relatively easy to use these models to analyze




-3-

Russell

the interaction between fiscal and monetary policy: indeed,
realistic descriptions of open market operations virtually
require explicit consideration of this interaction.
Perhaps the most important difference between OLG and
RBC models is that the former pr